flow-general・🌊

[rough halo]

buddy. are you delusional?

[copper fog]

Understandable, I have had some success adding an image an it keeping my image in the new generation using the "+" icon, still hit and miss though

[rough halo]

they wont change it

anything you get out of it thats close to risky would be by accident.

besides, if you want that there's lots of options out there for it. weird that you'd go and try to get veo to do it.

[lunar root]

This is bad for people paying full price $300 just to get less than what they had without a big price drop for Ultra.

[rough halo]

thats what im saying. go look for other tools that wont block you.

instead of trying to make google loosen up. thats not happening

[pulsar coral]

You're right at least not around here, the proper solution would be to either re-enable original NB1 at Flow (which I've been telling them for weeks) or add a proper third model type for artistry with 18+ verification.

[pulsar coral]

Which yeah if delusional or not would be a solution.

[copper fog]

I agree and I meant that I was using frames and not ingredients

[rough halo]

LOL.. dude. you understand that this is the same company that will demonetize and/or delist your video on youtube if you say certain benign words?

[pulsar coral]

What's gonna happen it's that eventually the Chinese will catch up with their models and people will just stop paying here, it's happening with the video generation already.

[rough halo]

That doesnt concern google one bnit

[pulsar coral]

Once that happens they will wake up, people from Grok did.

[rough halo]

chinese will never be google. google isnt just an AI company

[pulsar coral]

idk man have you tried Kling or Wan

[rain ivy]

i don't use veo. i only mess around with image generation for fun. doesn't make anything worth while anymore. even the other day it blocked just the prompt "onion" for me while i was testing whether i was still rate limited

but yeah i doubt they'll even acknowledge it much less change it back. when has a company ever improved it's services?

[lunar root]

im shure it will conern them if enough ultra users drop off.

[rain ivy]

i wished they'd bring back nb1. nb2 is awful. even disregarding the sudden extreme censorship

[rough halo]

image gen or even text. if you are going for anything risky.. doing it through google's AI tools is just not the right place. there's TONs of alternatives.

hell if its something you think is mild, go try grok. its way looser

[lunar root]

i use them all to make my productions tight. if people are ok with getting less and paying the same is crazy

[pulsar coral]

It's still up for some weeks more at Whisk so you can get used to it, after that it'll only be plausible to use at Google AI studio or by making a canvas page with Gemini, Gemini canvas can reprodce this same site and Whisk itself, but takes some practice and tweaking

It'll also be at some other payed credit pages, but probably won't be worth the cost there, specially compared to how free Whisk is

[rough halo]

To go that far to generate content thats marginally goon material... you must be crazy. considering the easier alternatives out there.

[lunar root]

This is less. some already had less but now we all got it when 5/10 hits. for ultra at the cost of $300 per month

[pulsar coral]

My attempts are going like this atm, not all works and it's wonky, but the NB1 model is there for sure

[rough halo]

but your funeral. keep barking up the wrong tree and try messaging devs thinking they cant read right through your request. they will always play safe. if the tool blocks it, then the tool is right. because it protects them.

[rain ivy]

it isn't even that

even this is too much for it now apparently https://gyazo.com/9c4ad13bcaac20b7a192bad31c46bd6b. for reference this is what nb1 made for me in gemini when it still used that https://gyazo.com/ee815224e4e00c434a0a6aa90f08b298

[rough halo]

thats a child.. no wonder it blocks it

if you're looking at content that nearly hints at minors.. which is wrong in the first place, but even uncensored tools will block you.

[rain ivy]

how much of a prude would you have to be to think that should blocked? it rightfully never used to have a problem

[rough halo]

who are you talking to?

[fringe spindle]

I'm looking for volunteer Devs for a remote project (preferably USA or Europe/Switzerland, if anyone is interested)

[pulsar coral]

Yes we know Google is leaving the work to this kind of thing, we're with you on that don't worry. Let's keep centering more on them than on ourselves.

[rough halo]

CSAM is the most risky. no matter how "mild" you might think it is

the pic is clear. there is no doubt in my mind why its being blocked. and rightfully so

[rain ivy]

it's nothing of the sort. if you look at that and thing of sex that's on you. no sane person does

[rough halo]

oh its on me. right. that picture is of a grown fully mature mother of 9. im just bad in my brain

[pulsar coral]

idk someone might get turned on by piles of leeks, who are we to judge.

[rough halo]

minors is off limits

no matter what.

as it should be

[rain ivy]

"add more celery" https://gyazo.com/6a5fc0d6a0b06c4cdfe2dc2dd5221fd2

images like this shou;ldn't be off limits because some closeted pervert can't look at that without thinking of csam

[rough halo]

well. clearly its not only me thats the pervert. its google too. 🙂

[rain ivy]

exactly. and fortunately for you, it's them who decides the censorship. you'll sleep better at night knowing i can't make stupid pictures like that. too dangerous!

[rough halo]

all AI tools will block minors. to them, its playing with fire. your opinion = 0. as it should.

im doing you a favor by telling you why. you can get defensive and insult me all you want, dont know what that will get you, but sure.. take it out on the person thats telling you to set your expectations right instead of wasting your time

[rain ivy]

not here to insult people. like i said i came here to moan and see if anyone else had noticed it, or if google had said anything. even if they said it's intentional and it isn't changing, can't say i'd be happy, but at least i'd have an answer rather than be left in the dark

[rough halo]

look at it from their end. Lets say they allow your level of content. You may not know this, but AI bots are unreliable and dumb and easy to trick. Another one of their millions of users uses this permit to cause AI to make content that can get google into trouble. Was it worth it to please you? no. it wasnt. so they wont.

[rain ivy]

"my level", an anime girl jumping in a pile of celery

[rough halo]

You just wont get it.

Read it as you want. it states the reality.

Its not them being prudish. its them being risk-averse. logic

the potential trouble that can come from allowing it is not worth it.

[rain ivy]

yah someone like you might have a heart attack knowing i'm looking at an anime girl in a pile of celery (nano banana 2 edition). i already have the nb1 and imagen editions, it wasn't so prudish back then

[rough halo]

are you 12? do you even know how many prompt injections they face on a daily basis by users?

nvm you're just baiting and being intentionally dumb rn. im off

stop fixating on creating images of girls who can be interpreted as minors and trying to circumvent the rules. make your life easier.

[rain ivy]

alright bye lol

can you tell me why the image i posted shouldn't be able to be generated?

ok but can you tell me what's wrong with the image i posted? if you're going to bring up "images which should never be generated in the first place"? or did you change the topic to another image and i drank too much to realise we're talking about something different?

[lunar root]

for some reason no kids are renderable. i tried rendering a kid at a bus stop and it outright refused so i went with kling to get the shot.

[pulsar coral]

idk it's really been stated Nano Banana 2 is only good to make gingerbread cookies and tea parties with little antropomorphic furry animals now... whatever that's good for.

Doesn't seem to have probs when they are anthropomorphic chibi, which is suspicious from Google

[rain ivy]

well i even tried to generate 2 images with the prompt "onion" the other day to see if my daily generation limit had reset yet and one was blocked lol https://gyazo.com/8c36f22de2710eda4532d5dfec9c8d81

[lunar root]

Also not sure if anyone noticed...in nano banana if you edit an image miltiple times it wil continue to decrese quality and add a heavy color grain over the image and it gets worse with each render

[rain ivy]

ye quality degrades a lot with edits.... nb1 didn't do this. nb2 in general is just awful. i was willing to concede that it was better at scenes and backgrounds, and whilst that's still true, it's so censored now that i'd rather just have nb1 back and send nb2 to the bin. it's just entirelyt useless

[pulsar coral]

Maybe the word onion is copyrighted by this place. Yes NB2 is being that dumb now too, that's the actual prob.

[lunar root]

Really not liking no sound or ingredients in lite if we are losing fast with no ingredients. Ultra is looking really sad now. might have to make the jump soon. 😓

[pulsar coral]

Whisk doesn't have that prob because it has the "Seed" mode, and the seed mode surely is nothing but some layer of prompts you can reproce manually.

[rough halo]

there is no image editing tool where generation loss doesnt occur. not yet at least. thats why you should play around and edit and add until you get your finalized image then go back to the original and combine all of it into one singular prompt

[pulsar coral]

Yeah Whisks seed is an illusion too, but it sorta does it

[rough halo]

whisk would do it more clearly if you catch it in a chain of "dont change anything else" prompts

[pulsar coral]

Yeah you got it, that's pretty much how it works

[lunar root]

I work around it with photoshop taking the 2 images and blend together the change to get the result. Just odd a new render cant remake a render it already made. this is definately a bug. Photoshops nao banana doesnt do this at all.

[pulsar coral]

You gotta end your prompts in things alike.

[Complete image's quality: Keep everything true to reference image especially the facial structure, but all Professionally remastered in best known quality, HD ultra high resolution, sharpest definition, denoised, unblurry, clean lines, High contrast, fix any jpg artifacts, 4k detail, fine textures, clean edges. ultra detailed, sharp focus, high frequency detail, 8k detail]

[rough halo]

NB is just better at listening to you when you tell it not to change anything else. the nb1 that whisk used sometimes ignores those instructions. at least moreso than the new ones

[pulsar coral]

You can remove the "remastered" part if it's not working for some images. Can work shorter or short as azzzz said too. point is gotta try to do something.

[rough halo]

another better alternative is to use the marker in flow. draw a square or circle where you want to edit. that causes the quality to be maintained better. especially if your edit area is smaller

[pulsar coral]

The same chat models alike gemini and GPT also started reccommending doing all that

[rain ivy]

i've honestly noticed no difference when i do that. the best way with edits is to just do what you said and combine the original image and thed edited part in actual image editing software

[pulsar coral]

It might depend on where the image is losing quality too, that's why I clearly add face there

But yeah it really ain't as easy as it seems, whatever Whisk is actually doing with that

[rough halo]

thats what i do, and if all else fails i do a sloppy job with photoshop, upload it, and make NB do the heavy lifting and fix it with my idea in mind

doesnt have to be photoshop even. crop and paste into MS paint and orient things how you want them. then let NB blend it all

when i really want characters to remain consistent and i have issues with it, i do that. it works. its more work. but for my line of work its better for me

[narrow pilot]

if you want characters to be consitent please check out https://newcharactersheet.com lets you easily create persistent characters with a prompt or ref images

[lunar root]

i make character sheets, but for some reason Veo still manages to miss some details sometimes unlike Kling the images and faces stay exact on renders. i have had to delfect to Kling form time to time as a workaround.

[rough halo]

to be very honest with you, ive had better luck using smaller models for faces in some of my projects. the high fidelity and extra 4K quality rendering sometimes changes how a character appears entirely, and i find a lower (but still very good) quality to be more consistent. i think maybe its the new tools focusing too much on lighting and ultra 4k etc. that they leave less for the details

[narrow pilot]

with veo its beter to take advantage of batch generation esp if your injecting character sheets every scene

[rough halo]

but thats just a theory

seedance for example replicates some of my characters way better than veo

NB^

[lunar root]

yes indeeed... You have to. sometimes takes 10+ batch generations just to get one clip to work. This is why Veo fast worked but now without sound that sucks. if you need sound to produce a short film those credits are gonna go fast trying to get what you need. And now with the addition of charging you for ingredients is wild.

also noticed the discord button just dissapeard after i used it to come here lol. they are just doing anything now. features just come and go with no announcement out of fear of backlash.

[pulsar coral]

And for worse they activate them friday leaving them as they are without any testing then don't come back until Monday evening.

That'd be fine if they'd leave them as they are for people to give them lhem feedback about it, but it's been the same complaints over and over for over a month and they haven't listened to more than a couple of things.

Actually, I think just one, which was the aspect ratios, other than that it's one new unrequested update next to the other which only breaks things.

[crisp basalt]

what

[pulsar coral]

Maybe it was some message that went to their recycle bin and we can't see now.

Staff can see those, we can't

Use the modmail in blue from above the user list. Something does seem a little fishy here.

Discord chats also have a moderation log that admins can see, to check who took the actions.

At the server settings

Seems some other accounts got banned today too. Well you alll better take that to the modmail or the general chat, derailing the Flow topic might only make it worse.

[lunar root]

like i said the discord button magically disappeared from the top of flows page. it was there a few hours ago when i got the update message.

[west laurel]

any solution for this?

[pulsar coral]

Anyway, to all banned people remember it's the weekend so staff might take some more time to check what's going, but they'll be on it I guess.

[pulsar coral]

That's spreading fast but doesn't happen to me yet tbh, wait for some hours and when it works again don't try to generate more than 12 images at once, wait till even the last one finishes generating or failing to generate, then wait at least 1 minute before generating 12 more at most and so on. Just don't rapid click again.

Don't constantly refresh your browser window either, refresh it as less possible, that surely also triggers Google's recaptcha security.
Because I imagine people refresh their windows to clear the failed attempts.

Helps to tab to somewhere else while the images generate so your finger doesn't want to click regenerate. This has turned into compulsive gacha for some people and all the new censorship didn't help.

Also it seems some people also fix it by clearing cookies or trying Flow from an incognito window.

Tho that last one doesn't work for everybody

[wise locust]

Anyone getting the "You're requesting generations too quickly" error?

[rain ivy]

that happens to me either when i'm spamming generate too much or i've reached the daily limit. don't think i've had it otherwise

[wise locust]

If it's because of the former, how long do I have to wait?

If you know that is

[rain ivy]

not sure

tbh when i've hit the daily limit it's usually when i'm about to go to sleep anyway and then it won't be until at least few hours after waking up before i'd try anything again anyway

[pulsar coral]

Yeah if you run into that error just after 5 minutes even, you more likely just ran out of image generations for the day for that model, Flow doesn't display the correct message for it, should be saying something like "You ran out of image generations for today, come back tomorrow 🍌 "

Check if you were using Nano Banana Pro and if so try swapping to Nano Banana 2 and vice versa, each model might have some more images left, if both are dry just come back tomorrow around the same time you successfully generated an image and it will work again.

Wont help to check in just 4 or 5 hours, needs the whole day.

