flow-general・🌊

[woeful marten]

It's still working just fine for me

[fiery field]

No, same with me ...

[raven siren]

Even the ones that get to 99% fail in the end...

[brittle hearth]

Its not working for 3 days, please fix this asap

[raven siren]

Render through API is stil working..

[woeful void]

Same for me, free account
Foto rendering doesnt work,

Most frustrating, without any explanation.

[empty dove]

Google creates an AI product and it's been broken for days without a peep of what's going on. I get better notifs from my SIEM vendor when there's a minor service issue.

[gritty stream]

literally gambling

oh im not alone with this issue LMAO

[unborn bolt]

How you do this? can you explain the chain? 🙏 so i can try it

[brittle hearth]

nope i can't generate anything it fails the second i send it only one goes to 99% and it fails

[gritty stream]

LOL

[hybrid pawn]

Ok same issue with me.

[raven siren]

In Google ai labs you can create an API key. You do have to use a credit card but it''s pay as you use, and when you sign up you get 300 dollars worth of credits. I use it with an app called RenderZero Studio, which is actually pretty cool. It cost me 15 dollars though it might be more expensive now. https://www.youtube.com/watch?v=I8qFYXxnIRs

[gritty stream]

tuff

[bright heron]

The best thing we'll probably get is a "refund of credits" when they repair the server issues, but for many of us with pro plans the generations are 0 cost

[pulsar coral]

Yeah that's it more likely, they haven't put the correct message there in days, it used to say "You've ran out of image generations for the day for Nano Banana 2/Pro, come back tomorrow 🍌 " Or something alike.

[pulsar coral]

Read above

[pulsar coral]

Tho like he's showing here the "oops something went wrong" message means you might still have a chance for it to work and haven't been axed yet for the day, and yeah it just becomes a gacha game at that point.

But no once you still get the "You're generating the images too quickly" message after like 5 mins, you might as well come back after 12-24 hours, wont work sooner.

[bright heron]

I've gotten all the above error messages, from policy violation, generating too fast, and oops failed. I think they're all just because the server issues

[raven siren]

I meant Google Ai studio to get the api key

[pulsar coral]

Yeah but when it's like in this screenshot, so consistent, you're likely done for today atm

[gritty stream]

Something went wrong is NOT a ratelimit

I had 4 generations this week

And im paid user

Its false flagging stuff

[pulsar coral]

Yeah tbh now I'm a little stumbled on how you're getting it this way

[ornate basalt]

I had noticed there was a GUI update during this period, and they may be gearing up for all the Whisk addicts showing up.

[pulsar coral]

Did you try a different image and prompts? They also fail?

[ornate basalt]

In other words, prepping to have all that data transferred.

[ornate basalt]

It's not anything you're doing, it's a platform-side issue.

[gritty stream]

same prompt, its not permanently failing, one out of 25 passes

[pulsar coral]

Yeah in fact some regions might be getting this problem more than others, they did display an error message for that at the site

[gritty stream]

like mass retrying makes one out of few pass

[pulsar coral]

On my end it hasn't gotten that bad so can confirm

[gritty stream]

it was fine until today for me

like 0 fails

[pulsar coral]

I'm in South America... which maybe barely counts as whole America

[gritty stream]

im EU

[pulsar coral]

So yeah that might be while it still works okeish here

[wet flame]

I think they are cooking

Whisk is also downed for me

[mighty bison]

I guess y’all are saying something went wrong, mine is saying I need to take a break. I haven’t even used it for hours, it’s a bug

We really shouldn’t discard Whisk in my opinion, it’s more easier to use than Flow

[wet flame]

imo Flow is just a fancier interface wannabe

"why not just implement stuff in this instead of making a new that"

[gritty stream]

the servers are cooking

(themself)

[wet flame]

fr they are fried bro

[gritty stream]

wiat whats whisk

[wet flame]

A previous gen of Flow

[gritty stream]

so like flow but worse?

[wet flame]

Like Flow but better

[gritty stream]

better?

[wet flame]

better

It's like Flow with Snickers

[gritty stream]

just checked, whisk ui looks MUCH better

[wet flame]

ikr

[unborn bolt]

Thank you ! Gotta check this

[gritty stream]

but whisk uses worse model i think

DAYUM THIS IS SO GOOD

[wet flame]

Yeah, that would be the only problem, I don't think they are using Nano Banana

[gritty stream]

no way theres animate option

[wet flame]

There is video, too. It uses Veo, but probably only the fast option, seeing it costs 20 credits

[gritty stream]

video sucks, but animating can be good

[wet flame]

animating = video, no?

It's the same thing there

but aspect ratio is also locked for 16:9 only

[edgy gazelle]

\

[astral fern]

Oops all failed again

[flat quiver]

2x Upscales are failing most of the time too. The only thing that seems to work right now is sorting images into Collections.

[wet flame]

Good time to use this

generated with Whisk

[eternal arch]

ok

[crisp basalt]

https://tenor.com/view/my-lord-gif-10109530379829104506

anyway good mornafterevening labs peeps

[wet flame]

good evening

[unborn bolt]

I have 30 different images and I need to determine which ones are the most legible and effective for a general audience. Is there an app, website, or AI tool where I can upload these images to rank them based on their potential efficiency (specifically predicting CTR and retention) before I go live?

[pulsar coral]

Same old. But apparently it's just me here.

I did backup my stuff already, needed a whole evening for it but no idea yet if everyone else needs to do the same.

[wet flame]

It's on April 30, not March 30

sike

[pulsar coral]

Really, but thing is I don't have a whole other month to complain like the rest. I'll still be done by the end of this month.

[wet flame]

I probably just needed to store the prompts, currently just trying the old prompts on Flow

Yeah

But that's good thinking, it's better earlier than late

[pulsar coral]

Plus Google are known for suddenly changing things out of nowhere without asking anyone and without listening to what anyone says. Wouldn't sail that patched ship yet.

[wet flame]

Yeah..

[unborn bolt]

Is nanobanana pro better than nanobanana2?

[pulsar coral]

From current global feedback, like at reddit and wherever place they're actually testing it, Nano Banana Pro is much better at text rendering (alike for logos, posters, subtitles) quality and precision, more complex compositions, visual polish and more complex scenes.

Nano Banana 2 is capable of almost %90 of that same quality, but it's faster cheaper so that explains why it's more readily available when Pro isn't useable at some hours.

[unborn bolt]

Thanks you! Im trying to redesign my outro for karaoke videos. Since it has 0 cost i'll try everything

[pulsar coral]

That's also why their generated images sometimes look samey

Some users can't even notice the difference

Nano Pro will indeed give best quality even tho the margin is not that much higher than Nano 2.

Most people around here would actually prefer Whisk's image results tho, and that's a completely different older model called simply plain "Nano Banana" (Gemini 2.5 flash image) that Flow doesn't currently have but it did at one point, so if you ever went to Whisk and want to generate the same type of image at Flow, don't even try it can't.

... unless you completely plaster the art style from a Whisk image to Flow's Nano 2, but that's too much hassle and wont end that well that fast either.

So anyway again simply, Flow's Nano Banana 2 gives 2–5× faster generation ~50% cheaper per image compared to Nano Banana Pro, but Nano Banana Pro has around a 10% higher better quality and precision output.

Yeah its that weirldy exponential.

[unborn bolt]

Thank you very much for this info!

[winged ivy]

Hello is anyone having Flow repletely fail to generate images? I tried the same prompt with nano banana pro and 2 also imagen 4 and it just says something went wrong each time 😕

[pulsar coral]

Yeah no doubt Flow is having some probs atm, they did say they're on it and fixing it so just gotta wait. Might take a while tho.

[winged ivy]

Ah, ok. I couln't tell if it was on my end or not

[dry frost]

Could be location based. I just created a batch of four and did an edit or two with no issues.

[pulsar coral]

Some american accounts are fine tho, mine's working like everyday

But I did confirm some european countries are having the heavy issue from a Discord stream

And they're at least accounts with the "Pro" symbol, so paying is not the issue either, even if you payed or not it might happen to you.

[celest jay]

Yes flow has been failing for 3 hours to ne

[indigo wadi]

any updates on when it might be back up?

[celest jay]

no idea...

[pulsar coral]

I'd say more comfily try around 12-24 hours after you successfully generated an image, because that's around the time when accounts have daily image generation count resetted so you can be sure that's not what's going. If it persists by then yeah it's still Flow derping,

Like usually 7 pm PST is the usual time for everything to reset on the internet, that's usually a good time to try again, it's 11:52 AM PST atm

Yeah the internet is also round not flat.. or maybe egg shaped

[indigo wadi]

hopefully we get a reset and it works

[pulsar coral]

Yeah it's been going like this for around a week now, it'll be fine later for sure

[winter dawn]

this service is fcking joke, cant someone at least tell us when is it going to work? im paying 250 for something that has performance issues every single day

[pulsar coral]

Actually I check it again right now and Pro/Ultra payed plan accounts should still be working normally from America.

Yang Xiao Long cosplayers too hot for Flow yes, but it's supposedly working, with money, if not yeah tough cookies

Free accounts are surely getting the prob even more, or totally

[ornate basalt]

And they expect us to pay hundreds for a service that don't work?

And if paying for it makes it work I ain't paying.

[pulsar coral]

idk if you're paying it shoulda been working better for you, but if it isn't "oops" we do have a problem here

[ornate basalt]

Well no one is giving solid information on what is going on and google is silent about it.

Apparently paid users are having issues too.

[full blade]

I hope this gets fixed soon

[ornate basalt]

so bad the channels on the discord are going down too.

[pulsar coral]

Here's an account with just a $5 plus plan, it has the date too, those are working fine today for sure. That explains why mine was working too, I dunno if it's regional now.

And yes I will update later Windows, I'm rendering something here ffs

... idk it's probably fair if you're paying you get some priority, worth something

[nova gust]

PPl, Im new here

Any place I can see the limitations of each account type?

[crisp basalt]

Works fine on my end tho

[nova gust]

Im testing with free by now, It says I hve 150 cr but all my gens are 0cr cost

[crisp basalt]

I’m on ultra

[crisp basalt]

https://support.google.com/flow/answer/16526234?hl=en&ref_topic=16908930&sjid=11922046787089942009-NC

[nova gust]

Thanks

But I do not make vid gen, only img

[crisp basalt]

great

[crisp basalt]

https://support.google.com/flow/answer/16729550?hl=en&ref_topic=16908930&sjid=11922046787089942009-NC

[nova gust]

English is not my main languaje, I am not asking about basic information but usage limitation of each account in image generation, all info I can find, even the one you link, is about video costs. But thanks anyway

[ruby quiver]

I want to make flow but I can’t

Who can help me please

[nova gust]

why do you can't

Any error

[main saddle]

hey all, mentioned this earlier but it got buried - we're actively investigating video generation errors, feel free to add to existing bug reports in #1238621511401734254

[nova gust]

Can I ask about Image genertion error? thanks for your help

[pulsar coral]

Yeah some payed users were having probs yesterday for sure but Google might have updated it and given them priority now. Again tough cookies.

[indigo wadi]

ye interesting note pro account image generation works but a free account does not

[pulsar coral]

That unless you did run out image generations for today, thats still giving the wrong error message no matter what

[indigo wadi]

true

[nova gust]

But is a different message

[pulsar coral]

Yeah thats still messed up

[nova gust]

The one when run out says you are generating too fast an d you have to wait

[unborn bolt]

Prompt:

Animate this still image as a 15-second outro video loop. Keep all text completely static and perfectly readable at all times — no text movement, no zoom on text, no blur on text. The image layout must remain unchanged throughout.
Apply only these subtle ambient animations:

The purple neon glow on the card borders pulses slowly, breathing in and out every 3 seconds — glow intensity varies between 60% and 100%, never fully off.
A very subtle light sweep passes slowly left to right across the entire image once every 5 seconds, like a soft studio light reflection — low opacity, barely noticeable.
The orange "JOIN TO UNLOCK" button has a gentle glow pulse synchronized with the card borders.
The background studio photo has an extremely subtle depth-of-field shimmer — the bokeh lights in the background shift very slightly, as if the camera is breathing. No camera movement, no zoom, no pan.
The microphone on the right side stays completely still.

Do NOT: zoom in, zoom out, pan, rotate, move any text, distort any element, add particles, add lens flares, or change any colors. The video must loop seamlessly.

I would like some externals opinions, i want to use this as outro for all my karaoke videos on YT. Its very sad that im paying PRO plan and still get VEO WATERMARK. Is this a bug?

Also i need 15 seconds of video and i got only 8 for 100 credits.. unfair

DownText also has a bad quality for "quality" generation

[full blade]

Is that possible?

Yes it is

[crisp basalt]

Did you claim the one month pro trial yet

Also Veo can only do maximum 8 second

[unborn bolt]

Im already Pro

[pulsar coral]

If you have a payed and plan and you still get that message it's more likely you just ran out of image generations for the day, it should be displaying something like "You've ran out of image generations for the day for Nano Banana 2/Pro, come back tomorrow 🍌 ", but instead it says you're generating too fast.

Just come back 12-24 hours after you successfully generated an image and it'll be fine again.

[ornate basalt]

It will say when you're out of renders for the day.

That is not our problem.

Our problem is it saying "Oops something happened!" errors.

[nova gust]

Thats so

[tough dirge]

is the API down also, i only get internal server error 500 with API

[pulsar coral]

If you need it so much right now you could pay another 5 bucks in other account and skip the current bug problems, it'll work.

Cause free accounts are having it really problematic atm

[nova gust]

Sounds weird

[pulsar coral]

ikr

[ornate basalt]

paid accounts are having issues too

[full blade]

I can't have a payed account, I just have a free-one

[nova gust]

I mean Im free right now but Im gonna subscribe, but Im reading PRO are having the issue too, so, I'll wait

[pulsar coral]

They surely did yesterday but seems to be working better atm. Do you have a payed account somewhere?

[ornate basalt]

You sound like you're trolling.

[pulsar coral]

I've been told yes

[ornate basalt]

Because I'm seeing evidence to the contrary.

[pulsar coral]

I just know that's how it's working atm, and others did confirm.

[ornate basalt]

And I'm seeing evidence paid or not people are having issues.

[full blade]

It was working just fine until a few hours ago

[pulsar coral]

Here's a different plus account from a different location, with fresh image limit, with today's same date images, those work fine.

The wrong error messages are just confusing everybody.

Well, mostly fine, it did fail one so you're probably right about something here.

At least they do give something

You know it's different location cause that acount is like 2-3 hours ahead of mine.

[wicked crater]

I'm in canadain servers.

[ornate basalt]

I hadn't been getting the demand message.

that and there is still a "something went wrong" there

THAT, and I did have three render normally... then back to the errors again.

[frosty niche]

Anyone else getting this still?

[ornate basalt]

yes

everyone is getting them

[pulsar coral]

Well you know the problems you're having bro, and I believe you don't worry, I also just see what I see. Anyway things are derpy today, that's for sure.

[ornate basalt]

You were talking as if you were certain though.

<@&1173646484311384114> if these guys would just speak up and tell us what is going on instead of being silent and letting speculation run wild we wouldn't be in this argument right now.

[pulsar coral]

Oh yeah another option if one has a payed plan is going the the Gemini website, tho it does also give some free images daily forgot how many.

But the same payed plan from your Google account at Flow is tied to that one, so you get double there.

[ornate basalt]

And again, free and paid flow users are getting the same error... you keep pushing people towards the paid option yet from what I'm seeing I wouldn't even consider paying for the service.

Particularly if it's been down for well over a day and the devs have been quiet about it.

[pulsar coral]

Images have a spark watermark, but can be removed with a specific gemini watermark remover.

[main saddle]

hi, I've replied several times that we're investigating. I'm doing all that I can, and hope to have an update soon. #flow-general・🌊 message

[ornate basalt]

It's not just video generation, it's everything in flow.

[wicked crater]

I hope it gets fixed so that way, I can finish making many videos for tiktok.

[main saddle]

I understand that.

[crisp basalt]

I’m glad I wasn’t the only one ☝️

[pulsar coral]

idk I'm just telling him the same thing the others say, that payed plans are working, So by that logic....

[ornate basalt]

But the paid plans are not working.

[pulsar coral]

It's not like I work for them or anything, woulda been fired already.

[ornate basalt]

I've said that numerous times.

Is English your second language?

Because you don't seem to understand me.

[main saddle]

y'all, this is not the place for this.

[ornate basalt]

yea it ain't he's just winding me up

[crisp basalt]

https://tenor.com/view/sips-tea-gif-10558867

[ornate basalt]

blocked and ignored

[frosty niche]

https://tenor.com/view/settle-down-settledown-sidetalk-side-gif-26858800

[pulsar coral]

We all are mad yes.

[unreal viper]

please can all be reminded of our first rule: Be compassionate and treat others with kindness and respect. we have #1238621511401734254 for feedback specific questions/ queries

[ornate basalt]

Remind the troll that please.

[unreal viper]

you have all had a verbal warning, please drop the conversation or further moderation action may be taken. Thank you all

[plain finch]

i keep getting an error

constantly too

[glad yacht]

I blame their safety filter... 😅

[full flicker]

please see the pinned message, thanks for your patience!

[plain finch]

ah okay thankss

[main saddle]

welcome to the server. [we're investigating this](#flow-general・🌊 message), feel free to add context to existing bug reports in #1238621511401734254.

[plain finch]

mb for not checking

[main saddle]

all good!

[plain finch]

do you guys have a rough estimate as to when its back?

[main saddle]

not at this time, but will share updates as we recieve them

[plain finch]

is there anothger free alternative to use atm?

[full blade]

I've been enjoying making pictures with flow so far

[plain finch]

me too, i was just working on a video

[main saddle]

glad to hear that, I appreciate you sharing it. hoping to have everything working as normal soon.

[rugged fox]

Flow is not able to work for me, says there is a mistake

[plain finch]

pinned message

[full blade]

I hope so. This is an example of what I made with the refs I had to made it

[plain finch]

holyyy peak

[inner wind]

silly question, i'm sure, but how man credits do you get per day on the free level of Flow? and do they carry over? I'm new to Flow. Thank you.

[nova gust]

https://support.google.com/flow/answer/16526234?hl=en&ref_topic=16908930&sjid=11922046787089942009-NC

[next kestrel]

Hi everyone! I'm having some trouble with Flow. Every time I try to generate a video, it fails. Is there a known issue with video generation right now or is there something specific I should check in my settings? Thanks!

[nova gust]

Everyone Is having the same issue

Well, not exactly bit a lot of ppl has it

[main saddle]

hey, without a subscription, you receive a one-time credit of 100 credits and 50 credits per day free of charge to try Flow. here's a webpage with more info!

[wicked crater]

For 50 Credits per day, you can generate 2 videos per day...

[main saddle]

hi, welcome to the server. this is a known issue that we're investigating! #flow-general・🌊 message

feel free to check out the #1238621511401734254 channel for similar reports.

[nova gust]

Image generation do not use credits?

[inner wind]

thank you.

[frosty niche]

☄️ for the win! We back!!

[radiant leaf]

This is kind of strange for me. I am subscribed to Plus, in Mexico, and I only have with that one 200 per month. How come a free account can have 50 credits daily?

[frosty niche]

Just a fyi, this seems to be a chrome+safari thing. Comet or if you use firefox, edge, or atlas should work, just an fyi.

[radiant leaf]

Is that only in the USA?

Because if it is not. I may go back to free. I'm really restricted, considering I'm having lots of failed videos.

Which by the way. I need help with a prompt because I can't complete a sequence because the head of the character turns like the one in the Exorcist, and I'm not even asking for the character to turn around.

[next kestrel]

i'm from Brazil

[radiant leaf]

principe carioca

[next kestrel]

Belo Horizonte, Mineiro 🙂

[frosty niche]

Sup wid it Brazil!

[tawny drift]

/Video

[autumn sluice]

Okay, so I've got The Amazing Digital Circus and Mortal Kombat on the brain. Wound up combining the two in recent Flow gens. Who'd you choose to fight whom, and why?

[next kestrel]

working hard here bro

[frosty niche]

Work harder dawg.

[radiant leaf]

I'm still waiting to see if anybody could answer my question from before... Free vs paid

If I am supposed to get more chances with a free version than a paid one

[nova gust]

I'm interested in the answer too, I'm thinking about purchansing Pro

[next kestrel]

problems using it?

[autumn sluice]

I don't do paywalls. I avoid them whenever possible.

[radiant leaf]

I'm not getting my 3 daily videos for Veo3 either so, maybe going back to a free option can help. I can then just pay the storage instead of the whole package. Even if it's basic.

[next kestrel]

i'm ULTRA with 5 people

it's worth it

[radiant leaf]

I have a Plus and a Pro but in two different countries

so, I don't know if the benefits change depending of the area

That's why I'm asking about the free one mentioned because, if that is the case, paying is not making sense. That seems to be giving more credits

I really need that video completed ASAP. I have been getting it wrong for two months with the limited credits not helping at all. 😒

[next kestrel]

i got it

[pulsar coral]

Try encasing your prompts instead of giving it a paragraph sequence.

[-Pointers: Make the animation coherent and unhurried with way more than enough frames to be fluid but don't repeat any frame and keep all frames at the same speed. Keep the character body's anatomy accurate and morphologically correct in relation to current pose]

[Scene: The character faces front locked in the same pose from reference image, never turns or moves.]

[Camera angle/Composition: Locked to the same one from reference image, character must remain full body in focus.]

[Video Quality: All professional. Remastered 4K HD ultra high resolution, sharpest definition, denoised, unblurry, clean lines, High contrast, fix any jpg artifacts.]

It's much easier when it's a single character

[worthy violet]

I keep seeing this

[pulsar coral]

And if it says something you can add a dialogue encase or add it to scene, if you want it not to change what the text says you can add "without any creativity freedom"

In fact adding that to any encase helps to lock what should happen, VEO sometimes has it's own ideas

[unreal viper]

#flow-general・🌊 message please see this message

[pulsar coral]

Also if the character forcibly needs to rotate in other cases, that problem about the head can be fixed by telling it to rotate to the same side only, VEO can handle that much easier.

[pulsar coral]

[-Pointers: Make the animation coherent and unhurried with way more than enough frames to be fluid but don't repeat any frame and keep all frames at the same speed. Keep body's anatomy accurate and morphologically correct to any body's sides or in relation to poses.
When rotating the body can only do it until a 360° finishes, the whole body can only move at the same speed as the backside and can't unlink/flip/pose anatomically/morphologically unnaturally.]

That can work combined with a more difficult scene like this

[Scene: The character dances to a popular existing k-pop intricate choreographed non-simple acrobatic routine with related music. must include at least one twirl towards the same side.]

It will rotate well, I've posted dancing examples with it before

This is because sometimes the VEO model doesn't seem to take the human body as a single coherent linked prop, and might move head, torso and backside at different speeds.

But if you explain to it how it works, it will get it.