[rain ivy]

yeah threw me off a lot when they replaced the "out of daily generations" message lol. bit of an oversight at least

[pulsar coral]

Imagen 4 will also run out of images at the same time, since those Nano Banana models also run as a version of that one or something. All this goes the same for if the upscaling also fails since it also runs on Imagen 4.

Another option is going to the Gemini website, your same Google account from Flow links to the same account there, which is a good option specially if you're paying for a plan. You can get some more images there.

And if you prompt it to generate like 4 separate images at once it can also do it, or try to make the website you want with canvas mode, those can also make Nano images and Veo videos.

I've gotten surprisingly similar websites to Flow and Whisk there, helps to tell Gemini's canvas exactly to do that.

[rose vigil]

what is this? I subscribed for one year because unlimited generation. does this mean no more unlimited generation?

[tepid zephyr]

#flow-general・🌊

Guys please review my video, what is wrong with my channel, it’s not getting any views. Is it the editing? Wrong thumbnail? Or just bad AI quality?

https://youtu.be/M6lQf3ijNTg?si=5ceEiWrkm4pL9K1Z

[lunar root]

played it here....looks good. just keep going. it will take a bit for play-thru plays to count on youtube.

[tepid zephyr]

@DBot thank you for watching mate, it’s really annoying ti thinking youtube hates new comer like me and just stop recommending my video… 😓

[lunar root]

same here..... It means veo 3.1-Fast (Lower priority) is going away on May 10th 😓 😡 and now you will soon be stuck with a lower quality version veo 3.1-Lite (Lower priority) with no ingredients at 720 which is a no go for me. So it's good news for people who never had it but bad news for Ultra users who used Veo 3.1 Fast to get better results with sound.

Impact on Ultra Users: While 3.1-Fast was previously available at low or no cost (0 credits) for Ultra, it is increasingly being updated to cost 10 credits per generation, or is being replaced by lower-tier alternatives, frustrating users who relied on it for higher-quality, faster output. 👀

[tepid zephyr]

@rose vigil this happened to me too

@near gate comfyui is a great tool so far… free tier is still useful too

[tired sedge]

Is there anyway to make the style stay consistent on flow? I have a character and I often need it to be in different scenarios but I get stuck for hours because it doesn’t get it right

[lunar root]

make character sheets with all angles of your character. it works better but sometimes gets it wrong. using ingredients was good for character sheets under Veo 3.1 Fast but now they are phazing it out with the new Veo 3.1 Lite so not good and sad news for those who depended on it to keep charcters looking the same. It's a big step backwards for a lot of us.

[tired sedge]

Yeah that’s bad news, they are even removing whisk which I used most and it was way accurate than flow mine is image generations so whisk used to understand that way better. Flow is just wasting hours and still doesnt get it right

[lone notch]

maybe use reference image from time to time? the nature of llms causes them to drift overtime, even if you explicity tell it not to

[tired sedge]

Yes I attach clear reference each time yet it messes it up

[clear glacier]

frame is still not working ?

[modern dragon]

How to use the application

[exotic shell]

seems like most people havent noticed yet, that their ultra subscriptions soon are useless. else there would be riot

https://tenor.com/view/peepo-peepo-riot-riot-protest-peepo-angry-gif-21768097

[lone notch]

wdym?

[exotic shell]

veo 3.1 low priority will be deleted. the only free generation will be veo 3.1 flash

[jagged marsh]

Also the whole "You're requesting generations too quickly. Please wait a moment and try again." is a fvcking joke. What's the point of paying $$$ for Pro/Ultra if every few generations I get "yOu'Re rEqUeStiNg gEnErAtIoNs tOo QuIcKlY. PLeaSe wAiT a MoMenT aNd TrY AgAiN" which forces me to have to use Banana 2 which is not good enough anymore visually?

[lone notch]

roughly half of my generations "violate policy" -all SFW, the ones that don't get refused aren't what I'm looking for, so I regenerate and get hit by "too quicky"
I'm starting to question the whole premise - if I'm only allowed to generate what google deems acceptable (again, I do no spicy stuff, not even remotely close) then what's the point of the limit - just tell me what's unacceptable so I can try something else that is acceptable for me -otherwise if I have to choose to take what's acceptable to google but not to me - why the would I even pay for that?
"please pay us money to generate the content WE like"

[neon raven]

what? the lite will still be available

[exotic shell]

call it lite then.

[vernal remnant]

wait wait wait, are they removing the ingredients to video in fast mode ? because ingredients to video cannot be used in quality anyway ?
are you referring to the fast low priority mode here, or just general removal of fast mode for ingredients to video ?

[tepid zephyr]

#flow-general・🌊 #flow-videos・📹

Hey guys, please review my AI Video, tell me what is wrong with my video, it’s not getting any views

https://youtu.be/obLwwCB3P1M?si=CVA42EXjjUi9CVIY

[lone notch]

harley darvison ahhahaha 😄

[crisp basalt]

it’s only the low-priority model getting removed for Ingredients-to-Video. swear to god, the number of people here thinking the Flow team is killing the fast model entirely… that needs to be studied

[lunar root]

Lite = No Ingredients: it is optimized for speed and cost over high-fidelity detail. It does not possess the same capability as the standard/pro models to handle complex "ingredients" (structured inputs) efficiently. THIS IS BAD 😡 This was a 0 credits for Ultra users to use ingredients. now you have to pay for it in the only option available.

[crisp basalt]

no sound? where did u get source for that

[lunar root]

Swear all you want. FREE turned into a FEE for people paying full price already.

[lunar root]

i asked GOOGLE lol

[exotic shell]

@crisp basalt you deleted some messages. so much about that. 😄

i dont get why @crisp basalt is so keen on getting his paid service reduced by 90%, while price stays the same.

[crisp basalt]

here we go again

[full flicker]

Hey everyone, a reminder to be respectful and kind to everyone. Please move on from this topic!

[exotic shell]

so we are not allowed to talk about the fact that google changes the subscription to our disadvantage? that is weird 😄

[patent fulcrum]

Need b-roll. Got bored watching the same gestures and same zoom and back action

[lunar root]

i'll leave him alone. he asked me and i rsponded. that is all for that. Now back to the real point of Ultra users having to deal with now paying to use ingredients when thats the one option that was working well in Fast 3.1 for 0 credits. sometimes it takes multiple trys to get it right. now it will drain credits while doing it

[crisp basalt]

https://tenor.com/view/grass-gif-6112271

[tepid zephyr]

@patent fulcrum b-roll… hmm i wonder how I can generate good b-roll, I can’t come out with good prompts

[crisp basalt]

uhh honestly I’d say start by watching some real b-roll examples first. not everything is about writing some perfect prompt half of it is just understanding pacing, camera movement, and what kind of filler shots actually feel natural. once you get that, your prompts get way easier because you’re not guessing anymore, you’re just describing what you’ve already seen.

[patent fulcrum]

Just ask gemini to take your script, put it into tabular format, and for each sentence, generate a prompt for b-roll video to illustrate the idea. Table should have the time stamp, sentence, b-roll prompt, goal (what emotion are you trying to evoke) and then priority... meaning if you can't generate all of them, which ones will be the most effective (value between 1 and 3, where the highest impact broll would be generated first). Play with it and see what happens. Your sound is awesome, just the visuals are lacking.

[rough halo]

what does this mean?

clicking on help centerr only takes me to the front page of the help center. nothing directing me to the issue itself

[small mirage]

We all got this issue, probably a server issue. Just wait for some hours then try to generate. Don't click repeatedly, give some time between generations.

[lunar root]

you can try to report the bug in "⁠ labs-bugs-feedback・you can use the "Flow" tag to search reported posts and issues. "

[neat lava]

It make veo 3.1? Or more editing?

[small mirage]

probably not helpful... as long as google fix it. They already know we are facing this problem

[lunar root]

You might be right....but that was from a direct message from the help team pinned in upper corner. just a place to see who all has the same issue and know when its over. sometines an error can be just one persons account that needs a reset.

[rough halo]

is the extend button not working?

i keep clicking it nothing hapens

[cold totem]

I can’t seem to upscale any images to 2k with nano banana pro in flow, keeps saying failed, has either anyone had the same issue or can anyone test it?

[rough halo]

the extend tool is not working.

[old bay]

been broken for a while, grab a still of the last frame and update the prompt then try stitching in edit.

[merry sapphire]

Hi! I have Ultra plan but all my videos
have "Veo" watermark. Is this a bug?
How can I remove it? Thanks

[mighty ember]

Anybody Know Jesuscrist?

[lunar root]

"Google Flow Ultra users are seeing watermarks on Veo video generations primarily due to a temporary synchronization bug following a platform update, which causes the system to fail to recognize their premium status." Hopefully this will be fixed after the weekend. you can try to post the bug under "Watermark Regression: Ultra Tier Sync Failure"

sure pal. i test them all anyways lol

[narrow pilot]

I sent a fr!

[crisp basalt]

what country r u from

[crisp basalt]

lol no.

[crisp basalt]

google hasn’t really released an official list of affected countries yet, but from what’s been confirmed so far, south korea is definitely one of them

[stone garden]

Any news on this thing?

[lunar root]

this is from google itself so LOL yes. please stop respoding to my comments

[crisp basalt]

you replied to me too lol. if it’s from google, just drop the source next time no need to get worked up over it

ask google how to spell google cuz u needed that

[lunar root]

That what the "" was for... please stop talking to me before i make you run to cry to the admin again. unless you work for flow you are in the same boat and have no clue whats going on till they tell us. im going to the source

[upbeat sierra]

We hear you on Ingredients with Lite. Our team wants to ensure continuity with creators workflows, so am hopeful we'll find a resolution that works for everyone.

As for Veo Lite, would really encourage the community to try the latest Veo checkpoint, it's a strong model and we've been very impressed by the performance.

We value the community's feedback as always, please keep the feedback coming.

I've seen you mention 'no sound' a couple of times. Veo Lite generates Audio, so not sure what your issue is.

Please start a thread in #1238621511401734254 with details and we can look into this (prompt used, error received, screenshot etc)

[crisp basalt]

i can tell it’s a different kind of model when using the Veo 3.1 Lite especially on extension , the workflow feels like proper extension now, not just image-to-video. it actually continues the video instead of just relying on a single end frame to generate from.

or maybe im just being delulu

also ,, veo 3.1 lite still doesnt ... produce "stitched" extended video on my end

[pulsar coral]

Are we done with the vid stuff, I'm lazy there.

[pulsar coral]

Upscaling amount of images has different limits according to payed plan tiers. But surely if it isn't working anymore atm you just ran out of upscaling daily quota, it also runs on imagen 4 model same as the Nano Banana models, so it will work again the next day around the same time you successfully upscaled something today,

So if you got a lot of upscaling to do might wanna start by upscaling all you need the next day instead of going straight to generating images. That or gotta pay a better plan.

[cold totem]

So turns out it was literally just one batch of images that weren’t upscaling, old ones worked and new ones after them worked, kinda strange as I hadn’t done many today 🤷‍♂️ works now anyway haha

[pulsar coral]

idk it does also work wierdly sometimes too.

If all started working again, maybe it did just reset itself, there doesn't seem to have a set hour of the day to reset.

Another option is to download the images as they are in 1K and create an 2K or even 4K image upscaler at Gemini's canvas mode, it's the exact same type of imagen 3 upscaling anyway.

Really had to do that when in a hurry tho the prompt needs to be REALLY specific explaining it everything in detail... which I'm not perfect at but yeah. Anyway let's run through it since it's Sunday here. Ok here's reporter girl.

1536x2752

Gemini canvas mode prompt: "Create a batch 4k upscaler that uses Imagen 3 to upscale, it should allow me to upload any amount of images and allows me to download them in literal 4K resolution, it must have a working 4K resolution download button able to select to download them in either jpg or png exactly."

What sorcery is this, that's what I got, that's a free account atm

This is 3840 x 2160 tho I think Discord can't display it right.

The prompt can also accept other enhancements like clean lines, no bluriness, remove jpg artifacts etc

Oh yeah I forgot the upscalers seem to have this bug where if images are upscaled they download with a wrong weird extension, this happens at both Flow and Gemini, another user and me did ping some staff about it, so might wanna add to the prompt to allow to download them in either jpg or png exactly, tho I dunno if that corrects it yet. People just fix it by resaving them manually with an image editor, just renaming the extension doesn't fix it.

[dusty bridge]

SCENE LOCK: static tripod camera, vertical 9:16 framing, wide 24mm lens, eye-level height, centered on luxury mansion facade in background; fixed landmarks include mansion front door, symmetrical windows, driveway curve, and central lawn area; bright natural daylight with soft shadows, consistent lighting direction
STAGE: pristine empty lawn before construction, freshly cut grass, no structures present
DETAILS: subtle grass texture variations, faint mowing lines, slight soil patches, natural sunlight reflections on mansion windows, clean driveway edges, light breeze effect on grass tips
NEGATIVE: no text, no logos, no watermarks, no warped geometry, no new structures, no floating objects, no construction materials, no workers, no lighting inconsistency

[hazy acorn]

What is the deal with all the "This generation might violate our policies. Please try a different prompt or send" responses??? I am not doing anything that could remotely be considered in violation. See this screenshot for example. Flow has become unusable.

[pulsar coral]

That man does look like an older pierce brosnan, that surely is triggering the filters tho it's not intended from you. Even the font from that screen above him seems to have been taken from a James Bond movie.

To "help you" Nano Banana 2 looks for references that look like what you're doing right now and try to complete the image accurately with them, even tho almost no one asks for it to do that. And in doing that it often trips on itself.

The original Nano Banana wouldn't know how to do that, so it happens to it less or not at all with Whisk's precise mode.

[exotic shell]

me and many others would still be happy if veo 3.1 low priority stays. we would like to keep the same quality.

[minor badger]

Is there a channel here to request a feature?

[rain ivy]

there's a "feedback/requests" tag in the labs-bugs-feedback channel

[minor badger]

Thanks!

[void ledge]

I'm new to Flow. Is there a way to see which prompt each image was generated from, like in Whisk?

[dry frost]

Click the image in the project to take you to a scenebuilder type view and on the right will be the prompt along with all the alterations to date.

[void ledge]

In Whisk, you also request edits, but in the final results you receive a prompt for the entire image.

[pulsar coral]

Yeah that's not what Whisk does, it writes a whole prompt for each image you upload or create. It's not really as useful as people think, but it does help you to learn.

And no Flow does not have this option at all, in fact has none of the image models from Whisk, none of the options either, and all of the problems Whisk doesn't have.

[void ledge]

I entered the query: cyberpunk city
I received a result with the full prompt

[pulsar coral]

Flow has all of the censorship Whisk doesn't have too. Abandon all hope, ye who enter here.

Also notice how did Whisk recognize that as Pierce Brosnan too.

[void ledge]

Whisk is good at crafting prompts. The prompt varies with each result, but the keywords from your prompt are retained for future use.

[rain ivy]

they made a post about "policy violations" in the feedback channel. if you don't like the recent censorship changes you should respond to it too

[pulsar coral]

I don't really think that's the exact prob tbh, the way the model's consistency was destroyed by whoever tampered with that is visibly much more of a bigger issue, like with Pierce Brosnan example up there.

That shows it's not only about censorship but more about "I got enough money so I can do what I want-ery" with this model from Google.

[rain ivy]

well i don't think you have to make it specifically about that prompt example. just the issue in general with how they've made it so much more harsh recently. it's worth responding

[pulsar coral]

Yeah but yesterday some others did also make it about the censorship, I don't think they've noticed how much more it goes beyond that. Something really got borked with it.

Like I'm trying to make ocean beach water at it's shore, which should be normal... but it instead places all the beach's water behind some sort of invisible glass.

And that only happens when I say the water has some sea life.

I don't even say "inside" or anything that would make it understand it like that, and ruins some beautiful prompts.

[void ledge]

Could you give me an example?

[pulsar coral]

Yeah I think I got something

[potent patrol]

Anyone else having trouble with upscaling?

[pulsar coral]

This'd be when it works normally before the "Updates"

[rain ivy]

don't think i've noticed something like that

sounds like two separate issues unless it's sudden "weirdness" like that is also causing it to misinterpret prompts and therefore block them. because when i spam the prompt 20 times and get 2 outputs, those 2 don't look much different to how they did before they increased the censorship

but regardless, the censorship has increased recently by a lot. so you should still reply to that post

and maybe make a second one about this issue

[pulsar coral]

This is when I add "Sea life" cutting it a bit to avoid snowflakes just in case

Hold on I'll show the glass things that happen now

I've tried everything to prompt against it, or just avoid it without negatives, it just wont work. Also notice how cartoony it looks, I've also had a fixed art style for it like in the two first ones above.

It's like some kid with crayons is drawing the images now, like I said earlier.

All that I've found can be done about it is to generate at random and hope it will get how things should work.

It's like the model got brainwashed by disney chanel

[void ledge]

Yes. It feels as though she’s standing in front of a large aquarium.

[pulsar coral]

And that did really cut on my week's budget, I was supposed to make those

[void ledge]

It’s as if someone’s switched on nursery mode

[rain ivy]

i just tried "anime girl in swimming costume walking down the beach. she's looking back at us. water has some sea life" and the output looks more like the first one https://gyazo.com/7bd00c6f46c66d616e5eea8d15e9fd19 does look a bit more like they're floating though lol but i'm assuming your prompt was more detailed

strange that your output looks.... as crap as that though lol. i do see what you mean

but like i said maybe whatever they changed about the model's prompt interpretation is also what "inadvertently" increased the censorship. my prompts which do eventually get past the censorship look basically the same as they did before they increased the blockages. i still think it's worth responding to that post

or make your own post about your issue. or both lol

[pulsar coral]

Yeah that also happens, the animals sometimes float unnaturally.

Tha'ts not how it looks in the second one I showed, I had no probs with that before.

[rain ivy]

yah these ones are definitely just hovering lol https://gyazo.com/b8ad7649c1d2ef9e83f59f77d79cbfb3

of the 4 outputs one was blocked too for the recent overly prudishness too lol

[void ledge]

A similar issue cropped up in the models managed by MS. But it wasn’t fixed. A completely new prompt had to be created. Perhaps you could upload your work to Gemini and ask for a detailed prompt?

[pulsar coral]

Yeah this is what actual water looked like before.

It's just awful now

[void ledge]

Can I have a prompt? I’ll check it too.

[tepid zephyr]

@patent fulcrum Thank you for kindly shared input, i’ll try your way

[pulsar coral]

Nah I'm sure I removed it all and tried it back forth, added all sorts Light distortion, refraction and water realism prompts, it just makes it worse.

Because it seems to then associate them with refraction from glass

[tepid zephyr]

@crisp basalt oh great suggestion, haven’t thought of that, so I just copy the video and ask gemini how to recreate it like @patent fulcrum said, thank you for the kind advice guys

[tepid zephyr]

@lone notch 😅

[pulsar coral]

And this was last week

[rain ivy]

well you should still make a post in the feedback channel about it. just with a load of examples like you showed here and when you noticed it became useless like that

maybe copy & paste it into the "send app feedback" form on flow every now and then too lol

and respond to the censorship one lol

feels like a waste of time i know but nothing will ever change sitting around in this channel complaining about it

[pulsar coral]

Nah I know who to send this to and it's not them here, it's better to send it directly to Gemini support

They have a board for the model there

[crisp basalt]

I do that all the time ahaha just tell Gemini to uhh convert this video to text to video prompt something like that

[pulsar coral]

But thing is they broke it last week and that was very irresponsible

[rain ivy]

oh really? how do i contact them on there? i'd like to send some feedback there too lol

but it is strange. i don't think it's at flow's end, i think they mucked something up with nb2 in general as using it in gemini is a waste of time too https://gyazo.com/b9d5d37a741712c6a1545fbf848de7ec

though i don't know how it all works behind the scenes and our ui commuicates with the models etc

[void ledge]

Which model is having the problem?

[pulsar coral]

Once you send feedback there the whole board opens, one can send a detailed post in those

[pulsar coral]

Nano Banana 2 and pro in general, bottom line is tho, this is just one thing, there's actually tons of problems now

There's also way worse anatomical consistency, huge heads, cartoony small bodies

I had all solved that by prompts last week without probs but now they made it much worse somehow

[void ledge]

My test NB2

[pulsar coral]

Yeah you got it much better there. But after a while you will see some bad ones too

[rain ivy]

oh that's just the send feedback form. yeah waste of time really i think. i do have a few walls of text i copy & paste into it every now and then though, but i doubt anything will change. better than doing nothing though

that said i did just google it and found this https://support.gemini.com/hc/en-gb/articles/209113466-I-have-a-suggestion-for-a-new-feature-or-feedback-regarding-an-existing-feature

don't know if it actually goes to a different place but i ought to spend a moment of time writing walls of text into it

actually is that even google? crap lol

[pulsar coral]

idk I've indeed seen them responding there... unlike here

Everyone knows discords are just made to play anyway

[rain ivy]

looks like google staff actually responds to a few of them in the feedback channel

probably gets more attention than just using the send feedback form which probably just goes straight in the bin, assuming it even goes anywhere

[pulsar coral]

Up to here wasn't so bad either, at least they don't look like bath toys. It will just do it when it wants

[tepid zephyr]

@crisp basalt I guess I’m just need to think simpler. LOL

[minor bridge]

Hey will the google ai ultra plan let me upscale my flow omages to 4k?

[sterile garden]

https://www.youtube.com/watch?v=G85pGajUbGQ

[brittle hearth]

yes

image and video

[desert grove]

hi, im getting this error after every 15 - 20 images i generate and have to clear cookies for it to work again

[pulsar coral]

Veo's around midtier atm, been trying Kling but I do think Wan's new version is much better tho you can only make one vid daily for free or 3 if you have patience for a 10 hour queue or so.. most than all cause decent amount of NSFW allowed, not everything.

More than all it accepts my encased style of prompting just as well as VEO

[pulsar coral]

Yes clearing cookies and incognito browser window help sometimes. When it works again don't try to generate more than 12 images at once, wait till even the last one finishes generating or failing to generate, then wait at least 1 minute before generating 12 more at most and so on. Just don't rapid click again.

Don't constantly refresh your browser window either, refresh it as less possible, that surely also triggers Google's recaptcha security.
Because I imagine people refresh their windows to clear the failed attempts.

Helps to tab to somewhere else while the images generate so your finger doesn't want to click regenerate. This has turned into compulsive gacha for some people and all the new censorship didn't help.

F the censorship tho, I can bypass that just as always, my real problem like said earlier is it lost a lot of ability to understand how physics work,

idk maybe someone copyrighted water or something, Disney probably since the little mermaid swims in the ocean

[wise flame]

I wanted to create a flow video animations for kids (playing in garden or something like that) but Google doesn’t let me do it because they are minors

• anyone has a way to make it work or what did they do alternatively?

[winged plinth]

What's your prompt?

[winged plinth]

It can handle it natively without issues, might just be the inputs [prompt or image].

[wise flame]

I’m trying to create a realistic person not animated one. Simple uploading an image of teen playing football is not allowed

[gloomy verge]

You want create image by prompt, or add as reference image? I tried it as prompt and it created image for me.

[winged plinth]

Realistic minors [children and teenagers] are an industry wide standard restrictions across almost all platforms.

[gloomy verge]

Left image is by Imagen 4, right image is by NB2,

[shy hatch]

can you share the prompt

[gloomy verge]

Same as he wrote. "Teen playing football".

[shy hatch]

ohh really cool

[gloomy verge]

But, it created me only 2 images, because all other images I have as problem, same as other users. So, it create only 2 images from 9 images. All other 7 images has this problem.

[shy hatch]

im new to this ai things , i barely understand the things you guys say

[wraith niche]

Buddy, you again getting problems ??

[gloomy verge]

When you create image in this times, all or lot of results will be this problem. Sometimes image is generated correct.

[wise flame]

That’s for me

[gloomy verge]

Yes, all of time. sometimes one-two images are created, buit a lot of this problem is still.

[gloomy verge]

Yes, because image generator disallow generate images of kids. Person must be adult.

[gloomy verge]

Why are you asking? Do you have it too?

[wraith niche]

noo..

[gloomy verge]

Interesting...

[wraith niche]

I'm not getting any problems

how many images you are generating at a time?

[gloomy verge]

I have still same problem. 🙁 Sometimes a picture is created, but this problem is more often than created images.

[wraith niche]

i have created these 12 different images without facing error

[gloomy verge]

By NB2 or Imagen 4?

[wraith niche]

flow

[gloomy verge]

Yes, but model you used NB2 or Imagen?

[wraith niche]

nb2

1st image is from NB2 and second one is from Imagen

[gloomy verge]

I don´t understand it. Now it created all images in NB2.

[wraith niche]

this image generated by using imagen model.

I'm not getting what u wanna say

[gloomy verge]

Now, I tried "Christmas scene: Christmas dinner" and 8th image is with error.

[wraith niche]

let me try

[gloomy verge]

7th, sorry.

Now again. 9th image is error.

[wraith niche]

wait, let me try

this time not the perfect one.

but without any error

[gloomy verge]

Now I tried create 6 images with Imagen 4 and 4th and 6th is error.

[wraith niche]

bad luck continues for u buddy 😆

[gloomy verge]

Can it be by internet browser? Or browser haven´t influence for it?

[wraith niche]

may be the internet

i think in india peoples are not getting so much error in generating images like this

[gloomy verge]

Okay, I tried it on other internet. But, when it created errors too, so I don´t know, where can be problem.

[small mirage]

Nope, this is happening randomly. I am from Bangladesh and the same thing happened with me but my friend got no issue, even tho we use the SAME INTERNET

I am facing this issue from this week, last week I haven't seen any kind of issue like this one

[topaz lichen]

hi, anyone got trick. how to generate 0,2 seconds video (6 frame) with AI (i got first scene and last scene). any website/google feature or must use local computer? i also think the name is frame interpolation

[deft osprey]

the whole "You're requesting generations too quickly. Please wait a moment and try again." is a fvcking joke. What's the point of paying $$$ for Pro/Ultra if every few generations I get "yOu'Re rEqUeStiNg gEnErAtIoNs tOo QuIcKlY. PLeaSe wAiT a MoMenT aNd TrY AgAiN" which forces me to have to use Banana 2 which is not good enough anymore visually?

[violet ruin]

I'm trying to make another image and I want to make it as high quality as an opening first frame image. Can you please help me?

I know you said to use Nano Banana or Nano Banana Pro 2 and you mentioned that you use a high-quality image but I'm not able to get a proper prompt. How did you prompt your image to be that type of high quality? What is your prompt? I'm going to share with your mind right now. I really want to explore this even very much further.

[autumn spear]

Man...this unusual activity notification is pretty rough

[rigid nest]

Flow seems to be broken for 2 days. Lots of inconsistencies. It was working well for me earlier.
It does not follow character consistently. Photos looks like CGI render.
Anyone else noticed such recent changes?

[gleaming lynx]

second one looks pretty realistic bro

you are overthinking it

(you can add more freckles but i would advice against it)

[full blade]

If you don't mind me asking, where can I pose some of my Flow artwork?

[gloomy verge]

Flow-gallery, I think.

[spark hornet]

Yep, 4 days since that started

[rigid nest]

There is a flow-gallery channel here

[full blade]

Thanks^w^

[stable helm]

What is with the unusual activity notifications. I have been fine for weeks and last 2 days it fails 99% of the time with an unusual activity error

[sinful lotus]

An annoying, clearly incorrect message that is really a pain when it starts. I've already put this up on the labs bugs site, but I did not hear anything back.
Fortunately, today it has not been doing it. But then again, I've been on Higgsfield doing videos there instead to see how it performs (yes, checking out an alternative for when errors like this start and I have to get stuff done)....