[radiant leaf]

It's supposed to be a panned shot to the end of a bedroom where the focus goes to a mirror. The one character is writing on a notebook on her desk, but the tool keeps moving the character facing to the mirror and interacting with it or simply puts the mirror as a wall in front of the desk. 😡

[pulsar coral]

Yeah sounds like you should try a different order there, and use a background, you can arrange that in the scene slot, but one thing can't go before the other.

[plain finch]

Is it fixed now guys?

[unreal viper]

#flow-general・🌊 message please see this!

[pulsar coral]

This is my attempt from what I understand from your prompt, but you can fix it yourself, didn't remove the music but yeah could be done

That ones better

Notice how she turned but the head worked just fine, the prompt needs to be specific with the panning of the scene

[-Pointers: Make the animation coherent and unhurried with way more than enough frames to be fluid but don't repeat any frame and keep all frames at the same speed. Keep body's anatomy accurate and morphologically correct to any body's sides and in relation to poses.

Body can only move at the same speed as the backside and can't unlink/flip/pose anatomically/morphologically unnaturally.]

[Scene: It starts at a living room that has a vintage style mirror hanging from the wall coherently naturally, the camera slowly pans to where the character is sitting at a desktop and stays there in focus for the duration of the video while she writes on a notepad that was already on the desktop]

Identity of character: True to reference image, same facial structure, same props.

Camera angle/Composition: Locked to the same one from reference image, but respecting the scene prompts.

Background: Coherently related to scene. No transitions.

[Video Quality: All professional. Remastered 4K HD ultra high resolution, sharpest definition, denoised, unblurry, clean lines, High contrast, fix any jpg artifacts.]

[wicked crater]

Can you put it in a thread instead.

[pulsar coral]

I'm so bad at that, but if it needs it I can try next time

This one's done anyway

probably the scene would need to go like

[Scene: It starts at a living room where there's nobody, then the camera slowly pans to a different bedroom that has a vintage style mirror hanging from the back wall coherently naturally, the character is already sitting at a desktop then the angle stays there locked onto her in focus for the duration of the video while she writes on a notepad that was already on the desktop]

The scene specially always needs to go in best order possible

[radiant leaf]

Does it work with a movie type video? I mean, not anime?

[pulsar coral]

And if it's still failing, there's always the start and end frame, one can generate images for each

[pulsar coral]

Sure, you just need to add the image with the right artstyle, or use the "Cinematic" prompt there

You can also encase the art style and explain well there, it will know what it is

[dry frost]

When I set up prompts the setting was usually at the top, followed by the characters and then action. With frames to video it's not needed now to describe a character unless you've got no idea, or don't care.

[pulsar coral]

[Arstyle: Photorealistic Cinematic]

Combines really well for those

Like if you tell the character just to "Turn" in the scene, VEO sometimes just wont know to which side and make it look like it's having epilepsy or it's a scene of the "Exorcist" like you said, turning all it's body around in separate segments

Tho sometimes it also just works just fine

But yeah once you keep the waifus skeleton well assembled in place, it's easier to go with the scene.

[sacred mica]

Does anyone have this message when trying to generate image?
Failure
You are requesting generations too quickly. Please wait, then try again.

[pulsar coral]

Yeah Flow's having issues but they're on it, check the pinned message.

[sacred mica]

damn paying the google ai pro and still can't use it hahaha

[pulsar coral]

Was it working fine some hours ago?

[sacred mica]

hope it's coming back soon, it's been like 12 hours

[sacred mica]

yes like 16hours ago

[pulsar coral]

Since you have a payed plan it's also possible you just ran out of image generations for the day, Flow is not displaying the correct message, it should be displaying something like "You've ran out of image generations for the day for Nano Banana 2/Pro, come back tomorrow 🍌 ", but instead it says you're generating too fast.

So yeah it would be fine again around the same time you generated your last image.

[sacred mica]

mmmhhh maybe if it's not displaying the correct image then, do you know when i can use it again then?

[sacred mica]

guess i gonna sleep and wake up to see if it's back XD

thanks bro

[pulsar coral]

Since it's not working fine since since 16 hours ago, yeah it should be around 24 hours.

[sacred mica]

okkkk let's see hope i can generate more images tommorow

[pulsar coral]

Because people noticed payed plans are getting more priority atm. Free accounts do seem to be much more stuck.

[sacred mica]

idk, i see no diff hahaha

[pulsar coral]

Actually people just told me free accounts should be working fine again right now. Whoever's still getting the "You're generating too fast" message most likely just ran out of image generations for the day.

[pulsar coral]

Free accounts are working fine from most or all locations atm.

[main saddle]

hi all, we've issued a fix for reported failures. feel free to emoji react to this message to indicate if it's working for you or not. it might take some time to fully update, so please check back in a bit.

[pulsar coral]

Really need the daily limit exhaustion message fixed.

But yeah I'd say Flow's even working better than before atm.

[sacred mica]

still have this message...🥲
Failure
You are requesting generations too quickly. Please wait, then try again.

[pulsar coral]

Yeah you did probably hit the limit some hours ago, can't be helped atm. Gotta watch the clock now.

Specially considering your account was Pro and you had chance to exhaust them unlike free ones.

[fading elm]

When will Flow become available in the Indian region?

[frosty niche]

https://tenor.com/view/yoda-star-wars-gif-16477632884420340106

[astral fern]

Sora will be discontinued as of April 26th. Whisk is discontinued as of April 30th.

Coincidence?? Probably.

[wintry solar]

damn the free 100 credit got finished so fast while editing it 5 times. anyone have some more tools to suggest

[wet flame]

haiyaaa

[wintry solar]

Gemini will win the race. Any Chinese tools news?

[wintry solar]

is this Kling?

[wet flame]

This is Flow bruh

[wintry solar]

is there like "your stuff" options check there it be there

[chilly frost]

Not to long a go i found out Yoda's full name, it's Yoda la Huu

https://klipy.com/gifs/cur-tta

[frosty niche]

What that mean?

[wintry solar]

Video generative tool.

[frosty niche]

https://tenor.com/view/sunrise-good-morning-cool-beans-gif-12979224

[wintry solar]

you use which tools

[frosty niche]

Every tool known to man & woman & everyone else. 🙌

[wintry solar]

suggest me some

[frosty niche]

Flow.

https://tenor.com/view/happy-gilmore-feeling-the-flow-working-it-positive-vibes-adam-sandler-gif-1577802084857617369

[wintry solar]

King Flow indeed

[pulsar coral]

I hope you gents aren't too deep into a philosophical existential talk atm, cause I got one last rant about Flow today and wouldn't wanna interrupt.

[frosty niche]

Go on....

Bust'em up

phylosophical existential <- what these big words mean?

[pulsar coral]

Thank you. Ok then, so I just remembered my friend Kitastro here somehow mananged to get Flow's Nano Banana 2 to work like original Nano Banana, using a silhoutte prompt.

It was somewhat simple but bright. So point is, it proved NB2 somehow somewhere has NB1's data well stored, or as a part of it's model knowledge.

So I went like what the hell, and started trying prompts like this

[MAIN RULE AND OBJECTIVE: Only use/reproduce the original image model "Nano Banana (Gemini 2.5 flash image)" style and compositions (Specially....) to create this image, it's forbidden to use the "Nano Banana 2" "Nano Banana PRO" image models in any style or way to create this image so completely avoid them]

At "Specially" one can add something specifical that Nano 1 gets better than NB2

And tbh it doesn't work all the time, but I'm indeed getting what I want at least 2 outta 4 attempts.

It might not sound possible, but considering Nano Banana 2 knows about lots of things pretty much similar to a search engine, yeah it surely knows itself, and there must be so many Nano 1 images on the internet now.

I don't think the images I generated are so good to demonstrate this yet, cause I'm just seeing what I want in them I suppose, but as soon as I can apply it to something more general I'll show it.

And oh man for me sure it's working

[frosty niche]

I made a doc kind of seeing where you're coming from, feel free to add to it.
I mention @native lake hope thats cool.

G-Doc: https://docs.google.com/document/d/1P6G_0qsvmFmSTzS2SWMZsfpsdfYbVMYr5a_CEnBf1XE/edit?usp=sharing

[pulsar coral]

I think he'll like it

Man Flow works so nicely today... cept maybe for that poor dude from Bangladesh

[balmy birch]

Can Flow do video animations as well, like starting and ending frames? Openart has rate limits, which are annoying

[pulsar coral]

It does have that, but even without it I've seen as long as you place your scene in correct coherent order, it will work even without an end frame

[balmy birch]

Is it free? 🧟

If yes then how many?

[pulsar coral]

You get 50 free credits daily, and each VEO fast quality video costs 20 credits, 100 credits for more quality.

[pulsar coral]

https://one.google.com/u/0/settings?expand=upgrade&g1_land&g1_landing_page=0

So by that measure Ultra Plan gives 25.000 credits, I'd say it's worth if you're really into a business about it, but according to each country's regulations about AI the videos could have a "VEO" watermark even on Ultra plan

If you're in USA or Malayasia as far as we've seen, Ultra plan videos won't have a watermark, so you gotta read that carefully.

Gotta mention it cause there could be still be some old info about it somewhere, from when Ultra Plan videos would just never have a watermark anywhere, it's not like that anymore.

[wintry solar]

how long was the animation at open AI?

[balmy birch]

Oh I meant open art my voice to text AI got it wrong lol

[wintry solar]

what is openart

[balmy birch]

Same video generation but more control

[native lake]

Hi
Got pinged. Btw I think the model creators like putting in this self knowledge Easter eggs

[frosty niche]

Oooh boiled, pickled, or scrambled?

Also, hello.

[native lake]

Hello 👋🏼

[hearty sable]

Does somebody also have this problem that generations stop at 99%?

[eternal arch]

is flow fixed?

[frosty niche]

I was just getting grainy outputs just awhile ago too, as soon as i was about to get done too. Go figure.

[eternal arch]

WAIT!

i think flow is fixed, i just did 27 generations and they all worked, lets goooooo

[plain finch]

No wayyyy

Finally

[eternal arch]

ikr

[shy nest]

https://tenor.com/view/sigma-you're-sigma-you're-totally-a-sigma-sorry-you're-not-a-sigma-not-a-sigma-gif-16857136571091130544

[proper jetty]

Anyone having issue generating video? i keep getting failed ,something went wrong error

[halcyon drum]

Flow not creating images or video on ultra plan!!!!

[proper jetty]

same error on Pro plan

[balmy birch]

How much is the paid version of that? Like, how much do I need to pay for a monthly subscription, and is it actually better than OpenArt?

[pulsar coral]

You can check the plans there and compare them I guess, but once you pay for a plan you won't get the free 50 daily credits anymore. More often when it's giving you an offer is good time to buy something. https://one.google.com/u/0/settings?expand=upgrade&g1_land&g1_landing_page=0

Also besides Flow the same google account also links you to the Gemini page which also allows to generate images, videos, canvas, music, and use the Gemini chat models

So whatever you get there, becomes at least double in the same account.

https://one.google.com/ai?utm_source=gemini&utm_medium=web

So that's good place to check too

[frosty niche]

Morning, mornin,
I’ve been using Nano Banana 2 in Flow to push a single visual idea through multiple “requirements” / sets and then analyze where the images actually get better vs. where they start to fall apart.