(unusual activity).

btw: try signing out completely and logging back in. 99% of the time, it goes away for me, at least for a few more generations.

[stable helm]

I’ll give that a try thanks. It’s just so silly because every now and then it just works for a minute or 2 then stops again. Is the unusual activity just trying to use their product 🤣

[pulsar coral]

Clearing cookies and/or using incognito browser window helps sometimes. When it works again don't try to generate more than 10-12 images at once, wait till even the last one finishes generating or failing to generate, then wait at least 1 minute before generating 12 more at most and so on. Just don't rapid click again.

Don't constantly refresh your browser window either, refresh it as less possible, that surely also triggers Google's recaptcha security.
Because I imagine people refresh their windows to clear the failed attempts.

Helps to tab to somewhere else while the images generate so your finger doesn't want to click regenerate. This has turned into compulsive gacha for some people and all the new censorship didn't help.

Free accounts seem to get the problem more than payed ones, but payed surely do sometimes too.

And yes it's amazing how they find ways to add new error messages when other ones haven't been fixed yet or had their causes clearly stated.

I'm sure I did see the error yesterday cause I cancelled my Google subscrtipion cause NB2 is way less useful now, I only use it to write text to Nano Banana 1 images. But I did that and worked fine.

Yes it did come to this, don't need to struggle if I can just use NB1 in some Gemini canvas.

Also that's how you send feedback about your discomfort.

[rain ivy]

i wished flow let us use nb1. it's honestly just better for a lot of things even before nb2's censorship changes. i'm glad it still has imagen though even if it isn't as good as when it was in gemini

[pulsar coral]

Ah yeah, you could somehow apply "pro" mode to NB1 at Gemini too. Some say it doesn't work that way, but when you actually tried it was surely noticeable and somehow working.

I think a canvas can also do that

It's like how Nano Banana Pro is actually just a more juiced mode of Nano Banana 2 not a separate model, and this can be proven when used at Gemini since the "pro" feature is also used for other things. NB2 was put up to use less resources and be less expensive... I somewhat like NB2 more than Pro tho, Pro pedals to the metal a little too much sometimes.

Or used to like it, I'm not really into creating pics of cute chibi furries tbh and seems that's mostly what it's good for atm.

[frank apex]

help

[split raft]

can you tell which one you need help with?

[distant ferry]

Hello everyone, is anyone else experiencing a similar strange issue like I am? 😕

[pulsar coral]

Clearing cookies and/or using incognito browser window helps sometimes. When it works again don't try to generate more than 10-12 images at once, wait till even the last one finishes generating or failing to generate, then wait at least 1 minute before generating 12 more at most and so on. Just don't rapid click again.

Don't constantly refresh your browser window either, refresh it as less possible, that surely also triggers Google's recaptcha security.
Because I imagine people refresh their windows to clear the failed attempts.

Helps to tab to somewhere else while the images generate so your finger doesn't want to click regenerate. This has turned into compulsive gacha for some people and all the new censorship didn't help.
Free accounts seem to get the problem more than payed ones, but payed surely do sometimes too. [/Pasted]

Flow can't generate images faster than that anyway.

I'm on a free account too now so I know it's working fine this way.

[smoky wolf]

I'm generating a talking head video with Veo3 Fast and I want zero background music. It keeps adding BGM even when I tell it not to. Any tips on how to fix this?

[simple stone]

please don't close WHISK. because flow STILL BROKEN

[analog radish]

Why I can't put ingridients on the 3.1 Lite model?

[lone notch]

anyone noticed that both NB2 and NBPro output visibly lower quality images than a week ot two ago on the same prompt?

[sand nymph]

hey! has anyone faced this problem? it's been around 8 hours since my last generation. but flow text me that i'm sending requests too fast. what's wrong?

[burnt canopy]

same bro

[grim nest]

Same

[distant ferry]

Same

[agile yew]

why i cant use google flow?

[autumn night]

how to fix this? got an ultra plan but doesnt let u generate 🙁

[lone notch]

probably bugs or new wave of safety-ism

probably both

[pulsar coral]

Yeah if you run into that error just after 5 minutes even, you more likely just ran out of image generations for the day for that model, Flow doesn't display the correct message for it, should be saying something like "You ran out of image generations for today, come back tomorrow 🍌 "

Check if you were using Nano Banana Pro and if so try swapping to Nano Banana 2 and vice versa, each model might have some more images left, if both are dry just come back tomorrow around the same time you successfully generated an image and it will work again.
Wont help to check in just 4 or 5 hours, needs the whole day. [/Pasted]

^ This is not the same as the unusual activity error tho, her message is different I translated it from Russian

@burnt canopy@grim nest@autumn night For the "Unusual activity" error message, Clearing cookies and/or using incognito browser window helps sometimes. When it works again don't try to generate more than 10-12 images at once, wait till even the last one finishes generating or failing to generate, then wait at least 1 minute before generating 12 more at most and so on. Just don't rapid click again.

Don't constantly refresh your browser window either, refresh it as less possible, that surely also triggers Google's recaptcha security.
Because I imagine people refresh their windows to clear the failed attempts.

Helps to tab to somewhere else while the images generate so your finger doesn't want to click regenerate. This has turned into compulsive gacha for some people and all the new censorship didn't help.
Free accounts seem to get the problem more than payed ones, but payed surely do sometimes too.

Also after checking this it's also possible you also start getting the "You're going too fast" error message. Same would apply after that If you keep getting that same one after like 5 mins trying more likely your account just ran out of image generations for the day, gotta come back tomorrow around the same hour you successfully generated an image last time.

[pulsar coral]

People have often fixed that one by contacting here: https://support.google.com/googleone/gethelp?authuser=0&hl=en

[pulsar coral]

Google staff have also said this:
Check if your Google Account completed the age verification.
Verify if you are using VPN to an unsupported area.
If you are using Google Workspace subscription plan you can follow the instructions here.

Here's more info.

[hollow furnace]

I use Ultra, but I still encounter these errors all the time. There are especially a lot of them today. I pay $250 a month to monitor errors?

[burnt canopy]

I did all of these, but still nothing changed.

I also cant download any image or videos i have previously generated

[pulsar coral]

Ultra Tier does have a limit for the amount of daily images you can generate, it should be pretty high tho way above free/plus/pro that reach around 400, Ultra has above 1000 images daily or so, but the "You're going too fast" limit is not permanent.

[tight osprey]

Ultra not worth subscribing bro, please i want refund

[burnt canopy]

also using ultra

[hollow furnace]

This applies not only to pictures, but also to videos. This applies to the entire f*cking Flow.

[pulsar coral]

Yeah also the failed attempts count, even if you got nothing.

[hollow furnace]

its problew start 1 week ago

[tight osprey]

REFUND

REFUND BRO! I CANT GENERATE IMAGE SINCE YESTERDAY

[hollow furnace]

Same question. Do I pay for a subscription to see the errors?

[burnt canopy]

I just wish we can fix this right away, higgsfield nano banana pro takes way too long when i generate on it

[pulsar coral]

The errors just don't clearly state their causes sadly and their service isn't unlimited, more often they error messages are just trying to tell you that you exhausted your resources for the day or to slowdown a bit. We've reported these to staff before but yeah, doesn't help the error isn't clear about what it's causing it.

But none of the errors is permanent for the day from what I've seen, I did finish all my image generations for the day yesterday and currently waiting for the reset.

You just get used to how it works.

[hollow furnace]

This doesn't work at all. It's impossible to get used to it. Google simply cheated me out of $250.

[pulsar coral]

That goes for the vid attempts too, and yeah like said if they failed to generate that also counts.

[hollow furnace]

You advise me to just pay Google and do nothing.

[burnt canopy]

Is there a reason why i cant download any image or videos i have previously generated?

[pulsar coral]

No just wait for tomorrow and you'll see it works again around the same hour.... it's just how it works really.

[hollow furnace]

It worked fine before. Someone just broke it with the latest updates.

[hollow furnace]

I've been just looking at the mistakes for a week now.

[pulsar coral]

Once you see that, you can take a decision for real.

[bold nacelle]

if the issue persists flow is simply not worth it. Maybe they need to raise the price? I would rather pay more but for something that actually works

[pulsar coral]

idk I did just complain about it in General too, wasn't being rude about it and they deleted my message.

[hollow furnace]

This used to work perfectly. Now it doesn't. You're suggesting I just shut up and wait instead of letting Google fix the problem.

[pulsar coral]

Yeah I do too, but it just gives you some hours to work daily.

[hollow furnace]

So maybe it's worth updating to one that clearly states all the limits? I'm buying the Ultra package as unlimited.

[pulsar coral]

I don't think they've stated it to be unlimited anymore, tho at some point Ultra did have some unlimited features, this is what I know they've said https://support.google.com/gemini/answer/16275805?hl=en

[hollow furnace]

Its Gemini apps limit, not FLOW

[pulsar coral]

No Flow and Gemini share the same Google account. This one's for Flow too.

[hollow furnace]

its dont working, i dont cant 500 prompt in day in flow

[hollow furnace]

There is nothing written there about FLOW.

[pulsar coral]

Your same Google plan links you to both Flow and Gemini accounts, so that's a good way to get some more images and videos.

[bold nacelle]

yea it says i can do 1000 banana generations, i can do maybe 50 if i am lucky.

[pulsar coral]

Thing is they just increased the censorship and copyright protection, so you might get a lot of failed attempts, and those count.

[hollow furnace]

This doesn't work. I can't even do 20.

[pulsar coral]

tbh I did also get like 10 good images from like 200 attempts yesterday

[hollow furnace]

So, do you think this is normal? Instead of writing about mistakes, you defend them.

[pulsar coral]

I don't like it either, it's just most of the old users already got used to how it works.

[hollow furnace]

I've been using FLOW for over 5 months and every time it gets WORSE AND WORSE

[pulsar coral]

Yes it's gotten worse no doubt. Eitherway knowing all this you can probably make a thread at labs-bugs-feedback to complain about it.

[pallid violet]

Hey, check https://support.google.com/flow/answer/16526234?hl=en

[hollow furnace]

And instead of writing bug reports, you sit and say that we need to wait, we need to be patient...

[hollow furnace]

its not limit

[pulsar coral]

The credits are only for the vids yes. For images there's not a stated counter, all one can do is get used to it.

All these tips do apply more to how to use a free account tbh, I did also cancel a couple days ago.

Being fair the credits do refund if it missed to generate a video, but even so you do have a daily limit for attempts be that they succeeded or not.

[hollow furnace]

To be fair, it's worth noting that Google has gone nuts.

Personally, I'm going to sue Google. My work just stopped and I'm losing money every day.

[fresh scroll]

I am an ingredient to video user. Once the Fast Low Priority is gone how can I do it for free? The current Lite Low Priority doesn't have that.

[pulsar coral]

Eitherway what we know about Flow's error messages is:

-If you keep getting the "You're generating too quickly" error message, swap the model from Nano Banana Pro to Nano Banana 2 and viceversa and if you still can't get more images after like 3 minutes that just means your account hit the daily limit, just try again tomorrow around the same hour you successfully generated an image. According to tier you get a set amount of image generation attempts, and the failed attempts do count.

-If you keep getting the "Unusual activity" error message try clearing cookies and/or using a different browser, if none of that works for some reason just wait for some hours until it stabilizes again, when it works again don't generate more than 10-12 images at once wait till even the last one is generated/fails to generate then wait around 60 seconds to generate more. Refresh as less possible or not at all because that triggers Google recaptcha security, otherwise you'll have to do these steps all over again. (idk this did work for me and haven't gotten the error again.) This won't stop you from getting the "You're generating too quickly" error tho and it's reasons are stated already, once your account is dry for the day gotta try next day, and the missed image generation attempts do count.

With free tier accounts you for example get around 20-30 free attempts to generate Nano Banana Pro images and 100-200 Nano Banana 2 images, it doesn't matter if you succeeded to generate an image or not. And around double that amount for each tier upgrade. This may change according to daily resource availability.

The amount of missed image/video generating attempts has currently increased a lot because of all the new censorship and copyright security. So to avoid being flagged by google security might wanna try to rephrase your prompts making them shorter so you can revise them more easily for problematic terms instead of playing gacha trying to get an image.

[pulsar coral]

Also there's these "Batch image generators" and auto clickers some people might be using, yeah just don't. Or make them work really slow about a minute a click.

[pulsar coral]

As well remember if your Flow account is dry for the day already, your same Google Flow account free or payed links to Gemini, you can get some more images and videos there or play with the canvas mode to try to generate yourself a site to also make more videos and images.

Gemini's Nano Banana 2 can be told to generate like 4 separate images at once similar to Flow, but yeah needs to be told to do it by prompt. Same way it can generate the image with requested aspect ratio by prompt, end your prompt in something like "Aspect ratio is 4:3" and it will obey you.

And yes the images have a Gemini spark watermark there, so be sure you use the download button on the image not right click save as, that will give the image 2K resolution automatically, and it will also allow you to use this watermark remover on it, it can do it in batch and the images won't lose quality: https://geminiwatermark.org/batch-watermark-remover

There's surely easier ways to make it understand how to do this, but it can.

There we go. Man what's with the huge heads, NB2 does need to be told everything now. NB1 doesn't have this problem.

Regenerated in 3:4 aspect ratio, Downloaded in 2K, watermark removed, easy peasy.

[merry mango]

when you use the stroke on flow to show a part to modifiy , the stroke stay in the picture modified !

[gleaming lynx]

is anyone here making something high quality long form?

[lone notch]

is is dead curently?

[plain finch]

can any1 explain why its saying prompt must be provided

[pulsar coral]

At your Flow's project's screen Go up right to the cog icon for "View tile grid settings", there turn off the "Clear prompt on submit" setting and it will stop forgetting your prompt.
Unless you refresh, but then you can still right click on your image/video to find the "reuse prompt" function

[pulsar coral]

It's a bug yes, but doesn't always happen. Might wanna take a screenshot of it and report at labs-bugs-feedback. I don't think enough people have reported that one.

Hasn't happened to me yet tho, cause I usually just try to get my images in one run. But they'd call it more the "select" tool around here.