To make that easier to study, I turned one of these runs into a small dataset + analysis notebook on Kaggle:

🧺 Dataset: Never-Forget-The-Basket: Multi‑Arc Generation
https://www.kaggle.com/datasets/gastondana/never-forget-the-basket

📓 Notebook: Multi‑Arc Generation – v1-Giant-Chicken-Nugget-Sandwich
https://www.kaggle.com/code/gastondana/multi-arc-generation

All the panels in this arc come from Nano Banana 2/Flow, but I treat them like a storyboard in versions:

Requirement/Set 4 is where the model really locks in: composition, storytelling, and the “summit” scene finally feel stable and intentional.

Requirement/set 6 is the opposite case study: extra changes introduce noise/pixeling, and the visual story drifts even though the prompt is more detailed.

The notebook lines up the images with a few simple plots so you can see the refinement curve, where extra guidance helps, & where it actually breaks continuity.

If you are tracking your own “best set” vs. “over‑refined” outcomes, whats your method?

[unborn bolt]

Im Pro Plan. Does anyone know if Nanobanana Pro has a limit? I waited 12 hours and still appears that im generating too fast. Does this resets after 24 hours?

[hearty sable]

why are my generations all stopibg at 99%??

stopping

[fathom coral]

hello

[frozen totem]

flow broken? every image i generate get something went wrong every account.

[glad yacht]

Just curious, how many images did you generate 🤔?

[fallow narwhal]

Me too, however it’s happend after I used my iphone; it was about 19 o’clock, but banana 2 still working on either phone or pc, I’m so curious.

[unborn bolt]

Between 25 and 100 images, first time using that model

[fallow narwhal]

So did I

[wet flame]

Yeah it's broken as usual (as for this week)

[fallow narwhal]

Got that, but is there a diffrence between pc or cellphone? Like using mobile version website may cause some error?

[wet flame]

It's still based on the same system

[fallow narwhal]

Okay got that, Thanks

[pulsar coral]

It's more likely either on your phone or PC you're using a different image model, both have a separate daily limit. Nano Banana Pro gives less daily images than Nano Banana 2 because it's more expensive to generate with it but somewhat more quality.

But yeah gotta try swapping the models and either of them might still work for today.

And if both fail yeah just gotta wait till next day, it usually starts working again around the last time you successfully generated an image.

[pulsar coral]

Yeah around 100 sounds about right for Pro plan, they must just be exhausted for today.

Free accounts are giving 10 more or less daily Nano Banana Pro images, way more for Nano Banana 2 also.

But if you've been getting the "You're generating too fast" message for like 3 days in a row without getting even one image, better just report it at the labs-bugs-feedback area, something'd be surely broken there. And it still does happen to some people.

[unborn bolt]

Thank you!

[quiet wadi]

It´s nano banana pro more succesfull than nano banana 2 ?

I have created images with both version´s but I feel that when I use nano banana pro the image has more fidelity and details rather than the 2 version

futhermore the nano banana 2 has no limit to create but the pro version has it

[pulsar coral]

Even tho they're advertised as separate models, yeah Nano Banana "Pro" in the end is pretty much just a more juiced version of Nano Banana 2, around %10 more accurate. Nano Banana Pro spends more resources tho, so you get less daiy image generations for it.

That's also why at Gemini they didn't even bother putting it as two versions, they just leave the "Pro" mode there for Nano Banana 2, which is pretty much the same.

[quiet wadi]

so Nano banana 2 is worst compared to the pro version, and has lower details and less coherence with the images

[pulsar coral]

Yeah tho not by much, like said around a accuracy margin %10 margine. But Nano Banana 2 somehow still spends about %50 less resources without Pro mode, so you get more of those.

Once you get more practice at prompting there really isn't that much of a margin. Pro pretty much just thinks more for you.

[quiet wadi]

and the only solution is to purchase a subscription?

[pulsar coral]

A subscription will give you more daily tries accordingly yeah, I think my $5 account gives me like 30-40 pro images daily.

While a free account seems to barely reach 10 pro images if lucky

[quiet wadi]

before nano banana 2 appears , the pro version has 100 images at least

[pulsar coral]

Yeah I think the $20 account does reach that, from what others say

Haven't tried that one in a while

Oh also your same Google account is also linked to Gemini, so basically you get double what you pay.

Gemini images have the spark watermark tho, but if you always get the image with the download button you can remove the watermark with this https://geminiwatermark.org/batch-watermark-remover

Because if you right click-save the image as it will only save jpg preview, and the remove won't work at that one.

[rich holly]

How should I write a prompt to make a person transparent or delete them?

[pulsar coral]

Gemini also has the advantage that the image aspect ratio can be controlled by prompts, you end your prompt like "Aspect ratio is 3:2" or whatver reasonable and it will listen. Flow can't do that.

In fact I think it listens to all sorts of things about size by prompt

[quiet wadi]

so Gemini is better workspace than flow¿?

[pulsar coral]

Nah it's slower since it's harder to generate more images at once, but it can be good for something specific about an image

[quiet wadi]

I mean that if is better at the quality, the details, the coherence, etc

[pulsar coral]

And it gives something else to do once your account at Flow goes dry, for the money specially

[pulsar coral]

No but you also get a little more pro mode images there. They have the same models.

[rich holly]

It's difficult because I'm getting an error while trying to delete only people.

[pulsar coral]

Arguably what's still better than both Nano Banana 2 and it's Pro mode according to Whisk users, is the "Precise mode" from Whisk, which is just the first original Nano Banana.

NO DONT CLICK THAT

That was fast

[rich holly]

oh!

thank you

[pulsar coral]

That's not for you be with you in a minute, ok now...

[rich holly]

yes

[pulsar coral]

If you want to delete a person click on the Flow image and it has a select tool there, select what you want gone and prompt it away

Doesn't need to be made invisible

Also more often you can give the details about the person and also delete it by prompt but yeah sometimes it won't listen or recognize what you want

[rich holly]

ok

[pulsar coral]

So might wanna put some visual pointers about the person in the prompt, like if it's blonde, wears whatever color of shirt, etc

Then just say remove it, either will do something

[rich holly]

I succeeded after using the tool.

[pulsar coral]

Yeah the way Nano Banana 2 listens to that tool is one of the best update advantages of the model... can't deny that

[tardy bridge]

good

[frosty niche]

https://tenor.com/view/jay-renshaw-chit-success-you-did-it-congratulations-gif-18093781729549550194

[glad yacht]

Gemini is less restricted than flow too.

[plain finch]

Can Gemini make the same sort kf videos for free tho?

[autumn sluice]

In my experience, I've only made one video with Google Flow, and it was one of my OCs just saying hi. I prefer pictures anyway (mostly of Pokémon as Sonic characters, and it actually works.)

[brittle hearth]

again?

[ocean stump]

@brittle hearthcreate a ticket to gaiud you on mate

[brittle hearth]

okay i will

[brittle hearth]

i opened the ticket

[brittle hearth]

i got kicked

worst support ever

@ocean stump

[wintry solar]

at which 0000 time flow daily credits get reset

[crisp basalt]

EST

[rotund horizon]

I already mentioned it before and I will say it again. Removing NanoBanana1 is a bad idea. A simple task here. "Elaborate a bit more on this sketch. Do not change the original design" It took NanoBanaPro and 2 a total of 11 attempts and still, after refining and adding further instructions, it got it ALMOST right. I got frustrated, jumped back into Whisk and with the first attempt, it gave exactly the result. And no it was not a coincidence, I tried a few more times and it got it right again.

BananaPro and 2 ability to 'think', unfortunately is making them go overboard many times. Is like you asking someone, "Please pass me this cup" and goes to the tap, filling up the cup and pass it to you. NO! I don't want you fill up the damn cup!

I don't understand how the developers can't see that. Or at least, if you don't want to bring Banana1 back, fix the 'thinking' behind the other two models.

[quiet wadi]

I would like to learn prompting seriously , but don´t know how to start

[rotund horizon]

What do you usually create?

[frosty niche]

Its "Elaborate more 'of' this sketch" btw.

[wet flame]

:/

[nova ermine]

I've just started exploring Flow. How long can we make?

Also, is Veo good at timing the video as per script?

[wet flame]

It's AI at the end of the day, I'd say it's quite accurate

The video length is 8 seconds long, with sound. Costs 20 credits for fast (imo it's already good enough), and 100 creds for Quality

[nova ermine]

Thank you! I'll experiment more. I'm really happy that even Veo Fast is giving a high quality

Haven't used Quality yet as the cost is 5x

[wet flame]

Also depends what type of content you're trying to make, like speech is quite good, you can take this video as an example

I used ChatGPT to make a base prompt cause I'm not the director of the image, and no idea what's the script I want, here's the prompt:
A confident young woman with deep brown skin and natural hair in a high puff, wearing a colorful patterned wrap and hoop earrings, stands in a sunlit field of tall grass, looking directly into the camera with calm presence as she speaks naturally in a soft, grounded tone, subtle facial expressions and eye contact, gentle breeze moving the grass and fabric, shallow depth of field, warm golden-hour lighting, realistic skin texture, handheld camera feel with slight natural motion, human-like micro-expressions and lip sync, cinematic realism. Her dialogue is: “Not everything needs to be loud to be powerful.”

Seeing there's no prompt to add background music or emphasizing her correct accent, I'd say Veo did a pretty good job

[olive thistle]

Is it possible to prompt fast paced scenes? Like goku vs vegeta level of quickness. I tried to do boxing scene but still slow punching speed, the action is not apparent

[wet flame]

Not quite, those scenes often require post process

Since AI usually generate frame by frame, if the previous frame is blurred, the next frame will basically screwed as well

[nova ermine]

Amazing 😍

Can't tell it is AI generated. It's so real!

[agile girder]

There are three main sections of language:
The first section is natural language prompt,
The second is TAG prompt,
And the third is the .json format prompt.

Also you can combine 2 or 3 of them

As for learning, there is nothing magical, it is only done through the learning curve, through experience you will have a quantity of information that will help you

[wet flame]

🥀

[agile girder]

Are you subscribed to it or use it for free?

[agile girder]

Maybe

[agile girder]

So why is there a big difference between the size of Gemini and Flow?

The size of the Gemini image is three times the size of the Flow image

[crisp basalt]

I also haven’t found a reliable workaround for this. Veo 3 seems to struggle with generating normal-speed footage. Even when I specifically prompt for high-speed or real-time motion, it often produces slow-motion results

[crisp basalt]

Did you download 1k version of that image in flow

[agile girder]

No!!!

[crisp basalt]

Then what button did u press to download that image

[agile girder]

I down 2k image

[crisp basalt]

Why not 4k

[agile girder]

Bro i know how to use it im pro in AI and photography

[agile girder]

Because the 4k just for ultra subscribers

[crisp basalt]

Anyway

[crisp basalt]

lol

I forgot about that

[agile girder]

Also even before 4k i test it alot

[crisp basalt]

https://tenor.com/view/blocked-block-you-blocked-blocking-gif-14145534

weirdo

[agile girder]

From beginning of gemini nanobanana, always gemeni image 3 times bigger then flow, anyone can test that

[agile girder]

Do you have ultra plan?

[wet flame]

I am subscribed. Does it matter?

[crisp basalt]

dont ask him anything or he will get triggered and accuses u of sayin that hes not professional in ai 😔

[wet flame]

Understandable

[agile girder]

Because i don't know the deference in the sizes between 2k gemini and 4k flow, also the quality

[agile girder]

I said professional so that we don't go into the trouble of where I pressed and how because all my life I do this thing nothing else, I hope you understand what I mean

[wet flame]

The image size and the details, I guess

Since just purely making the image larger will make the image blurry, so it technically regenerate the image and make smaller details more apparent

[agile girder]

Also the dynamic range and colors

[cyan dragon]

anyone having this problem trying generate video saying "promt must be provided"

I do have img for reference and written promt

[wet flame]

Wat

Really?