Waddayaknow it did just happen to me, yes this is a bug. Such a good meme ruined.

I gotta go donate blood like a good boy scout now, but I might report it myself later.

[soft tiger]

Anybody know why every time I try to make a video using flow I get the failed option about policies? And also how do we remove the veo from the video it’s annoying

[plain finch]

its still not wokring

[pulsar coral]

Maybe save the prompt refresh and try again, doesn't work on the same screen immediately. Or start a new project.

[plain finch]

icl i started a new project like 5x

only doing this on my pc asw

[pulsar coral]

That's weird, maybe try clearing cookies and logging back in, and if nothing made it work you might just wanna report at labs-bugs-feedback. Could be something about the image and prompt.

[plain finch]

ill try

its doing it on a different browser too

https://www.tiktok.com/@painter_professor?_r=1&_t=ZN-95XQ3lrJrmA

Could sm1 try explain 2 me why im stuck at such low views aswell please 🙏🙏

[pulsar coral]

It's probably recognizing your image as having a prominent figure or something unsafe, might wanna run it at Flow's Nano 2 or Whisk instead to edit something in it alike "Remove unsafe elements" or change something about the face. And the VEO watermark is there due to policies about AI from each country, USA residents don't get it for Ultra plan. Otherwise you need a specific VEO watermark remover because at this point the problems about it are too many just gonna say this one worked for me https://reel.money/tools/remove-veo-watermark

Watermark

No watermark. If you use a random watermark remover for other things it'd give a moving blurr there or worse, needs to be specific for VEO

[lone notch]

Could sm1 try explain 2 me why im stuck at such low views aswell please

everyone SEVERELY overestimates how much people watch AI slop
it's fun for like 1 day for the "wow effect" but after watching 50 few seconds videos about nothing people just scroll past
I suspect someone is going to vibecode some kind of AI sslop-adblock at some point

[pulsar coral]

Yeah AI does get a lot of hate, however it doesn't get it that much for little unsafe stuff at art sites that allow it like Deviant, and there's many alike.

Youtube and TikTok are full of haters tho.

Same for most other social media alike Facebook and Twitter/X. You just gotta know what to do with it or be REALLY good at it so it's well received.

I've seen even at Facebook AI waifu pages get thousands of likes and followers when they are good, NB2 ain't very good for it anymore tho.

So it's mostly LoRA stuff.

[unreal viper]

hey! please use #1238621511401734254 to submit any bugs you find!

[neon coral]

i dont think it was a bug tho. its frustration it happens every time no matter what, i dont quite understand

[unreal viper]

it sounds like one! if you use the thread the team will be able to see it and be aware of it easier than looking in here where it will get buried!

[neon coral]

ah found one similar

in the bugs feedback

[unreal viper]

you can add to that one then to bump it up so the team know its not just one person having the issue!!!

[tough spoke]

this is just my opinion, but some people are scared about change, and AI is clearly an invention that is and will be life changing.

This has an uncanny resembelance to the Luddite Movement in the early 1800s which was against Industrialisation, strangely with the same arguments that people bring up about AI now.

[neon raven]

whats up with the unusual activity errors

im paying 200 bucks per month for an ultra subscription and i cant even generate?

[lofty kraken]

Mine just stole my credits and doesnt let me generate video

[unreal viper]

hey! please use #1238621511401734254 to submit any bugs you find!

[pulsar coral]

Yeah but sometimes the images really are ugly. Some people use some very old LoRA models and that doesn't help. Many just stopped using LoRAs when they tried Google's Nano models...

Sadly it's really not that much fun anymore.

[burnt pilot]

i dont know why its happening, when i use my images with my face on it using nano banana pro on any other platform the face turns out fine, but on labs its just so bad, looks nothing like me, why is that?

[pulsar coral]

Are you sure it was Nano Banana Pro? Some sites still have the original Nano Banana/Gemini 2.5 flash image, and that one would base itself much more on the reference image.

[burnt pilot]

Nah im sure its pro

[pulsar coral]

Like if you were using Whisk for example yeah that's totally not Nano Banana Pro neither Nano Banana 2.

Anyway you can try prompts for Nano Banana Pro to obey

Stuff like

[Identity of character: True to reference image without any creativity freedom, same facial structure, same body, same props.]

or something like this

Facial structure of character: Completely true to reference image without any creativity freedom, this is not any existing character so never use any external sources to create it.

[cold schooner]

Hello I AM FROOM BRAZIL

[pulsar coral]

Hello Froom Brazil, I'm Buster

[pulsar coral]

Thing is Nano Banana 2 and Pro more often tend to try to help you to complete your images by using references it was trained with from the internet somewhere or just uses the internet somehow to find similar images.

Even tho no one asks for it to do that or seem to like it doing it.

[cold schooner]

Where in the world are you from?

[pulsar coral]

The internet.

[burnt pilot]

Ok ill try

[cold schooner]

I understood

So, is it possible to use Flow on a PC?

yes or no

[pulsar coral]

From the website yes https://labs.google/fx/tools/flow. But if you mean like running the model locally on your PC? No that can't be done.

[pulsar coral]

But the original Nano Banana wouldn't do that, even tho it is perfectly capable of recognizing prominent characters be them real of fictional.

Some head high up in development thought it was a good idea for Nano Banana 2 to search for existing material to help people complete their images as if most everyone would be creating stuff with characters/people that exist already.

And in practicality, it would indeed be a good idea, like for me it works really well.

But yeah this is not what the actual majority of people actually want... whodathunk it, most people want to make their own images and OC characters, in fact the amount of people who'd prefer doing this is inimaginably huge for some reason... I don't get it either seems a little boring.

So NB2's little "Search whatever" ability actually gets in people's ways unlike how NB1 works.

But the worse part is all the copyright blocks now, NB2 trips on itself cause of the same reason, sometimes it stumbles onto copyrighted material similarities and can't create the images.

... which is why NB1 is still better at it since it doesn't do all this, I don't get why they keep wanting to remove and hide it from everywhere, when having the model enabled is the actual solution to all this.

[burnt pilot]

Its so weird

It messes up my face

But when i use someone elses face it turns out fine

Tf

[pulsar coral]

Maybe it recognizes you as some movie star or something

[burnt pilot]

😂

[pulsar coral]

Yesterday it was confusing someone with Sean Connery

[terse raft]

hello world

[burnt pilot]

Bruh this is annoying

[terse raft]

guys where can i get nice flow propmts

[terse raft]

what is

[burnt pilot]

I keep trying to generated photos with face and it ruins my face completely

I try someone elses face and it turns out fine

[terse raft]

lol i guess the main picture has some problem

maybe try a better camera

[pulsar coral]

I bet if we saw the image someone would notice who is NB Pro confusing him with.

Not that it would fix the issie.

If nothing else works you might have to go to Whisk while it still exsists and make the pic there

I've indeed seen NB2 and Pro can be unnaturally stubborn about it.

[burnt pilot]

What does whisk do

[terse raft]

whisk is not that good as nanobanana from gemini it self yk

[pulsar coral]

It uses a different image model that will surely respect your facial structure much more. https://labs.google/fx/tools/whisk

[pulsar coral]

And yeah if nothing else works remember to report the bug at the labs-bugs-feedback thread, gotta complain about it.

[burnt pilot]

Like an example on lmarena my face turns out fine

Or on gemimi

[pulsar coral]

I don't ever run to that cause they're really slow about it.

[lusty sinew]

Love when I'm trying to create images and Flow keeps telling me that people I created with Gemini are "prominent figures"

[pulsar coral]

There we go

But yeah if you see Whisk is creating the same type of image you saw elswhere, more likely it wasn't NB2.

[lusty sinew]

Where is this thread?

[pulsar coral]

I would link the thread but this discord wont allow to link to it's own threads for some reason.

Read the pinned guidelines on how to post tho, and might wanna tag it well.

[lusty sinew]

Thank you. I'm getting errors with the prominent people filter so much today that it's getting frustrating. And it's only after I had to log in and out.

[pulsar coral]

Yeah and meanwhile people are also still getting that "unusual activity" error message, idk I'm just chilling pressing the generate image button on sets 12 at 2 minute intervals and so far been getting this bread normally today.

[lusty sinew]

Usually I do gen a lot at a time, maybe I should start doing 4 images simultaneously instead of 2 and see if it helps. Or if google themselves fix it. I'm getting less "unusual activity" errors though.

[pulsar coral]

Yeah that's how to avoid that one, clear your cookies but if that doesn't work just wait till it works later again, when it works again only generate like 10 or 12 images at most, wait till all of them generate or fail to generate, then wait like 60-50 seconds and you can try to make 10-12 more again.

[lusty sinew]

Yeah, no this is every single one I've tried to do

I don't like wasting my gens on this

[pulsar coral]

Ah yeah that one's all about rephrasing the prompt, some terms pass much easier than others.

[lusty sinew]

See, I'd be less annoyed except for the fact that the wording I've been using to fix this is suddenly not working today

Not only that, but at one point I couldn't even see my projects in my browser

[pulsar coral]

Whoever censored NB2 isn't as bright as he thinks anyway, things he thinks he censored do still pass and instead it ruined the model for other more average prompts too.

Everyone has some complaint or other about it anyway

idk how things are gonna end if it isn't rolled back to how it was like 8 days ago... but doesn't look good really.

[lusty sinew]

It's every generation now. Are they really not going to roll it back?

[pulsar coral]

idk probably would have to ask at the Gemini discord instead, I mostly see Google staff replying around here only when it has to do with VEO not the image generation.

But what can be seen is an avalanche of very angry users, Reddit is like this atm too.

A lot of people arrived here mainly cause of Whisk's image generation, so it's obvious they'd think that'd be handled here. It's also how I arrived.

[pulsar coral]

About this also don't refresh the page too much or at all unless really necessary. Refreshing has higher chance to trigger recaptcha securities.

pff my NB2 images are beautiful today, if any staff has doubts NB2's censorship ain't really working feel free to ask for my deviant account link, gonna upload images later today, it's not really very unsafe stuff tho but it sure was being censored for no good reason now. And causing errors at Flow for no good reason either.

Gotta admit VEO did up the censorship to untouchable levels tho, so props on that, got nothing to show from it. Not that I use it much.

[remote crypt]

Flow is so unusable now because of the censorship system, that I think I'll consider lowering my subscription.
Here's one example of an image I couldn't even upload - with the warning message "we do not allow uploads of minors at this time"

[pulsar coral]

Today I was out at the street for hours and noticed like every sign promoting schools had some kid in it.

[remote crypt]

I don't see any kids in my image tho 😄

[pulsar coral]

Not allowing kids in images for absolutely every possible reason is not very marketing intuitive

[remote crypt]

that too

But I was able to run it in Kling, and worked just fine.

[pulsar coral]

Gotta not go out a lot to it to think that's how the world works

[faint coral]

Whenever I try to create a video or image in VEO, this message pops up and it won't generate. What on earth is the problem?

[remote crypt]

idk why it would do this, but are you on VPN?

I've had VPN block things for me sometimes, not this though.

in this case, "unusual activity" can mean that it might think someone other than yourself is using VEO

[faint coral]

I am not using a VPN. However, two people use Veo with the same account. Even so, I never experienced this error before, but it has been happening recently.

[remote crypt]

that's probably why

report as an issue

[leaden radish]

I’m also currently experiencing the same error multiple times, and I’m unable to generate images or videos.
Is this an issue that all users are experiencing?

[open spindle]

I'm so tired of all the system errors.

[warped wind]

same here

It seems that nbp works while using API

[unreal viper]

hey! please use #1238621511401734254 to submit any bugs you find!

[burnt canopy]

Bruuhhhh

this is the same exact issues yesterday, im using flow fine in 5 mins or so, then this happened again

[analog radish]

Is there a way to upscale videos which are getting errors like "Upscaling Failed!"

[harsh lantern]

I've experienced this problem too.

[lean sun]

Is flow still down? It is on my side

[south yoke]

I previously made a veo3 video without a watermark, but today why does the resulting veo3 video have a watermark?

[obtuse flint]

same. google's video game was already weak and now thier image models. cant wait for june.

[harsh lantern]

Right now, at this very moment, the same problems are still being encountered.

[woeful void]

Guys, why aren't the selection tools to indicate, what should be changed or removed in the image, NOT working at all ?
They are imprecise and inattentive. I've tried ten times, and Nano Banana isn't changing what I specify.

How to fix this ?

[old bay]

Seems to be "working" for me somewhat, though it uses a very ineffective form of generative fill that doesnt always yield the desired outcome. A more sophisticated selection tool might help, on that lets you draw around the shape youre trying to remove around the object maybe... the red drawing tool doesnt feel very effective.

[civic kraken]

Flow is not working?

[sinful lotus]

Today, I have only used Flow for video and not nano banana at all. I haven't had any of the "activity" messages or other random failures. If I need nano banana, I go to a gemini screen and do it there (and I did a few today without incident). Maybe there is a success formula in this? Or maybe Flow figured out I have done countless generations (without any messages) on Higgsfield the last few days and got nervous 🙂 🙂 🙂

[civic kraken]

Ohh Nano broken :c

[sinful lotus]

Go to a Gemini screen, and likely you can get it done there. I'm assuming if you have Flow, you have gemini (at least I do on the plus plan).

[pulsar coral]

New day same bugs. I'm not gonna translate all those error messages to know who's getting what and how to fix it, might wanna write/send them in english tho more often it's the same thing.

Except the one in spanish about "An error has occured" that totally is Flow assigning less resources to a free account, having a subscription even the lowest value one helps against that error when Flow's having too much traffic.

But don't think that's gonna help you not to get other types of error messages, it wont.

Eitherway what we know about Flow's error messages is:

-If you keep getting the "You're generating too quickly" error message, swap the model from Nano Banana Pro to Nano Banana 2 and viceversa and if you still can't get more images after like 3 minutes that just means your account hit the daily limit, just try again tomorrow around the same hour you successfully generated an image. According to tier you get a set amount of image generation attempts, and the failed attempts do count.