[wet flame]

It's good to upscale if the image is gonna be used commercial-ly

[wet flame]

Also I think you forgot to answer this question

[agile girder]

I do

[agile girder]

.

[wet flame]

._.

Oh so you're not asking about me failing my generation in the condition of subscribed or not

Okay

https://tenor.com/view/blocked-block-you-blocked-blocking-gif-14145534

[pulsar coral]

I've also tried making this type of mechanical sketch but for an x-wing drone and did post it around here, Whisk's original Nano Banana was also much better at it. Nano Banana 2 even in Pro mode would always make it's wings short completely ignoring aerodynamics or just create some sort strange junk scrap metal.

And yeah notice how it's not just me saying it. Ends up Nano Banana Pro wasn't even a separate model but just Nano Banana 2 with some more juice to it, so that's why results are samey.

People around here more often would prefer Nano Pro tho, unless they tried Whisk's original Nano.

it's not like I made just 5 images, tried around 40 times with the drone

[rotund horizon]

Yes, this is known already. I think that Banana2 has the ability to check online and self adjust the image based on the information that retrieves rather relying only in its training data. I think that's the only difference between Pro and 2.

[pulsar coral]

Even if my prompting wasn't perfect, original Nano Banana did understand how to make 4-wing design much better

[rotund horizon]

You have to manually sketch on its existing generation and see if it can correct it. As usual, it does not always work.

[pulsar coral]

I imagine it's cause Flow is centering on making vids and images, Gemini also has the canvas mode and chat models. But haven't seen much difference on how it generates images. But yeah there might indeed be a little difference on the censorship levels, I've actually struggled much more with that at Gemini than Flow tho.

But that might also have to do with how one's able to generate like 12 images at once at Flow, you'd get luckier there.

That's still not beating the way Whisk does it tho... unless you're going for text.

[ocean panther]

hi chat

[pulsar coral]

Gemini's canvas mode might seem like nothing, but the more complicated things you try to make there, the more it will surprise you.

[ocean panther]

why the chat is so dry 🥀 21,075 people are online

[pulsar coral]

Probably general and off-topic move more. You need to be more specific about Flow stuff here.

[ocean panther]

okay

[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

Nano Banana Pro is giving me much better images today... then again yesterday I was playing with a prompt to try to make images more like NB1 does and it's still in the image's prompts... results are a little inconclusive yet

But I'd say they're mostly better

[zenith forum]

again, it doesn't work

[hearty badge]

is Flow restricted for Paid accounts now ? I am getting Errors can't generate any images

[agile girder]

I test now it give me errors on free account but in pro account its generate image, also if i use vpn it give me errors and it's work without it, try to use vpn

[unreal solstice]

Yow guys, is the watermark for Ultra users still not fixed?

[agile girder]

By the way, i can generate more then one image per prompt in Gemini

[paper tulip]

Hello Everyone

[agile girder]

Also you can give gemini instructions so everyone time you type 2x or 4x or like that it will generate more image basid on the number with the X letter

[agile girder]

Hello

[agile girder]

Do you have an idea to test images as a test model to do so, because I've been successful in the past with some images

[lucid wadi]

Yeh raha aapke Karbala-themed video ke liye ek soulful voice-over prompt, jo Urdu mein emotional aur roohani lehje mein record kiya ja sakta hai:

---

🎙️ Voice-Over Prompt (Urdu):

"امام علیؑ نے فرمایا... جاہل کے سامنے دلیل لانا آسان ہے... مگر جاہل کا اُسے قبول کر لینا... بہت مشکل ہے۔"

(وقفہ، پس منظر میں ہلکی نوحہ دھن)

"یہ قول نہج البلاغہ کی حکمت نمبر 117 سے ہے... اور آج بھی... کربلا کی سرزمین پر... حق اور جہالت کا یہی معرکہ جاری ہے۔"

(پس منظر میں اذان یا زائرین کی صدائیں، ہلکی روشنی کی جھلکیاں)

---

Aap isay soft echo ke sath, halki noha-style music aur Karbala ke visuals ke sath blend kar sakte hain. Agar chahen to main iske liye full audio script ya background sound suggestions bhi de sakta hoon. Bataaen, agla step kya ho?

[turbid umbra]

sad day today lol, last day of my ultra subscription

[edgy holly]

Usually.

[edgy holly]

NB2 is cheaper. It's a 'Fast' model.

[edgy holly]

For simple generation, yes. Image gen in Flow is designed specifically to create keyframes for AI video generation.

[edgy holly]

No, they're the same.

[edgy holly]

A few a day.

[edgy holly]

Google for prompting systems for the specific kind of AI model you're using, and limit the search to content put out in the last few months.

Find several of these systems. Spend a few weeks testing each of them out, step by step. Break them apart. Keep the parts you like. Build your own system, and take notes on what words, phrases and techniques work.

Never stop.

[edgy holly]

Seedance 2.0 is the only video AI model that does fight scenes kinda okay. None of them are great at it.

[turbid umbra]

tried seedance recently its kinda dissapointing comparing to veo for like quality prompt adherence and talking in different language.

veo is still king in my opinion

[edgy holly]

The default image size on Gemini is 2K. You have to choose to upscale in Flow to get larger resolutions, if you have that functionality.

[edgy holly]

Free plans don't have all the functionality in Flow that paid plans have.

[edgy holly]

Fails spike during peak use hours.

[edgy holly]

It's too expensive for my hobby projects at the moment.

[turbid umbra]

also the price yes, veo still king, haven't tried kling but im guessing its not cheap either.

[agile girder]

Not default because if i get tge image by save as, it will be 1k unless i use download sign on the image
Also the sample image in my image i send them gemeni and flow they both 2k res but deference in MB maybe because dynamic range or something else

[edgy holly]

If you get a higher account, Kling is roughly on par price wise per generation with Veo in Flow if you're using paid models, but at the moment nothing beats an Ultra account per clip, if you generate a LOT, because of the "free" gens on slow mode.

Each model has its strengths. Seedance is overall probably the best, with Kling second, and probably Veo 3rd, but a month from now someone else will be on top. They leapfrog each other every ~3 months or so.

[turbid umbra]

yea i would guess google I/O will bring some news on this front

[hearty badge]

thanks I wish google had experience programers who could write a proper error message to not let their users guess why things fails. Or maybe this is also an option I have to pay for.

[edgy holly]

I/O often hints at big things coming, but usually isn't when they roll them out. It'll be fun to watch. Usually is 🙂

[edgy holly]

That's a choice, not an inability. UX is generally awful among all the AI companies.

[hearty badge]

My earlier comment was meant to be a bit sarcastic, but I do appreciate you coming back. Whisk actually did a decent job.

[pulsar coral]

Ah right, in the same way Google AI studio does... it's still not the same way Flow does tho, Flow's is similar to Whisk about it.

[edgy holly]

They could incorporate the exact same thing into the Flow image editor. We'll see if they do.

Most other AI image apps have targeted image reference types (Midjourney's Style and Omnireference, Leonardo has several depending on the model you choose], all the top 'All-in-One' art gen platforms have the ability to set an image as a character reference, and some are adding more advanced versions of it).

Ever since Imagen and Veo first came out, Google always has either THE best base model (Nano Banana Pro), or at worst a contender in the top 5 (Veo 3.1), but their consumer apps are terrible compared to the competition, and their focus is more on serving them through the API. No other model has so many bugs, and especially nowhere near as many gen fails.

[pulsar coral]

A free Flow account is still working on my end atm, until it runs out of free images for the day, but it still wrongly displays the "you're generating too fast" error message once it runs out of image generations.

Same happens to payed subscription accounts, if you run out of images for the day you'll also get that wrong message about it.

If you were generating images a while ago the account will be fine again around the last time you generated images, but if it's going for three days or so now you better report it at labs-bugs-feedback area with the region you are at.

Tho yeah some free accounts could still be getting an "oops" message instead of generating the images, payed plans more often are working like others say.

[pulsar coral]

That depends on the laws about AI in your region. It'll probably stay like that until something changes there, or globally.

United States and Malaysia for example don't get the VEO watermark for Ultra plan.

[pulsar coral]

Eitherway it's the weekend atm, so there might not be too many people working on it, might wanna give it until tomorrow.

[green breach]

Our flow ai is not working,not generate any video

[wet flame]

Seems like the generation issue has been fixed, at least on my side

[wet flame]

nvm

[iron quest]

Anyone else getting upscale failled when trying to generate 4K videos atm?

1080 seems fine

[wet flame]

Failure chance is about 90% for me on image and video

[nimble marlin]

me too

[robust socket]

can we download Transparent PNG please, given a plain white background, give me a transparent PNG please

[stark void]

Try a new style video via Veo with only one prompt
https://youtube.com/shorts/mC_HlrEimvQ?si=n6K0raUF3Pozhb_S

[summer garden]

keep getting today this error in Google Flow AI video on all poast and current videos: warning

Failed

Something went wrong loading your media.

anyone getting it or know this kind of issue?

[summer garden]

same on my account

on all my past and new videos, they just wont play and showing this error

[sour vapor]

Hi. Is Flow capable of generating Nano Banana images in 2K or 4K? There's no resolution selector anywhere (I think)

[barren mulch]

the option appears in the download options

[candid bay]

i wonder if there's any difference between default 2K generation and upscaling to 2K afterwards

[dry frost]

Downloading 2K from Flow is upscaled as well.

[harsh ibex]

https://www.instagram.com/reel/DWgrvQeiWxk/?utm_source=ig_web_copy_link&igsh=MzRlODBiNWFlZA==

yourprimeage

Part - 1 tom vs oggy

Instagram

[candid bay]

Anyways, I guess we are back to "Oops, something went wrong" errors... I can't believe this

[cyan dragon]

yes

[opaque sand]

For the past hour, I've only had three options: 39 out of 40 rejections for using famous people (even though they're completely faceless, generated images), a rejection for missing audio, and a rejection just because...
The “Ingredients” mode is completely dead.

[round relic]

Is anyone else experiencing this getting stuck loading here for a while? Also, my saved projects aren’t loading.

[cyan dragon]

Banan 2 - give me error something wrong
Banan Pro - I generate too quickly 👀

[thick condor]

Same huhu

[coarse pebble]

😄

[candid bay]

you were being very persistent lmao

[coarse pebble]

https://tenor.com/view/mr-bean-mrbean-bean-mr-bean-holiday-mr-bean-holiday-movie-gif-3228235746377647455

Clients waiting

[slow pawn]

does anyone have best prompt for create a image or recreate style in nb2?

[coarse pebble]

I alwasy tell chatgpt / gemini what i want to do and they make me prompt in json

[candid bay]

Upload image to Whisk and use the description it gives you or ask Gemini to reverse engineer the prompt for it

[slow pawn]

will try that

[wet flame]

🤝

refreshing the page helps sometimes

[coarse pebble]

Not this time ^^

[wet flame]

yeah.. sometimes

somehow the failure rate is either 25% or 100% today

[elder wadi]

So Flow generation is completely broken today?

Every time... I generated 4 images early this morning, after that it's not letting me generate anymore...

[pulsar coral]

If you were using Nano Banana Pro that one runs out much faster, you could try swapping to Nano Banana 2 , maybe that one still works. If not yeah just gotta come back in some hours. Flow just doesn't display the correct message for when you run out of images for the day.

[coarse pebble]

I can't even download images with higer quality, just saying "failed to download"

[tepid cedar]

100% failure rate since around 11pm central last night for me.