-If you keep getting the "Unusual activity" error message try clearing cookies and/or using a different browser, if none of that works for some reason just wait for some hours until it stabilizes again, when it works again don't generate more than 10-12 images at once wait till even the last one is generated/fails to generate then wait around 60 seconds to generate more. Refresh as less possible or not at all because that triggers Google recaptcha security, otherwise you'll have to do these steps all over again. (idk this did work for me and haven't gotten the error again.) This won't stop you from getting the "You're generating too quickly" error tho and it's reasons are stated already, once your account is dry for the day gotta try next day, and the missed image generation attempts do count.
NEW With this "Unusual activity" error it's also possible you're sharing your account with someone and he's making those mistakes, or if you're in a VPN it might also flag you, so yeah try to stop doing any of those doings.

With free tier accounts you for example get around 20-30 free attempts to generate Nano Banana Pro images and 100-200 Nano Banana 2 images, it doesn't matter if you succeeded to generate an image or not. And around double that amount for each tier upgrade. This may change according to daily resource availability.

The amount of missed image/video generating attempts has currently increased a lot because of all the new censorship and copyright security. So to avoid being flagged by google security might wanna try to rephrase your prompts making them shorter so you can revise them more easily for problematic terms instead of playing gacha trying to get an image.

As well remember if your Flow account is dry for the day already, your same Google Flow account free or payed links to Gemini, you can get some more images and videos there or play with the canvas mode to try to generate yourself a site to also make more videos and images.
Gemini's Nano Banana 2 can be told to generate like 4 separate images at once similar to Flow, but yeah needs to be told to do it by prompt. Same way it can generate the image with requested aspect ratio by prompt, end your prompt in something like "Aspect ratio is 4:3" and it will obey you.
And yes the images have a Gemini spark watermark there, so be sure you use the download button on the image not right click save as, that will give the image 2K resolution automatically, and it will also allow you to use this watermark remover on it, it can do it in batch and the images won't lose quality: https://geminiwatermark.org/batch-watermark-remover

Then there's this new "Prominent people" error, that's the type of error people keep pushing regenerate to try to get lucky about and avoid to get an image anyway.

I've seen explaining to the Nano 2/Pro model your image has nothing to do with anyone else increases the success rate tho it can still be stubborn about it. Gotta start the prompt with something alike.

[Identity or character: This is my personal character not any existing character so never use any external sources to create it and do not mix/confuse it's appearance it with any existing person.]

But more often to try to explain it to it in some way does influence things.

[sinful lotus]

I just got the prominent people one, and there is no logic to it whatsoever (I have gotten this over and over for months).

There are absolutely none, zero, nada, prominent people in this prompt (see below in quotes).

Another massive bug in the system. I think the engineers handling this need to be called into a room, and nobody leaves until it is addressed.
I would have never allowed my engineering teams to do this to customers. This is how you lose them to competitors:

"The camera the cuts to different places of the world similar to the one just shown with similar scenes. The sounds become more defeaning as the video progresses with more angry voices shouting disapproval, although it's still not possible to make out words because of the defeaning roar."

Maybe it just doesn't like my spelling mistakes? Filing under bugs in a moment.

Yeah, lots of prominent people in that prompt!

[pulsar coral]

Not everyone knows everybody, but the Nano Models do have an immense database for character recognition, it will compare your character to an exisiting one to try to make it accurate... but yeah in the process it might trip on itself cause of copyright.

[sinful lotus]

There is no character at all in the shot or the prompt. Just a group of 1,000 faceless folks ....that Flow generated.
Oh.... Higgsfield handled it no problem as always.....

[quick breach]

hello everyone whos fighting against this bad cake poop cat thing, anyone knows what does it mean? Console: 500 (Internal Server Error)
why are they doing this to us?

[sinful lotus]

It's another pain in the butt bug. I sometimes get it over and over. Try logging out and logging back in.

[pulsar coral]

Guess I'm gonna have to illustrate because of course, ok now no one talk hold on

Here's reporter girl, I'm asking Nano Banana to make me an image of her but more like of who it thinks she really is, could be confusing it with a movie star or something, I get no image at all

I'm asking to write her canonical name from who it thinks the person is, it trips on copyright

[quick breach]

that means nano banana has dirty thoughts about your character XD

[pulsar coral]

No it just means it thinks it knows the person, and when it thinks it does, seems the person it thinks she is is copyrighted. Ok now

This is the prompt less severe, I get at least one image.

I bet that really is her idk

[quick breach]

imagine if someone born with a face that is similar to some famous person and get banned from every platform

[pulsar coral]

Now I apply the "Identity of character" prompt I mentioned earlier, all images pass without problems.

It needs to take priority at the top of all prompts tho

[quick breach]

wow this is hot

[pulsar coral]

Then again this might not work with everything, NB2 will be more stubborn according tho who it thinks each person is and how famous it considers it.

Anyway, it's indeed possible to negotiate with the NB image models about these things, same goes when it thinks the character is underaged cause that's not always it.

[quick breach]

and thank you for this info

[pulsar coral]

Now I'm gonna push the copyright prompt a little more to know who is Nano Banana 2 confusing her with.

This is not her real name, it's who Nano Banana 2 thinks she is, it thinks she's Linh Chi and that's some famous Vietnamese soap opera actress/model.

... this doesn't really excuse the Nano Banana 2 model from doing these things tho, Nano Banana 1 doesn't have this problem.

[quick breach]

nb2 took a break for some conference

[pulsar coral]

Didn't know the cardashians had a jurassic park

[dusk patio]

so I'm sure this has been covered before but I can't find it in the search, how exactly are reference images supposed to work when it comes to videos, every time I add an image and say "make a video of X in the reference image" it ignores it

[severe stream]

FIX !!!!!!!!!!!!!!!!!! YOUR !!!!!!!!!!!!!!!!!!!! APP !!!!!!!!!!!!!!!!!!!!!!

[sinful lotus]

Try logging out completely and logging back in.... that usually works for me.

[harsh lantern]

But it didn't work for me.

[sinful lotus]

Or they could just fix it... Sorry, it didn't work. This is why I headed to Higgsfield today (flow started acting up again). When it starts doing stuff like this, I turn off VEO and go to Higgsfield so I can get things done.

[lone dagger]

In Flow, can we share projects with other team members or am i missing something?

[old bay]

Anyone else notice the video models switching automatically on their own changing not only the model but how many generations you can make? This is supremely irritating and quite cheeky because it pollutes the project and changes the token values guys... naughty naughty, we need hard coded user settings if you're going to play this game...

[hybrid meadow]

Hello people

i have a question

I've seen a lot of people making videos with Flow, however I see that there are many scenes in the videos, more than the paid plans can give in credits. Could someone tell me if there's a way to do that?

I need to make some videos with about 6 scenes per 5-second video.

Maybe 2 videos on a day
one it's ok for me

[frosty niche]

Been over 24 hours and this is still popping up?
Sheeeshh

[copper fog]

I just realized scenebuilder is really broken. All of my scenes start out muted and switching to a different scene remutes the previous scene and I have no way (without downloading all scenes) to know if the audio is good or not.

[pulsar coral]

I can't believe this, we just found out some accounts still have Nano Banana 1 but don't have imagen 4. While most other accounts don't have Nano Banana 1 but HAVE imagen 4. Just to the pile of bugs.

This is what the image model selection looks like for everyone else, or does it not?

Eitherway gonna go see how to report this, because it's a HUGE find.

Did no one else seriously see or say anything about this? Couldn't have been only Lenniboy here noticing it.

Man also this screenshot is so useful, if for some reason your account still has that model that says "Nano Banana" without the "Pro" or "2", you can still make the same kind of images as Whisk at Flow.

It might be tied to Ultra plan accounts tho.

[ashen ridge]

Agree , Flow is down for 24 hours since yesterday

[frosty niche]

I knew something was a mist!

[mighty sage]

Bro I used it for 1 hour. And my work is gonna finish🥹 and then gone

[ashen ridge]

I cannot create vdo since yesterday

[chrome agate]

First time posting here. I build cognitive architectures for AI systems — working on a project called VERA that treats multi-LLM setups as a single system rather than a collection of tools. Dropping a Flow generation tomorrow that’s a companion piece to a white paper on inter-agent friction (inspired by Microsoft’s recent Critique deployment in Copilot). Curious to learn from this community.

[harsh lantern]

Hey bro, when will it be usable again? 🥺

[ashen ridge]

always failed

[errant lark]

Just try another browser

[reef mural]

Well as what happened to components? it only works in veo 3.1 fast and fast lower priority? this is a very nice feature to use when having many diffrent things that has to combine to one in a video shoot. They atleast have to introduce it to veo 3.1 lite? as far i have seen veo 3.1 lite is a downgraded version of veo 3.1 fast lower priority? or what do you think?

[errant lark]

They have to release veo 4 as soon as possible, veo 3 is become trash towards other Chinese Ai video generator

[thorny lynx]

Ingredients are not working properly in flow . Whenever trying to add an image before a word in text via @ it's not giving any results just saying. Opps something went wrong !

[harsh lantern]

For me, it didn't work. 🥹

[errant lark]

I had same issue before a week, tried another browser immediately fixed, however you can try another email

[harsh lantern]

I have more than 10 email accounts. I've tried all of them, but it still doesn't work.

[errant lark]

Try vpn

[deft grove]

hello everyone can anyone please confirm Am I only receiving this Flow error or is it happening with everyone:

audio generation failed. Please try a different prompt or send feedback. you have not been charged for this generation. You can update your setting to return silent videos.

Its been a week and most of the generations are failed and receving this error. Any solutions?

[cobalt scroll]

Hi, I’m still getting this error in Flow:

“Failed. We noticed some unusual activity. Please visit the Help Center for more information.”

This has been happening repeatedly, even after waiting and retrying. Sometimes reopening the page lets me generate once, but the next generation fails again with the same message.

This issue is seriously affecting normal use. I’ve also seen other users reporting the same problem, so it does not seem to be an isolated case.

Is the team aware of this issue? Has anyone found a reliable solution?

[reef mural]

I can get it to work using 3.1 fast or 3.1 fast low priority

[cobalt scroll]

Is anyone else seriously frustrated with Flow right now?

For days, I keep getting:

“Failed. We noticed some unusual activity. Please visit the Help Center for more information.”

At this point, it doesn’t even feel like a user-side issue anymore.
People in this community have already reported:

• no VPN, still getting it
• waited hours or even a full day, still getting it
• reopened the page, maybe one generation works, then it fails again
• different accounts, different prompts, same result

So what exactly is triggering this?
Because right now it feels like normal use of Flow is being treated as “unusual activity.”

This is not just annoying — it makes the product unreliable and basically unusable for actual creative work.

How many of you are still getting this error right now?
And has anyone found a real fix that works consistently?

[reef mural]

I can just say as some other does. they need a veo 3.5 or 4 to get things back on track, longer renders, native 1080p renders at least no upscaling. 4k would be awsome but i dont thinks is possible at the moment to much computer power used. Multiple scenes in one shoot,m extended video rendering to , max 20-30 sek and the ability to add your own sound. This would put Veo back in the drivers seat. 🙂 big wishlist there: )

[ashen ridge]

I can't create vdo since yesterday Flow always failed something went wrong and another thing, when the vdo creation failed why does the credit charged T__T

[reef mural]

and one other thing thats is really annoying, render a AI-person, for the first few clips it ok then it suddenly think its a celebrery person 🙂 and cant render, can be to hard to add some metadata that says it has been rendered internally 🙂 just a tip 🙂 its really annoying

[cobalt scroll]

Same here. Mine is mostly failing with the “unusual activity” message, and sometimes reopening the page only lets one generation go through before failing again.

It really feels like Flow is unstable right now, not just a single-user issue.

[rain ivy]

i use flow in firefox private browsing. no issues. if it asks you to enable cookies at any time say no

as soon as i enable cookies or use browsers outside of private browsing mode, i get that error all the time too

[azure nacelle]

Hi guys !
Any french users here ? Do you know why the voice generation sometimes takes a québecois accent ?

[pulsar coral]

...Now you also gotta clear cookies to fix the "You're requesting image generations too fast" error message.

Well gotta give it to theme, they do give us something new everyday ain't that awesome

The errors for upscaling can also be fixed like that.

[lean sun]

Multiple image generation errors using chrome, does anyone have this problem now?

[neon leaf]

Yo bro new coming soon veo 4?

[snow matrix]

Does anyone have more info? I just saw this in a Facebook group, but without details.

[unique abyss]

That seems like made up nonsense.

[reef mural]

agree

But it shouldnt be to far away, alot of the chinese variants are beating the sh.... out of veo now. So hope google continues to give us new versions and dont do as chatgpt did killing it.

[reef mural]

very very nice

[unique abyss]

Yes, it seems both likely and welcome that there will be an update reasonably soon.

[terse arch]

GUYS, This problem is very annoying. Please don't tell me to reinstall AntiGravity because I already did that and the problem persists.

[stone garden]

This is drivin me crazy

[merry apex]

In case the servers are overloading... everyone really should just try generating ONE image just to see it's working or not. No need to try so many going at once. Let's just take a break.

[stone garden]

I tried one at a time... That's the result after waiting...

It does not even go past 3% before giving me the error BUT when I use any other platforms it works 🤷‍♂️

[unreal viper]

hey! please use #1238621511401734254 to submit any bugs you find!

[dense cairn]

Hi everyone, I’m reaching out to see how many others are currently blocked by the error: "Failed. We noticed some unusual activity. Please visit the Help Center."

I am on the Google Ultra plan, but I’ve been unable to generate content for week now. I've tried:

-Switching models
-Clearing cache/cookies
-Waiting 24+ hours
-Change Browser

This error is effectively wasting my subscription time. Has anyone successfully bypassed this, or found a way to escalate this to a human support agent? I’m currently looking into a refund that i've paid with credit card until the service becomes stable again.

[wise flame]

Did you try the help center?