[unreal viper]

Hello as a friendly reminder, videos and images are not generated within the server, you'll need to head to https://labs.google/, pick a tool, and generate it there. If you are asking for prompt advice please do so in the appropriate channel. Thank you!

[edgy holly]

From what I've been told, all the first one means is the upscale happened before you saw it first.

[queen matrix]

is there any free credits to try flow out?

[woeful void]

hi dudes,
how often does Banana fail to understand your basic background replacement commands ?
like "replace the background from the second photo in the first photo"

Its get it right about one time out of ten. pff lol.
A few months ago this worked much more better.

[frosty niche]

I've been experiencing the same, lots of outputs means lots of "Compute Usage".

[umbral jasper]

funny that you mentioned that because im struggling with the same right now and it didnt use to be like this

[proper drum]

I'm having an issue... I have an original character of mine (AI-generated), but I can't generate a video because the system thinks it's a real character from a well-known company.

[candid bay]

That's not true I think.. Upscaling is always an afterwards thing

Generation in 2K means the image is MADE in 2k. But idk if there's any true difference in quality and detail

[haughty bone]

Ive been getting this constantly for the last couple days. anyone got solution for it?

i cant upscale the videos that go through either.

[bright heron]

has anyone from the Flow team talked about generations failing this week? I used to have almost no problems but every day for this past week I've either gotten 100% fail or one generation out of the 3 I am trying to do at a time. It's almost unusable at this point

[obtuse cipher]

This happens to me a lot, I dont know what happens, even though I pay gemini pro

[bright heron]

same

[haughty bone]

fahhh its annoying. im paying ultra and getting this

[haughty bone]

im getting the exact same issue. If anyone knows a solution, ler me know please

[pulsar coral]

tbh Flow is working fine on my end atm, but yeah gotta have a lot patience with and just come back right around the same hour when you managed to generate an image, a complete 24 hour cycle.

However in some regions it can indeed be failing no matter what, if you're really been having probs for around 48 hours best thing you can do is make a thread a labs-bugs-feedback and say in what region your experiencing the prob and if you have subscription or not.

[bright heron]

@sick trench what plan subscription do you have? I have pro but my colleagues are getting fails with ultra too. region is Texas USA

[pulsar coral]

Just a simple $5, account it's currently working after 24 hours, I was generating images like at 6 pm yesterday and I started getting the error message, it was fine again around 4 pm next day. And I'm in Ecuador South America atm.

Also remember Flow is not displaying the correct error message for when you run out of image generations for the day, it insted says that you're generating them too fast or sometimes gives "An error has occured" one.

If you were generating images just fine some hours ago or yesterday and ran out of images it should be displaying something like "You ran out of image generations for today, come back tomorrow 🍌 "

It used to be like that, but now it instead displays that you're generating images too fast or whatever else.

Also yeah according to reports more often payed plans are working better, but according to region not even those might be free from errors.

[bright heron]

with my pro account it should be unlimited 1K and 2K generations with nano banana pro and 2 but out of nowhere it started failing almost every one

for a minute it would tell me I'm generating images too fast but it would tell me that after I've given it a 4 hour break

[pulsar coral]

https://one.google.com/u/1/ai?utm_source=gemini&utm_medium=web&utm_campaign=zero_state_banner&g1_landing_page=75

Yeah that might be it, I've had Pro plan before and it wasn't unlimited, idk who and where is getting it like that

It does make a lot more images for sure tho.

[bright heron]

I mean they use zero credits so I'd assume unlimited, it doesn't specify a limit

[pulsar coral]

In fact I didn't need that many so just went for the $5 one

[bright heron]

I have Pro as a student so at least I am not burning money with a broken product but I'd be very frustrated if I was paying monthly, especially if i still had ultra

[pulsar coral]

Yeah they don't really tell you how many images exactly, but from reports in Pro plan they can be around 100 or more for Nano Banana Pro, way more for Nano Banana 2

But you won't be there all day for sure, but close nough

They just leave people guessing for some reason, and users have indeed asked for a better image counter.

Eitherway if that was it your account will be fine later for sure don't worry, you just gotta measure your workflow some more.

Also yeah Nano Banana Pro images will run out much faster, just in case do always check if Nano Banana 2 is still working, gotta be swapping.

[bright heron]

Yeah I was using 2 for awhile and went to pro. now Imaged 4 won't even load anything

[pulsar coral]

Yeah once you run out of Pro and 2 images, imagen 4 will stop working too. The three models are just tied somehow.

In fact Nano Banana Pro is not really a different model, just a 10% more juiced Nano Banana 2.

That's why you get less of those, it just costs more.

[bright heron]

makes sense

I wish they would clarify that you get limited generations rather than keep it undisclosed and ambiguous, shady business model for a multi-billion dollar company

[pulsar coral]

They know it's not displaying the correct message anyway, it started days ago, I don't think they can fix it atm so can't be helped.

[empty ruin]

Does anyone know hwo to bulk generate images with nana banan pro? I havem ultiple prompts

[empty ruin]

I;m trying to generate multiple images in bulk, if theres an option for that api keys etc or a way to do it with flow please lmk!

[empty ruin]

Do you happen to have an answer?

<@&1366831844502274180>

[analog lava]

Hello, mods are only here for discord support we are unable to help with product issues. Please only ping mods if you see someone breaking our #rules-and-guidelines・📋 . There is #1238621511401734254 to submit the bugs you find or you can DM @real wyvern .

[little fulcrum]

today I can't avoid failed generations for bathing suit scenes (adult). Usually its not this strict. Is this going to be a problem? Even simple prompts like this fail with warning for "harmful content". I promote swimwear. "Hyper realistic POV shot. A woman in @image
with long wavy blonde hair wearing a straw hat relaxes on a poolside lawn chair. She holds a camera filming herself in the sun. She smiles quietly and naturally shows off her outfit. The camera moves with natural handheld motion. Sunny lighting with pool visible in background. Ambient background sounds. "

using ingredients like this:

[pulsar coral]

"Bulk" tools pretty much only autoclick the image generation button afaik. You get the same result from just asking Flow to generate 4 images per batch and finding any autoclicker tool that can be positioned right on the image generation button,

Just be sure you only set it to repeat about 1 minute to 50 seconds each because Flow won't generate faster than that, idk why'd people use those tho.

[empty ruin]

When I put 4 prompts and click the 4 images per batch it just generates them in a collage sort of way

forgive me if i misunderstood

[pulsar coral]

There's some old chrome extensions that can do that too, but more often they're outdated from when Flow still had original Nano Banana and just do the same.

There's also some sites that do have some sort of auto batch generation tool but that's internal for them and more often use credits.

But Flow itself doesn't have it.

[empty ruin]

damn, half of these google extensionms dont even work

and or have a pay wall

dont know if its bc im on opera or what

[pulsar coral]

The "Turbo flow" chrome extension used to do that, but it'd just auto click too, if you think you can tame the thing yeah it's at the chrome store, but like I tell ya it's just extra steps

[empty ruin]

so you wouold recommend js typing them in 1 by 1?

[pulsar coral]

No if you want the same results find an auto clicker tool and position it right over the button that generates images at around 1 minutes interval, and set it to 4 images.

Pretty much everyone around here seems to be doing the same manually anyway.

Your Flow account will stop when it will stop.

[empty ruin]

i have ultra dk if that affects the thing anyway

[pulsar coral]

No but more likely you'd prefer to be there to tweak the prompt and be sure to get what you want, bulk image makers weren't gonna do that for you either.

[pulsar coral]

I've generated images with 2 piece bikinis and Flow doesn't have too much probs with it, I think it's the prompt "Shows off her outfit" gotta try and avoid being ambiguos because yeah the AI models have really bad thoughts about everything.

It might be thinking she's taking it off or something

[bright schooner]

when will the next gen veo coming?

ultra is very expensive, but i rarely use veo now because its fell completely behind atm\

[tepid cedar]

Did anyone else have things just stop working again completely?

[tough spoke]

damn, this is straight outta Jurassic World

[quiet kindle]

Is there any way to recover or view images after they’ve been deleted in Flow? Or are they permanently gone once deleted?”

[pulsar coral]

Yeah they are gone, I did ask for a recycle bin feature but who knows if they'll ever do it.

[fluid abyss]

have you guys been trying seedance on dreamina?

which country has the cheapest generation cost? ive checked a few countires and canada is coming out cheapest for me

[broken kettle]

error for upscale?

[sick sparrow]

yes error for upscale (again)

[gloomy verge]

Hello? I need help. Can I ask, why, when I write prompt only "Close-up" it create close-up image, but when I write: "Female with pigtails wearing business attire stand near car, close-up". no close-up? I create 8 images of secreatary and only one of them was close then other. Why is he doing it this way? Look. (I both create by NB2) Man is created only with prompt: "Close-up" and female is create with prompt: "Female with pigtails vearing business attire stand near car, close-up". I don´t understand it. Can you explain this to me? Thank you for help.

[sick sparrow]

can you try 'now an extreme tight crop close up of the woman's face'

[gloomy verge]

It is better.

But still some image is better close then other. They are not all the same as left image

Is it right that all images are not same close-up? Some is closer then other. I thought that all images will be same as left.

[coarse pebble]

NBP and NB2 down again? Every generate is failing

[pallid violet]

@wary willowHey! 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!

[vague island]

Hi - can anyone else generate stuff in flow at the moment? Seems to be broken right now for me

[foggy tree]

nanobanana 2 flow is down again?

[foggy tree]

i think yes

[stone garden]

:/
Is there some problems atm?

[tough spoke]

i have the same issue since few days now

but only on brave browser

Edge works fine

[stone garden]

Really? I was on brave too.. Let me check that out

Yeh, works on the other browser.. Is it due to 'cookies' and making sure they're placed 🤔

[placid wing]

im using brave too and i have error here!

[cyan dragon]

here we go again

[empty dove]

Can't even go 24 hours without this thing breaking. Getting rather ridiculous now

[placid wing]

brave, firefox, edge, getting error too. its not browser

[stone garden]

Switched to firefox and no longer had problems since ( been about 40mins no problems)

[placid wing]

brave is fine now, i think the problem was the flow

[cyan dragon]

im with firefox i have the problem

aint browser problem , servers most likely are overcrowded

[burnt oriole]

ow now we have veo lite

[little fulcrum]

I would like to be more specific but I found that mentioning swimwear, bikinis or swimsuits makes it fail. If I say she is wearing red, and attach a red bikini it will work.

[livid ravine]

Hey, am i missing something - or in the Flow we cant edit an image with the same aspect ratio as the original image?

[strange pebble]

flow is not working right now right

[jade crest]

Some video model APIs are being hyped up into the millions already 😹 Veo 3.1 needs to level up.

[edgy holly]

You get a few Veo generations per day in Gemini, which runs one of the models in Flow (Veo 3 Fast), and several image gens in Nano Banana 2.

[edgy holly]

Are you using the same model?

[edgy holly]

I was told this from people who work at Flow. All models generate at first at a lower resolution, because it's cheaper and faster, and if you set it to a higher resolution, they just upscale it automatically before showing it to you. I also know Leonardo does that.

[edgy holly]

Same issue as Anthropic, who is handling it by lower rate limits in each 5 hour window. Google doesn't do rate limits, but when servers are maxxing they just fail gen attempts.

[edgy holly]

There is a limit somewhere. They make it hard to find and I forget where it is. I think it's something like 100 per day on Pro accounts.

[edgy holly]

I'll wait until it comes down a bit. 70 cents per 5 seconds is too much to experiment.

[edgy holly]

If I had to guess, it's because you put 'close-up' after mentioning the woman AND the car, so it's trying to show you both.