[ocean sky]

Hey guys, Is there a method to create TTV in large quantities, like 10 TTV prompts all at once instead of one at a time? I'm a pro Gemini subs and couldn't find one.

[bold falcon]

I reported it in labs, bugs, and feedback, and they acknowledged my post, and probably about all we can do, for now. I have Ultra as well and it is definitely extending my creation time on projects.

[quick breach]

you know sometimes i do some vibecoding thing and there are moments when agent has its plan but completely lose some pervious project features, i feel like they're vibecoding flow, 1-2 months back i was much more satisfied with this platform

[tough perch]

Not sure if I am missing something, or if this is JUST a bug, but...

I had been working on politically charged videos in Flow, where truck drivers are filling up with diesel, and the view switches to the readout on the pump, and the numbers are going up and up (actually pretty realistic these days).

In real movies, they have someone working continuity, so that scene changes, switches, and come back and look like they should look. Is there anything like this in Flow, or is my problem just a bug?

Specifically, we see a truck driver with a ball cap on. After the scene change, he is no longer wearing the ball cap. Another error is that sometimes the gas pump has the hose stuck in the pump, and other times the hose is in the truck. I have even tried it by being extremely precise in the prompt that I write, but it still switches things around. I even went back to Gemini and asked it to build me a prompt to get what I want (sans errors), and it generated a beautiful prompt, but the same problems come up.

Just SEEMS like there should be something to set for scene continuity, like in big-time movies. This way if there is a scene change and then you come back to the original scene, everything looks precisely like it did originally.

Maybe this is just wishful thinking on my part. I still have fun with it, but it bugs to burn off all my credits trying over and over and getting different results every time.

[tough perch]

Just an FYI... I posted mine under bugs and a Google person asked me to come here and post it... great idea, but ?!?!?

[unreal viper]

i cant seem to find where you posted it? would you mind linking me to it so i can fwd it to the team?

[tough perch]

"Video continuity?" is the title under the bugs section...

and the Google person was LucasFromGoogle

Also when I post the link to the messages it says this cannot be posted

Forward worked, so I did so

[unreal viper]

Thank you for messaging me! its seem you posted originally in the Gemini discord server which is why I couldn't find it! thank you for forwarding it to me!

from the message from Lucas he directed you to this server to post your bug in our bug reporting section which is #1238621511401734254 rather than in the Gemini servers bug feedback section! hope that clears things up!

[tough perch]

Yeah, no. He said go to the Labs Flow section and post it. I clicked his link which led me here. It did NOT say labs-bugs-feedback, so your message cleared nothing up, but thanks for trying.

"I'd suggest joining Flow's Discord Server to report this!"

I clicked your link and re-posted yet again. The thing that helped was your link (the statements not so much), but thank you... hopefully, in the right place now

[pulsar coral]

Do clear cookies to be sure, but sometimes it can indeed be a while longer than 24 hours, depends on the last time you generated an image successfully so that's really when you gotta be there, not exactly the next day because there isn't set reset time. But it will reset for sure.

Everything's burning down yes, and free tier accounts are surely getting the worst end of it, but pay plans do also have a lot of bugs too.

And for free plans the "You're requesting image generations too fast" message surely appears now when you generate above 6 images at once without waiting for all of them to generate. Only clearing cookies and refreshing fixes it, but when it doesn't get fixed at all anymore even with that you indeed ran out of image generations for the day and same gotta come back next day around the same time it was working, little later maybe.

But I've seen payed plans suffer from that one way less or not at all unless you generate above 12 images at once, until they do run dry then it does stay like that until the reset.

[reef mural]

A lot of generation errors going on now with nanobanana 2 dont get alot of correct renders just alot of fails

[spring prairie]

Hi All, If it's possible, I would like to speak with a specialist regarding Google Flow policies.
I need to clarify whether images generated with a free account can be used for commercial purposes and what the legal limitations are concerning intellectual property and copyright.
Could you please direct me to the correct department or provide the official documentation for these specific inquiries?
Many thanks in advance

[full flicker]

Hey everyone! Just a friendly reminder to keep all conversations in the supported language for this server. This helps our entire community engage effectively, and we've gone ahead and deleted the recent posts that didn't comply. Thanks for your cooperation!

[full flicker]

Hi there, could you please send your question to us via DM to @real wyvern . That way it won't get lost here in chat. Thank you!

[green needle]

Guys, it's already the 16th and my projects still haven't been transferred to Flow. Is there any estimate of when this will happen?

[crisp basalt]

whats this?

oh wait , does that mean even if the audio failed , it will still show the video?

[old bay]

Ooo interesting!! What is this 🤔

[pulsar coral]

I'd call it throwing the towel about trying to do it by prompt.

That might be better for the next version anyway

[fathom quest]

hi everyone. i find that nanobanana pro and veo 3.1 are now pretty much unusable in achieving realistic results when it comes to textures, motion etc.
the difference in performance between March and now is unbeliveably bad.
in Veo 3.1 Fast, i tested prompts that i had great results with in MArch, and now it is impossible to replicate the camera movement, body movements and even textures.
has anyone else experienced this?

this looks like a deliberate Nerf to me. is there anything i am missing?

Gemini Pro and Thinking also seem less smart. the halucinations have increased drastically and ability to follow prompt instructions reduced

[crisp basalt]

Veo 3.1 Lite “ingredients to video” model is now available.

[vale lintel]

Hey everyone! Flow’s safety filters are flagging me as a "known person" and blocking face generation. Interestingly, ImageFX (Nano Banana) generates my face with these exact same photos without any issues, but VideoFX (Veo) blocks it every time.

How can I legally bypass this trigger in Flow to generate my own face? Is there any verification process?

[bright finch]

Bikin program

[near aspen]

why im not seeing any edit option in scenebuilder can only trim the video

[sand nymph]

it's happening again😩

[spring crystal]

its the only reason that i stopped to use google flow.

[burnt canopy]

same broooo

[twilit bluff]

wow I've been using fast/lite for so long I just did a full quality mode video and forgot how HD veo can actually be lol

[small mirage]

me having this problem for 2 weeks and still not fixed.

[tropic laurel]

https://tenor.com/view/krasi-fifa22-fifa22rage-fifa-rage-krasi-rage-gif-26170968

[stable field]

any operation with Flow demanding more than one image referencing encounters an immediate generation Failure

[fathom quest]

what are you using instead? i m looking to switch as flow performance has regressed by about 1 year... incredibly disappointing

[burnt canopy]

yeah, what would you guys prefer to use instead? Higgsfield is the same with lots of issues lately

[fathom quest]

i was considering Higgsfield. what's your experience with it Ezekiel? has i gone downhill in the last few weeks as well?

[stable field]

photoshop powered with NB, but with more care with prompts

[worldly shadow]

@stable field

[lean sun]

video generation is down on my side

[crisp basalt]

Normally, when audio generation fails, the system wouldn’t return the video at all. But now, with the “return silent video” setting enabled, it still delivers the generated video just without any audio

[fossil python]

Upscale broken again. Mention multiple images also lead to Something went wrong. 😭

[bright kettle]

whats the issue guys?

[rigid nest]

Request to Google Flow Team:

There have been several threads about Flow being totally unstable for last 2 weeks.
While @wheat flicker has been collecting all the information and passing it to the developers, we have no update on what triggered this.

It would be really useful to get some information about what caused such behaviour at the first place and if there is an estimated time for a patch. We are completely in the dark.

There is also common feedback about improving the error messages. It's too generic.

Thank you for your attention and I appreciate the work which has been done to Flow so far.

[upbeat sierra]

Always impressed how fast you spot things @crisp basalt

Ingredients to Video is now Available w/Veo Lite

I'm sharing ahead of our official communications because we're really excited to get this out to the community. We have heard that many of you are excited about Lite, but had concerns about Ingredients. We listened and now Ingredients-to-Video is available for Veo Lite and Veo Lite (relaxed priority).

Excited to see what the community creates

[pulsar coral]

I'm pretty sure they've been trying to tweak the error messages at least, cause now they give different explanations of what really causes them... sadly that's not really much a fix to them.

[upbeat sierra]

Please file a bug in #1238621511401734254 with as much information as you're willing to share so the team can look into this:

• prompt used
• mode of generation (frames vs ingredients)

[upbeat sierra]

we hope this is working as expected now as well, but please let us know!

[crisp basalt]

I’m about to try it out right now. 👀

https://tenor.com/view/wake-up-bed-morning-messy-hair-gif-18298820

It’s 2am but sleep can wait

[proven onyx]

Why can't I delete projects? This is nuts.

[gentle ravine]

its11:17pm

[pulsar coral]

We do have 3 or 4 threads about that over there, also we noticed some Flow accounts have the Nano Banana (Gemini 2.5 flash image) model while others don't, which I did make a thread about there too, they haven't seen it yet I think.

[crisp basalt]

The extended videos are now properly stitched when using the Lite model. Thank you for fixing em!

[upbeat sierra]

you should be able to delete your projects from the 'Projects' overview page. this is on the bottom right of a project. if you're having any other issues please file a thread at #1238621511401734254

[proven onyx]

Yeah, I try and try and try. It says deleted. then they just come back.

[upbeat sierra]

those threads will be your best place for updates. please try to keep this space clear of repeated requests, as it is much easier for us to track at #1238621511401734254

[upbeat sierra]

please file a thread in #1238621511401734254

[proven onyx]

It all started when I tried to clear the entire library thru this// happened to TWO of my accounts.

[proven onyx]

I understand. I did follow up. There are others with the same issue. ty

[upbeat sierra]

thank you for flagging, let me make sure the team is aware / on this

[upbeat sierra]

can you share your thread from #1238621511401734254 with me?

[proven onyx]

ok, thanks. happy to help. dont want to SPAM the chat, but yes verifyable on two accounts, one Ultra, one Pro. my other play/messaround account I didn't do this and it seems ok. Other NEW projects seem to delete fine, but everything up and until/including the moment I hit Delete library data is corrupted and zombi projects..wont delete

sure stndby

[upbeat sierra]

thank you! can confirm we are actively looking into this, we will share an update in #1238621511401734254 once we've resolved cc @main saddle

[proven onyx]

ok thank you very much. I hope the info I provided is of some help @upbeat sierra

[grim nest]

why i am getting watermark in my ultra veo account, model lower pirority ?

@proven onyx

[stark grotto]

When we can expect Veo 4?

[brisk field]

Yeah, Veo 3.1 kinda sucks now.

[tacit aspen]

THANK YOU! Thank you!!! Thank you!!! Thank you!!! Thank you!!! Amazing!!! You guys are the absolute best!!!! Thank you for being advocates for us. #grateful

[stark grotto]

I mean i like Veo 3.1 but we need better quality and stuff

[proper drum]

Hi

[proper drum]

someone can tell me the credit cost of veo lite - relaxedpriority? is that aso available for pro?

[upbeat sierra]

Relaxed Priority for Veo Lite is only available for Ultra subscribers.

[fallen swan]

Tralalero Tralala Porcodilo Porcoala

[proper drum]

Understood. Is it possible that these modes will be released for 'Pro' users so we can generate content at half the credit cost? Also, is there any possibility to remove the Gemini watermarks from images and videos?

[fading dirge]

bro im not going to lie, veo 3.1 right now is bascially only good at lip sync, still one of the best, but not so good at everything else. It's fallen behind alot man. And i love flow but bro its getting hard to justify. Especially after using other stuff. Others fast mode are better than veo 3.1 quality mode. I cant wait until Veo 4.

[pulsar coral]

Veo trips too much on itself with copyright and censorship which it looks for by itself who knows where more than on the actual reference image. I've been having better results with Wan, it's doing better than Grok even. Probably Kling matches it more but I didn't like it.

Veo works much better if you don't use a reference image for sure tho, but I don't got no market for that.

I'd say without reference image VEO does work much better than all those... but the guessing game is too random

[fading dirge]

Even without ref images, it most of the time puts out a video thats more funny than good. I get it, model camer out a year ago really, but i got the ultra plan and man its just getting hard to justify spending that for veo 3.1, thankfully ultra comes with alof of other stuff and you can gen a ton of nano images

[pulsar coral]

Yeah you get images and other things, and a lot of Gemini usage with the same account too.

tbh initially I payed more for using Gemini than Flow, once they started with unruly censorship Nano Banana 2 became unusable and I cancelled subscription, but Gemini allows to make a canvas for original Nano Banana, that' still useable for something.

Can't wait for them to notice that's still good for something and remove it tho.

It's like they actively look for working features and sites that have an actual purpose and take them down.

[bright kettle]

Anyone know what's the issue?

[fading dirge]

Yeah my main reason i have ultra is for video gen though, but alot of veo videos have me laughing from being like bro, what is it even doing? Or bro, this video is so bad

[fading dirge]

you keep trying a prompt thats keeping getting blocked

[bright kettle]

Nah,
Actually I open 2 Same tab with 2 diff Project

Then Added promts on 2 tab same time,
But got blocked

[pulsar coral]

Try:

-Switching image models, from Nano Banana 2 to Pro and viceversa, you might still get some more images today.
-Clearing cache/cookies and refreshing.
-Change Browser

If none of those work your account is dry for the day, you gotta come back again tomorrow around the last time you successfully generated an image, a little later even, it isn't exact 24 hours sometimes. All plans have a daily limit.

Also if you're sharing your account with someone you could also be getting that error.

Also once it works again don't generate more than 10-12 images at once wait till all of them generate then wait like 60 seconds to try again, if you go faster than that you'll also get that error.

It's much worse for free plans, can be as low as 6 images at once according to daily demand and region.

[bright kettle]

I have ultra plan

[pulsar coral]

Yeah but even that isn't unlimited or safe from those. If you get those probs in ultra, free is hell. Eitherway one of those will work.

[pulsar coral]

Oh also don't use a VPN since other people could be in the same network. Payed plans do mostly work fine if you don't go faster than 12 images at once, and if your account/network isn't being shared by someone also using Flow.

[bright kettle]

im not having any issue on image gen, Im having issue on video generation veo3 - quality

[pulsar coral]

Video generation also has that type of limit by error message.