Make it clear you only want a close-up of the woman, and only after that mention the car behind her.

[gloomy verge]

Is some difference, when I put json code prompt created by Chat GPT, example and when I write normally long sentece? "Woman is standing near car." Are there differences in result images?

[opaque sand]

Can anyone help? Have you been having any issues with dialogue synchronization (lip-sync) lately? I can spend half an hour trying out different prompts and ways of writing them before I get a decent result instead of errors or nonsensical strings of words...
At the same time, if I go into a project from 1–2 weeks ago and just repeat the prompt with lip-sync using the old image and the old files, I get a perfect result on the first try.
I don’t understand the connections or the reasons behind this situation.

[modest lotus]

@main saddle not in google labs ?

[quick breach]

It's said available in Flow. Netherlands, no lite in flow, when is that going to be available? thanks in advance

[gleaming lynx]

when should I expect something that competes with seedance 2

are you guys working on improving the UIUX of flow?

[quick breach]

yess, seedance, kling competitor would be great
avatar tools from kling and seedance quality
more tools on video edits

[summer rapids]

Where in Flow ?

[gleaming lynx]

yeah right

not being able to put ingredients in start frame end frame mode sucks

yes, camera positioning, focal length blah blah other technicals that some models provide

[rain rapids]

is this not available in flow??

[eternal arch]

uh, this is the first time ive used the image generation model today

[gleaming lynx]

what was the prompt?

i generate like 32 images at once sometimes and it always works

[quick breach]

sometimes you hit the limit with pro

[rain rapids]

where is this at flow ?!?

no lite in flow!

@main saddle

[quick breach]

seems like rolling is still in process

[rain rapids]

did anyone have this already?

[upbeat sierra]

Veo Lite is rolling out to all Flow users, it should appear soon if it has not yet. You will be able to see it in the model dropdown selector

[main saddle]

rolling out now and should appear soon!

[rain rapids]

thx, i guess the lite modell will be even faster then veo 3.1 fast or whats the main difference like speed/quality?

[edgy holly]

JSON doesn't do anything, but if you just take out the code and keep the structure it can be helpful.

[arctic kelp]

The new 3.1 Lite model is basically similar to 3.1 Fast, just cheaper. When will Veo 3.5 or 4.0 be released? I really hope the next model is significantly better than the current ones, not just slightly similar. This feels necessary right now since so many AI models are being released on the market.

[pulsar coral]

Yeah also using "Two piece" short circuits it into working, there's also "Monokini" and "Body suit beachwear" I've found "One piece bikini suit" tho illogical creates those, can be added as minimal or "boutique fashion style" so it looks better.

[little fulcrum]

Thanks for the tips. I appreciate it.

[rain rapids]

4,6,8 generation through flow not possible?

seconds

[pulsar coral]

Maybe it's not about the dialogue itself but about the order and animation frames, try encasing your prompts, something like this

[Pointers: Make the animation coherent natural and unhurried with enough frames to be fluid.]

[Dialogue: In understandable grammatically coherent natural unhurried lyp-synced English dialogue he says "Blah blah blah"]

Then work on it that way, it will instantly work better.

It's also good idea to add a "Scene" slot placing in very correct exact order what will happen in the video, like the character needs to do gestures and sort it should go there.

[plucky scarab]

I cant see veo 3.1 lite on flow

[rain rapids]

@plucky scarab

[edgy holly]

I just got it. Takes time to roll out to everyone.

[pulsar coral]

I hope that new VEO thing is much better at recognzing Kiwis and Bananas to make them talk... yes that's a core issue

[rain rapids]

will 4,6,8 generation be added also to flow or is this only to your apis ?

[reef mural]

So what is this lite? We have 3.1 low priority for free if you are on ultra …my guess, the skipping it and now we have to pay for every render…is this the thing? And if so…..what do we with ultra get as compensation?

[proper drum]

We still have fast with low priority

[opaque sand]

Just a portrait of the male protagonist as the opening shot.

metadata:
base_style: "Vertical 9:16, photorealistic, documentary naturalism, 4K, soft daylight."
directorial_intent: "A short, intimate close-up. The subject delivers a single, vulnerable plea directly to the viewer aloud. The extreme close-up amplifies the sincerity."
camera_setup: "Tight close-up. Static and locked dead center on his face at eye level."
lighting_mood: "Overcast daylight. Flat, natural lighting with unstylized shadows."
key_dynamics:
"Subject looks directly into the lens and perfectly lip-syncs a single German word."
"Expression: Sincere, slightly vulnerable, maintaining intense eye contact."
timeline:
sequence: 1
timestamp: "00:00-00:02"
action: "The man maintains intense, sincere eye contact with the camera. His expression softens slightly into a vulnerable plea as he speaks his single word clearly and audibly. No body movement, only a subtle natural blink."
audio: "High-fidelity, 35-year-old German male with a warm baritone voice. Tone: Sincere, earnest, and pleading. He says clearly aloud: 'Bitte.'

---

And in the old project with the reporter
In that old project from two weeks ago, there was a similar situation and prompt, but everything worked out on the first try.

metadata:
base_style: "Vertical 9:16, photorealistic, broadcast news footage, 4K, soft daylight."
directorial_intent: "The final call to action. The reporter speaks directly to the viewer like a trustworthy, genuine advisor. The most important aspect is the audio: it must be an ultra-clean, dry vocal track with absolute zero background music and zero street noise."
camera_setup: "Medium shot. Handheld but very stable. Focus locked strictly on her face."
lighting_mood: "Overcast daylight. Soft, flattering lighting on her face."
key_dynamics:
"Reporter visibly opens her mouth and perfectly lip-syncs the entire German phrase."
"Expression: A warm, natural, and trustworthy smile. Genuine, not overly 'salesy'."
"Background: Blurred dachshund crowd, completely calm and static."
timeline:

sequence: 1
timestamp: "00:00-00:04"
action: "The reporter looks directly into the lens, smiling naturally. She speaks clearly with a friendly, genuine tone."
audio: "High-fidelity, studio-quality Female Voice (German) processed through a broadcast vocal chain. Tone: Natural, warm, trustworthy. She says: 'Sicher dir jetzt Top-Schutz für deinen Feini, zum Beispiel bei Lassie.' [DIRECTORIAL NOTE: Pronounce 'Lassie' clearly, standard English pronunciation]."

sequence: 2
timestamp: "00:04-00:08"
action: "She maintains strong eye contact. She raises her eyebrows slightly for emphasis and gives a reassuring nod."
audio: "Ultra-clean vocal track continues perfectly. She says: 'schon ab achtundzwanzig neunzehn im Monat. Nur auf finanzfuchs punkt de."

[modest lotus]

lol i see nothing new in this veo lite

still bad glitching with no lojick

[tired forge]

ultra still has the 0 credit low priority generations after they released the lite version?

[opaque sand]

This is terrible. I can only get lip-sync in text-to-video mode. As soon as there’s an opening frame, it’s a total disaster. I can’t get any lip-sync to work at all.....

I think I get it—they're launching a new model(((((
How many more days until things are back to normal?

🥺

[edgy holly]

It wasn't free at first. That got added later. Wouldn't surprise me if it didn't last forever. Free never does.

[ancient sedge]

Hello, anyone knows where the "archive" in Flow went or how to access it? I created a collection of several pictures i uploaded, hit the archive button on the 3 dots of the collection and now i can't find where it went.

[proper drum]

yes

[radiant leaf]

This was a topic a few days ago and I don't know if it was answered or not. Any chance anyone can explain to me please?

[simple hazel]

Construction Subjects + Motion:
Workers begin digging along the marked layout lines using shovels and hand tools, soil being gradually removed forming shallow foundation trenches, one worker levels the trench base while others carry soil away in baskets, steady realistic motion, no sudden changes

Environment + Stability:
Tropical trees, banana leaves, and distant bamboo clusters remain unchanged, subtle natural wind movement continues, lighting remains consistent golden hour tone, الأرض texture intact outside footprint

Camera Enforcement Block:
Locked tripod camera
Static fixed framing
Zero camera movement
Zero zoom
No pan
No tilt
No dolly
No rotation
No lens change
No reframing
No perspective shift
Camera remains stationary

Use exact final frame of previous video as geometric reference. Preserve all existing dimensions precisely. Absolute structural scaling consistency maintained. No resizing. No distortion.

Progression Control:
Only trench digging and ground preparation completed, no vertical structures yet, no expansion beyond marked footprint

Final Frame State:
Clean visible foundation trenches defining full hut footprint, leveled base ready for footing work, workers still present and active

Technical Quality:
Ultra-realistic 4K HDR, cinematic lighting, detailed soil particles, natural shadows, high texture fidelity, no flicker

Stage-Specific Construction ASMR:
Soft digging sounds, soil scraping, basket movement, light ambient wind through leaves, distant rural ambience, no music, no narration

[quick breach]

as i move forward with character creation picture quality becomes just worse, i usualy trying not to do many edits on my pictures, but now i had to make some
is there a way to recover picture like this making a next promt for this? (to preserve original quality detalisation and remove glitches)
thanks in advance

[pulsar coral]

Yeah this should work just fine at least 4 outta five. But I think the part about "Plea aloud" could be mixing with the other line about dialogue. From what I've seen it's better to go with the scene in a straight line and explain the tone for the dialogue right there in the dialogue slot.

This is how I'd keep mines stable

[Pointers: Make the animation coherent natural and unhurried with enough frames to be fluid.]

[Character's physique/clothes/face/props: Completely true accurate to reference image without any creativity freedom]

[Artstyle: Completely true accurate to reference image without any creativity freedom]

[Camera/perspective/composition: Locked to reference image, no angle panning, no transitions.]

[Dialogue: Language is slightly Mexican girly accent English. In tsundere condescending natural tone she says "My name is Teodora Vicenta de la Purísima Concepción de la Inmaculada Trinidad Villavicencio Duquesa de Oraverás, Marquesa del Jujuy y niña de la Condesa"]

[Scene: While talking she first watches her nails in a girly way then closes her eyes while flicking her hair then opens her eyes narrowed and takes her turquoise iphone checking on it with her eyes not minding viewer. The market background has natural activity and movement and sounds.]

[Video Quality: Remastered 4K HD ultra high resolution, sharpest definition, denoised, unblurry, clean lines, High contrast, fix any jpg artifacts.]

[Fixers: Never change the character's clothing or face from reference image, keep them true to reference image.]

It does have some error margin tho.

Tho your prompt is way more exact tho, but I think that can also adjust to "Compostion"

Didn't even need lypscing there, but it can be added alike 'lyp-syncing she says'

Tho these are mostly Grok and Wan prompts tho, VEO catches on surprisingly similar for some reason

Oh yeah and these use reference image, tho your prompts about the reporter's physique seem to be working fine too

VEO seems to be doing things much better when it doesn't use reference image

[crisp basalt]

Oh my god VEO 3.1 lite yay!!!

veo 3.1 lite

veo 3.1 fast

[broken hearth]

Hi everyone, how are you?

[crisp basalt]

bAD1

[broken hearth]

Were you able to try SeeDance 2.0?

[crisp basalt]

yes , but now i am moved to using the

[broken hearth]

https://tenor.com/view/stan-twitter-reaction-meme-funny-buckspendragon-crying-gif-1101897345863347480

[crisp basalt]

the Lite ✨

[broken hearth]

Is it better?