It's more associated with Flow than the video and image models.

It just hates that people ask for generations too fast.

[bright kettle]

agree

[harsh ibex]

image not generating , have any issue?

[near aspen]

yes same issue

[unreal gate]

Yesterday I tried nano banana pro to genearte images, and stop generate images "You try generate images too fast..." and now I cannot generate any image. I think I will cancel my subscription (PLUS) if thr problem persists 🙁

[crisp basalt]

Gemini already got veo 3.2 , flow when 👀

[proper drum]

3.2? No 4? 🙁

[gleaming lynx]

why are my image upscale requests getting cancelled

[pulsar coral]

It also has daily limit, you know the deal. Unless the "unusual activity" error triggered it, then clearing the cookies 'could' still fix it. If not yeah account is dry.

[gleaming lynx]

i didnt do anything today, no generation nothing just tried to download an old image upscaled

[pulsar coral]

It still allows you to generate images? LIke with imagen 4?

[gleaming lynx]

yeah nano banana pro too

[pulsar coral]

Maybe cookies could still fix it then, if not you could've hit a limit yesterday. But if you didn't do anything yesterday either you might just do better to report the bug.

[gleaming lynx]

clearled cookies still didnt work

will report it thanks

[exotic shell]

havent checked out the chat in a while:
does google still stick to the bad plan and give ultra user the downgrade to lite (and make 3.1 fast low priority non-free)?

[analog sedge]

just realized how much better seedance is compared to veo

Multi billion dollar company btw

Lol

[dire sigil]

I'm stuck on a project using Google AI Studio's Build Mode (the Vibe Coding interface). The app scrapes Instagram data for sales scripts.

The weird part is that the scraping tool actually finishes fine. The data shows up perfectly in my React UI, and the logs even show {"terminal":true}. But right after that, the AI Studio chat crashes entirely with this error:

generic::INVALID_ARGUMENT: [...] Please ensure that function call turn comes immediately after a user turn or after a function response turn.

It looks like the agent is trying to fire another model turn without waiting for the current one to close properly. I've tried remixing the app, resetting the VM, and spamming the Retry buttons, but it's completely bricked.

Since this is all happening inside the integrated Build Mode, I can't manually edit the message history array to fix the sequence. Has anyone found a hack to reset the agent's turns or keep tool responses from crashing the chat? I'm migrating to Firebase soon anyway, but I want to finish building out the UI here first. Any ideas?

https://klipy.com/gifs/sigma-boy-8

[icy bobcat]

what does it say?

[exotic shell]

seedance belongs to tiktok?

[pulsar coral]

The golden rule to sell anything to anyone is to give the customers what they are looking for, not to push personal ideas and products onto them at least not from the get go. Flow, VEO and the Nano Banana projects do still have time specially to reverse the really bad decisions from like two weeks ago, cause whoever just took the wheel doesn't even seem to be around to care to see how things are going...

I just hope in long term if things fail they don't blame the customers for it, cause I think we've all seen Google doing this before.

[rough token]

Anyone else having ""You're requesting generation too quickly. Please wait a moment and try again."" ???

thought my quota of 1000 images was up but have been waiting 12 hours and still have it...

and if this is due to the quota, why can't we see that anywhere?

[pulsar coral]

It's not 12 hours anymore, it resets next day around the same time you successfuly generated an image/video, so a little more than 24 hours. No there isn't a counter, but the failed attempts count, doesn't matter if you generated an image or not.

[rough token]

Ah.. Thank you for that!

Feel like they should definitely add a counter and a more clear message that you used up your quota

[pulsar coral]

Flow seems to be a website auto-generated by Gemini, not made by humans, so idk how easy it'd be for them to add a counter. One can play with those things at the canvas maker from the Gemini website to get an idea on how it works.

[rough token]

I know very little about coding/cs but with all gigabrains at google suurley it should be easy to add a little counter 😅

Well.. will have to go with nano banana 2 for a while

[pulsar coral]

btw your same Google account from Flow also links with Gemini, you also get some more images and videos there, but images have a spark watermark so be sure you use the download button so this bulk watermark remover works with them, otherwise it wont https://geminiwatermark.org/batch-watermark-remover

The canvas maker doesn't need you to know coding, you basically ask it to make you a website how you want and it will.

Kinda like how making images work, all functionalities from a site alike Whisk or Flow should be possible there... to some extent.

And I saw that gif grasslely.

[rough token]

You mean the watermark in the bottom right corner? It's actually removed on the ultra plan (even in gemini chat)

[pulsar coral]

idk never payed for Ultra to test it at Gemini, but if for some reason it appears yeah that's how to deal with it

[rough token]

Ty for the tips!

[pulsar coral]

Gemini also allows you to control the aspect ratio by prompt like ask it

Aspect ratio is 3:4

Or whatever reasonable, it will obey.

[rough token]

Yeah the thing I love about flow is how is it is to bulk create images

You can create like 25 at a time

[pulsar coral]

That too yes.

[rough token]

chat is much slower (you can have multiple tabs to speed it a bit)

[pulsar coral]

I've only tried with 4 tbh, never tried 25

[rough token]

Yeah you do 4x and then just "reuse prompt" do it like 6 times 😅

but would only do it on ultra plan

[pulsar coral]

Well gotta get as much return value as you can I guess.. did cross my mind too but I'd be lazy for that one

[quick rapids]

what is veo 3.1 pro?
its listed in featureset for ai ultra

when they announced veo 3.1 lite and changed the feature matrix to mention it, they also updated it to say "5 videos / day using Veo 3.1 Pro"
no other documentation mentions this model. seems like a leak

before the update these were the models mentioned

[river comet]

lol sorry for that mate you gonna try it again because practice make perfect or you need mentorship

[thorn drum]

does anyone have a link to the changelog for flow?

[mighty bison]

Did the wait time for credits get extended?

[green needle]

Guys, it's already the 16th and my projects still haven't been transferred to Flow. Is there any estimate of when this will happen?

[rough token]

Still blocked...

Sanity check: Is nano banana pro even working for anyone atm?

[pulsar coral]

No it should be working if you still have generations left atm. Was it around now when it was working fine yesterday?

[rough token]

Hmm.. Was quite a bit later, but I should have passed the time while I started using Nano Banana Pro yesterday for sure

I was editing a big project and spent my 1000 images throughout the whole process 😅

[pulsar coral]

No it needs 24 hours from the last time it worked more or less, there isn't a standard reset time for it.

[rough token]

Alight! Just need to be patient then

[pulsar coral]

NBPro might be working quicker than NB2 from what I've seen tho

[rough token]

NB2 still works for me, it's only pro that is cooked

pro is even blocked in gemini chat

think pro takes much more infra on googles end

[pulsar coral]

I've seen sometimes it does reset quicker tho, no hurt in checking after 12 hours too

Also idk if there's too much of a hurry and you need it now one can also pay $5-$20 for a second plan, those vids have watermarks but images don't. And yeah there's always Gemini also linked to your same google account. Like we were chatting up there it doesn't hurt to learn to use it until, it has some amazing stuff.

Then again being honest, NB2/Pro has a lot of problems with copyright and censorship now, it also used to have them two weeks ago but right now it's %200 more annoying, borderline unusable for some people.

But if it's been working for you as it was yesterday, then all's good I guess.

[rough token]

Yeah I prefer pro it has slightly better taste imo

but NB2 seems to do certain things better

[pulsar coral]

Pro runs dry real fast. idk for me it's the opposite, since I do anime stuff. The realism NBPro pushes ain't that good for it.

Specially with things alike water, if I want to see little fish in it, they float over the water instead of being underneath it.

It's kinda funny when seen

[rough token]

Makes sense

Rumours openAIs next image model might come out next week

[pulsar coral]

I did send feedback about it from the Gemini page, I'm pretty sure next update will fix it

[rough token]

will be interesting to try

btw this is how my screen looks... Might deserve the rate-limits lmao

[pulsar coral]

Yeah it's not good idea to make more than 12 images at once, I don't think Flow can process them faster than that anyway

[rough token]

The limit seems to be 25 here

[pulsar coral]

Aka you can generate 4 images per generate button push, so you can push it 3 times that gives 12, and wait till all are done

[rough token]

Might again be difference in plans

[pulsar coral]

Then wait like 30 secs and you can do it all again, or that's how the other plans work, might be a little more lenient for Ultra

You go above 12 images at once you will start getting the "Unusual activity" or the you're going too fast message.

[mighty bison]

We really need to change that message

[rough token]

YES! please just tell me if I hit my limit...

[pulsar coral]

Considering they just keep adding new error messages all the time, I do think they could tweak it. In fact Flow did use to give the correct error message for that, it used to say something like "You've ran out of Nano Banana 2 image generations for the day come back tomorrow" or something alike, one day the message just stopped saying it for whatever reason and instead we got the "You're going too fast" message.

Which yeah has proven to be way beyond confusing.

[hazy kestrel]

Why can't I extend in scenes anymore?

Oh ok, found it

[river comet]

Yeah same here. Since they switched to Flow, “extend” is either hidden or just broken. You have to open the clip itself now, and even then it doesn’t work like before.

[trail depot]

@potent shell what in the hell is this actually? every time you put this panel out containing the changelog in flow its blurry/out of focus and the font is broken (unless thats what youre going for).

an intern could have just prompted a better panel into existence USING FLOW ITSELF. How this stuff is happening almost daily at a trillion dollar company actually blows my mind.

[pulsar coral]

Remember to submit all your bug reports complaints and feedback to the labs-bugs-feedback area, it's the weekend so staff might not be around as much and your complaints will get buried here.

Over there. Gotta love how this Discord won't allow me to link to it's own threads, they get blocked by itself every time I try.

Remember to read the pinned guidelines and tag posts so they can sort it better, there's just a lot of complaints atm.

I'd say something about the video feature probs but I'm not very good at that one, maybe Wrigley will be around later.

[crisp basalt]

👍

skill issue

[pulsar coral]

And yes wrigley I was just out with some grass just now

[crisp basalt]

?

weirdo

oh

[pulsar coral]

Flow's Nano Banana Pro's limit does seem to be resetting a little earlier than NB2's, mine just resetted around 5 pm pst, I don't remember using it at that time yesterday, just for notes.

... funny cause I'm the guy who doesn't like or need Pro, my stuff just works backwards.

5 pm PST is 7 pm o' clock in my time zone.

[crisp basalt]

A person (Subject A), standing on a lush, overgrown suburban lawn, violently rips a thick slab of turf and tangled roots from the ground. With a sudden, explosive motion, Subject A swings the heavy, dirt-bottomed grass mat directly into the face of Subject B, showering them in wet earth and green blades. The action is immediate and aggressive, captured in a tight, fast-moving medium shot under bright, slightly harsh daylight.

[pulsar coral]

I just saw a thing at Comfy UI that could generate that exact type of action... as long as it's just 5 seconds long... idk where'd it come from. Seems similar to VEO

It's not one I've ever seen before or people talk about. But yeah just from watching that gif one could get an idea on how things are going.

And I mean it's similar to VEO cause VEO is usally good at handling two people doing things at random without reference image, but I'm sure it's not VEO.

[crisp basalt]

what video model

[pulsar coral]

Lemme go get it again

[dry frost]

I'm guessing Wan

[pulsar coral]

Nope

"LTX-2-3", yeah that doesn't sound very marketable

VEO does have a more catchy name, gotta give it that

Yeah that's also why I had to go get again, we use Ai we don't need long term memory ffs

[crisp basalt]

u do now

https://tenor.com/view/йокелемене-men-in-black-light-sunglasses-gif-15516530

[pulsar coral]

Seems it's been around for a month even, and heavy marketed as NSFW

[crisp basalt]

oh thats Wan

dont ask me how i know

https://tenor.com/view/spongebob-spongebobsquarepants-gif-9146238

[pulsar coral]

No I'm sure it ain't

[robust oak]

[pulsar coral]

We all so busy arguing about Flow here while time goes by, eitherway I'll go post something about it in offtopic

[crisp basalt]

https://www.youtube.com/watch?v=PEUcCF-jVCI&pp=ygUOZGVlcG1pbmQgdmVvIDI%3D

wow i suddenly miss veo 2

https://www.youtube.com/watch?v=qtLyyGkdh_U I have never seen this one before..

[pulsar coral]

People do often say it was better yes

I think while trying to tweak copyright and censorship something about consistency got busted

Everything was better like two months ago actually.

[crisp basalt]

gemini also having issues too

[gritty cloak]

I’m sorry to say but is terrible this flow thing

Whisk used to give me so many good pics it takes ages to get on good pic in this one

I dont know if its just happening to me but gives terrible small eyes and terrible faces i dont even recognise the people

[crisp basalt]

Sorry to hear that

[pulsar coral]

Flow uses Nano Banana 2/Pro which yeah is being handled very poorly and unprofessionally now but was working okeish like 2 weeks ago before whoever else took charge of it, Whisk uses the older Nano Banana 1/Gemini 2.5 flash image model for the precise mode which funny enough is surely more artistic.

There's tons of daily complaints and negative feedback about Flow's models and the takedown of Whisk, I imagine it's not being taken seriously or being ignored for whatever reason.

There's simple solutions like re-enabling imagen 3 and the original Nano Banana at Flow just like Google AI studio has it, but also falls on deaf ears, I don't think many people have noticed it's all about the image models each site has.

[serene heron]

the page is working?

[lyric hull]

I believe it might be down. It's not working for me when I try to generate anything

[near aspen]

ayoo why flow is not working

[lyric hull]

Gas is expensive these days. Can't afford to keep the machine running all night

[near aspen]

https://tenor.com/view/bruh-bro-really-omg-wow-gif-21835931

[pulsar coral]

Flow's online from south america

[signal hare]

Speaking in Hindi with confidence, dialogue, emotion should not remove glasses from his face ]<><>दरअसल, ये कोई और नहीं बल्कि बिहार के चर्चित बाहुबली मोहम्मद शहाबुद्दीन के बेटे ओसामा हैं।

[old bay]

Flows been working fine for me here in the UK