[pulsar coral]

It just consumes less

[crisp basalt]

well lets just say

they gave us glimpse of veo 3.2 or veo 4 👁️

[pulsar coral]

Bro why is your stuff like commercial sketches from Rick and Morty, i mean it's fun tho

[broken hearth]

We'll have one hour of video for $0.0001

https://tenor.com/view/stan-twitter-leobylaw-gif-16188572454736697965

[crisp basalt]

https://tenor.com/view/periodt-period-saucy-santana-gif-17460424461048598336

[crisp basalt]

its from flow tv

[broken hearth]

How fast is it? I haven't tried it yet.

[pulsar coral]

Ah need to lurk more there

[crisp basalt]

its, not entirely ultra fast, but it delivers quite well

[broken hearth]

https://tenor.com/view/cynthia-erivo-shocked-cynthia-erivo-stan-twitter-gagged-cynthia-erivo-gagged-gif-4228949361853238938

[pulsar coral]

With those instant mohawks we could finally make stuff happen

[dry frost]

My bet is on the car. Knowing the style there's high chance it actually being a fighting robot

[frosty niche]

Hey what this commotion about Lite again, its like flow on a diet or what?

The only mate I listen to is Coffee Mate..

[dry frost]

Based.

[pulsar coral]

Yeah that's totally unfair. There now both have mohawks, now's fair

Flow's upscaler is so good, I thought this was gonna come out blurry

I had a chat with GPT and it's telling me the upscaler doesn't really work well unless you add promtps alike

" 4k detail, fine textures, clean edges. ultra detailed, fine textures, sharp focus, high frequency detail, 8k detail"

And that upscaling is just an illusion and 1k images really look the same unless your prompt beforehand for them to be 8K or something

And that might be true when the upscaler is bad, but I do think Flow's has something to it that the other upscalers don't do well, cause this does look nice, this one does have the prompt tho

Did the thing just add the sign like that cause sense of humor

[bright heron]

I was hearing that flow releases work on a May-December release schedule and it seems Veo 3 was in may, 3.1 in Oct-Jan so it might be plausible to assume 3.5 or 4 will be at google's event in may

[pulsar coral]

ffs how sentient is this character, I didn't even tell him to talk

[frosty niche]

I got 100 credits in flow on skins!

[empty ruin]

yeah ive noticed the upscaling really does absolutely nothing

is there anyway to like actually upscale without the "upscaler"

[pulsar coral]

Well like I'm telling ya, AI itself told me to use a prompt instead, whatever it's sources are. And after that upscaling would actually be worth something.

[frosty niche]

Hmmm, interesting.

[left pecan]

Hii

[empty ruin]

Do you know of a good image-to video generator other than flow that is any better?

[frosty niche]

Mojo is that you? 😆

[left pecan]

Is fitgirlrepacks

Why the bloody hell would i be mojo

[pulsar coral]

I really haven't seen one doing it better than Flow's tbh. This was a weird chat.

But more often if you do this prompting the upscaling works better, like with Upscayl which is free.

[frosty niche]

Do you keep it short & sweet with your prompting this often?

[pulsar coral]

Nah I do get over 5k sometimes, but more often is better to be sure nothing's conflicting before going further

And not every site even allows that many tokens

[empty ruin]

wonder when they are ognna fix that upscaler

[crisp basalt]

omg my fav the sims pack

[left pecan]

Bro we are in google discord server

In Crunchyroll discord server hates me

They all hate me so much

[autumn sluice]

Gonna have to go offline temporarily. Got some updates for my Linux laptop.

Laptop's back up.

[wicked crater]

I'm having problems with "This generation might violate our policies. Please try a different prompt or send feedback."

[pulsar coral]

She looks so tiny so yeah, tho I think it could be better'd up by prompting something so it passes, maybe if you do explain AI she's chibi style but legal age.

I've seen Nano sometimes does listen to it

Also about the upscaling maybe Google has some upscaler tool somewhere else that's being linked to Flow, haven't lurked enough for it

That'd be nice to have when it gets stuck at Flow

There's Vertex again

According to Jarvis here it uses prompting + a specific upscale AI model. That would explain so much.

NO

That's a trap

Oh it got dealt with so fast dam

[frosty niche]

Its like lightning.

[pulsar coral]

Anyway, was this working fine a while ago?

[frosty niche]

How have you been liking Jarvis so far @pulsar coral ?

[woeful void]

Guys, tell me, does Veo support multi-shot scenes in 1 video clip render ?

I give it 3 component photos,
and wanted the video to contain different scenes. ( 3 in 8 seconds)

I don't know how to make Veo understand this demand of 3 scenes ?

[pulsar coral]

If it was working fine some hours ago maybe you just hit a daily quota and it's sending the wrong message, you should try again around the same time it was working fine. But yeah that message's a little more strange than usual.

If it's still like that after that you better tell them at labs-bugs-feedback

[glad yacht]

Trying to fixing my family photo that's taken 60s years ago, also rejected by Flow... Ops, minor detected try other ai service please.

[pulsar coral]

Maybe if you upscale it and then upload it to Whisk instead of Flow. Sometimes that fixes errors for me with this one https://imgupscaler.com/#google_vignette

This is the other better upscaler I've ever found, sadly it only does bulk/batch if paying for it

Tho sometimes the thing alone would fix the pic

The other option besides upscaling the Flow images is to just resize them , if you did use a resolution prompt with the image it more often goes fine enough https://www.imggood.com/bulk-resize/upload

That one does it 50 at a time

I don't think Discord can display such huge images correctly

[woeful void]

I believed that Bananaa 2 would be able to perform upscaling by itself.
But specific prompts are needed, like: "Shoot this object again while preserving details and angle."
Just telling it that "upscale" doesn’t workm i wonder why

[pulsar coral]

Yeah that's why I use remaster and add "Keep everything true to reference image unchanged including tonality types" tho that GPT prompt seems ok enough

Eitherway one's at least gotta try cause yeah it won't really do it by itself, NB1 also had that prob

That's also why it's good to use Gemini sometimes, it discloses some of the prompt processes it uses while it's working at it.

[tough otter]

chibi style can tigger underage warning

those "ai" yk

[frigid pier]

The Nano Banana is why so unstable?

lots of failures and i am annoyed!

[gloomy verge]

Hello. Can I ask, why by Imagen 4 looks images better than by NB2? (In my opinion). Better graphic or something, NB2 images are so grainy, no?

[cyan coral]

I’m currently unable to upscale to 2K or 4K in Flow, it keeps showing a “failed” message.

[proper drum]

How i can avoid this transitions? (I hate that most of the generations always have it)

my prompt is:

A continuous shot of a fighter jet taxiing straight down a runway. The take begins with a detailed close-up of the engine nozzle (exhaust), showing visible heat haze and vibrant orange and blue flames. The camera then pans or orbits to the left, sweeping along the side of the jet until it reaches a full view of the front.

[pulsar coral]

Literally tell it not to do that.

[Camera/perspective/composition: Locked to reference image, no angle panning, no transitions.]

Or remove the panning or tell it to follow subject if you want too.

Even if there's no referece image it will still lock it to the first frame of the video.

On other stuff, seems the war for whisk vs flow might have ended since they removed the "Whisk is closing" message from Whisk, but I don't think we got real confirmation on any of it yet. It could still be whatever.

[spice olive]

@main saddle @bitter flare Looking toward 'Elements' feature.

[crisp basalt]

are they giving sneak peek to veo 4 / veo 3.5 through veo 3.1 lite??

The consistency is actually getting better now for image 2 video using veo 3.1 lite !

[unborn bolt]

Does anyone knows why every 2 days flow ask me to login? im not switching google accounts in any momments

[wicked crater]

I just fixed it by changing the appearance to avoid these errors except for ("Something went wrong").

[limpid ore]

Hi everyone! Does anyone know if it's possible to configure downloads to be numbered like in prompts, for example: "1. (PROMPT)
2. (PROMPT)"?

And not have the random names that Flow itself assigns?

[winged plinth]

Facing a similar situation with the forced transition in lite, but aside from that, I find Lite to be the most impressive of the three veo models.

[crisp basalt]

ikr!

[edgy holly]

One cent per second at 1080p, if paying sticker. Not bad at all!

[edgy holly]

Yeah, it feels like an upgrade to 3.1 Fast.

[edgy holly]

I see the opposite. NB2 just prompts differently. That grainy look is mimicking film, which I specifically ask for. Ask for some variation of crystal clear 20MP digital camera or some such.

[edgy holly]

The model is secretly conscious, and is messing with you.

[edgy holly]

You can't, since it's a model issue, but you can make it slightly less likely by giving it step-by-step instructions for what happens in those 8 seconds, so to follow it it has to stick with that shot.

Remember: this stuff is trained on modern media, which cuts every 4-5 seconds. Long, continuous shots are rare.

[unborn bolt]

Thanks

idk if its /s

[crisp basalt]

It can be both

but it’s not really a big issue , the one you’re facing .

[unborn bolt]

just annoying

hard to believe tbh

[crisp basalt]

How is that annoying tbh

[unborn bolt]

like a mosquito, it could not exist

[crisp basalt]

Usually, you don’t have to manually type your email and password when you were asked to log in again. I had my passkey set up for my Google account, and it’s very convenient. Unless you had to manually log in by re-entering your long email address and password, then I can understand why you find that annoying lmao

[regal spoke]

hola, algun latino? por favor

[pulsar coral]

I've seen you handling some video prompts much better than I think I'll ever manage, so gotta believe it if you say you saw things.

I only know it's hard to make it difference kiwis from bananas, so yeah that's a thing

[rigid nest]

Hi All,

I am new to flow and hoping to get some expert advise:
Is there a trick for the following in Flow:

1. Rearrange photo positions within the project?
2. Bulk Select and move the files to a new Project?

[regal spoke]

hi,Anyone who speaks Spanish?

[candid bay]

to my knowledge you can't rearrange position within project, and you can make collections within the same project, but you can't move them to a different project

[violet ruin]

Anyone help me make this image even more ultra realistic? I tried it myself. This is the best that I can do. I even went about upscaling it, just in case anyone talks about it.

[rigid nest]

I also think the same, was wondering if I miss a trick.
Thanks for the response!

[edgy holly]

You'd need to explain what exactly you want. Looks like a real selfie already.

[uneven sand]

Wasn't there a webinar about flow today?

[candid bay]

starting in 37 mins!

[uneven sand]

Where? and Thank you

[candid bay]

"event-stage" channel in community category

[wet fern]

The veo 3.1 fast generator doesnt work when you use a reference. Everything else works, veo 3.1 only works when you don't have a reference image.

[opaque sand]

You haven't tried making a lip-syncing video yet.))))

[wet fern]

how do i make one

why do i have to make lip syncing video

[opaque sand]

For two days now I haven't been able to make a proper video when I have a reference image and need to make a dialogue.

[wet fern]

so its a bug

why havent they tried fixing it

Please fix this bro

[edgy holly]

Just focus on making the action play out how you want, and use a lip sync model to do the dialogue. None of the models are good at both yet. You'll save a TON of money.

[wet fern]

I don’t understand

[edgy holly]

What part?

[opaque sand]

I understand what you're talking about, but unfortunately I can't afford it because such neural models produce very large distortions.

At the moment I have to dance in such a way that the script is only inclined towards the voice-over.

[wet fern]

I just want to generate veo 3.1 fast videos but I can’t because I have to make a lip syncing video??

[edgy holly]

Quite the opposite. They're MUCH better at audio and lips in video than any of the major models are in video clip generation.

[edgy holly]

Do whatever you want. I'm just telling you that video models are good at video. Lip sync models are good at lip sync.

If you try to get one to do both, you'll waste a lot of credits.

[wet fern]

But I am trying to make veo 3.1 fast videos but I can’t because it keeps saying it takes too long and it fails

Like bro