flow-general・🌊

[deft summit]

That seems to have about a 25% success rate for me

Maybe 15% lol

[modest lotus]

or wait it works

try it guys

[fluid abyss]

ive been doing this for the past few days, This is a nanobanana issue, other platform like higgsfield and freepik are also suffering due to this

[digital spire]

Yes is still really bad but at least is something i guess

[deft summit]

I thought my Pi Hole was blocking it, but I guess something else is going on

[fluid abyss]

it has gotten worse today. yesterday and day before my plan was just to spam generate and wait for it to show failed and reload and i used to get atleast 1 image lmao

[worldly willow]

still not working.. any admins or mods... please update

[modest lotus]

@digital spire its working bro thank you so much you saved my day

[crystal talon]

When veo5 ?

Lets skip 4

[tawny sandal]

your flow is work or not work ?

[stuck parrot]

Flow not working today?

[tawny sandal]

Yes, it doesn't work. Has any information been shared? I couldn't find any.

[crystal talon]

I canceled for now. Too many policies. Everything is forbidden.

[tawny sandal]

Rules are necessary, but right now we can't even generate a picture of a cow; the flow is broken and not working.

[worn dust]

how do we post on the flow-bugs-feedback channel? I don't seems to have permission

[stuck parrot]

this worked thanks

[slate bloom]

why flow are constantly telling me something was wrong,and the video and picturea cannot generate

[manic berry]

Is there something wrong with Flow's service today?

[worn dust]

we all have the same issue for the past 3 days. hopefully the team is working on it. thousands if not milions of users are stuck using flow

[deft summit]

Policies are definitely needed. We don’t want Google to turn into Grok. I had to cancel because I was asking for vanilla images and it was making highly inappropriate content

[fathom pike]

Bro i got shi to do and this has been down for HOURS fix up if ur charging 200$ per month for a service that is down most of the time

[fluid abyss]

does anybody know if whisk uses nano banana pro or the old nano banana. its generating images for me

[deft summit]

Old nano

[slate bloom]

the whole day cannot woek,i can not do anything in flow,and I subscript the ultra.

[fluid abyss]

people over at higgsfield are also suffering as i can see on their discord

[crystal talon]

Flow is too strict, we dont expect XXX content but veo and nano became ridiculous at this point.

[honest shore]

give more freedom for image generation? There are too many rules right now. It doesn’t really make sense to pay if there are so many restrictions😠

[fluid abyss]

in a few months open source models will reach nano banana pro. heck z image is already somewhat there. you cannot really control what people choose to do

[crystal talon]

Agree, it feels like it is run by extremist muslims. I cant even generate a woman with a tshirt and a belly piercing.

[fluid abyss]

seedream is so much more open and cheaper but ive been spoilt by the unlimited nano banana in the ultra plan

[deft summit]

But a decent company doesn’t want people generating illegal content with their own products (unless your xAI). That exposes your company to risk

[crystal talon]

Nobody here wants to create illegal content.

[fluid abyss]

choosing not to train the models on any image that has any form of nudity or extreme gore is an option they could go ahead with

[crystal talon]

It is just too strict. Not possible to make a shortfilm.

[gloomy verge]

When I create image in Flow, can I place it to Flow-gallery? Can be there images too? Or only videos?

[deft summit]

I’m not saying anyone in the group wants that. Just that the more open a model is the more likely it is to be able to generate illegal content. Whether intentional or accidental by a user

[fluid abyss]

i think flow gallery only has videos

the entire grok issue is what prompted this to happen with google for sure,

[frosty comet]

my take on it is it should be even stricter on images containing photorealistic people, while loosening filters alot more on digital art involving adult characters

[vestal salmon]

olá

como faço um video realista?

[tulip helm]

Why image generation erro ?

Any one have same experience ?

[fluid abyss]

everyone brother everyone

[crystal talon]

Not good enough. The moderation detector is bad. Often it thinks a grown woman is a minor. Ridiculous

[vestal salmon]

não falo em ingles

Alguem ai fala portugues?

[frigid sapphire]

Opa

falo sim

[serene cradle]

https://tenor.com/view/touching-grass-touch-gif-21219969

[frigid sapphire]

I'm trying to generate images in Google Flow, but it's not working and keeps giving me the message: "Unable to generate the image. Please try again later."

Anyone else having the same problem?

[dusty cobalt]

Yes I am having the same problem and a lot of users too.

[tawny sandal]

Hey guys, maybe we should find the relevant people and tag them? What should we do to spread the word?

[dusty cobalt]

I think they already know but don't want to respond.

[glad isle]

Hello everyone, can we create a 60-second endless runner game background video using Flow AI, without a player or coins—only trains and obstacles?

[dusty cobalt]

Maybe it will be fixed when we sleep and wake up the next day.

[lime mural]

jesus, its like every day now

[worldly willow]

still not working.. any admins or mods... please update

[lime mural]

make sure you report on the labs

maybe that helps lol

[dense lotus]

Same hera, i hope they are preparing some suprise drop kek

[dull rapids]

Yeah they're having issues with their infrastructure per higgsfield, so I assume they're aware and working on a fix.

[dim solar]

me too

[fiery panther]

Still down?

[little stone]

yes

[dusky canyon]

Do you have unlimited nano banana pro generations inside flow studio with Google AI Plus subscription?

[turbid lintel]

yes

[little stone]

gemini can generate pics by NBP, but can't download the full size pics

It seems that there is a problem with the server program, which has resulted in the limitation of certain functions. While images can be generated, the algorithms for image enlargement are not available.

[oblique wind]

Is Flow down? I can't generate a single image.

[little stone]

use Gemini to generate now,

What is certain at this point is that the NBP can be used. However, generating four images at once using the FLOW tool would obviously overload the system, which is why its use has been restricted without any prior notice.

[strange forum]

This like day 3 Google flow been down

[little stone]

Yesterday, the system kept displaying alerts indicating that it was under high load, and that tasks might fail. Users were advised to try again several times.

[dim solar]

Friends, does anyone have any prompts to share that can restore ancient paintings into photos with a good realistic photography feel? The ones I always generate are not还原 enough and have inconsistent elements.

[little stone]

upload the pic to gemini or chatgpt, and ask him to create a prompt to keep all elements of the composition and atmosphere adhere to the physical laws of the natural world. The style has shifted from that of an oil painting to that of a photographically realistic landscape. then chatgpt will give you a prompt describe the painting, then ,use your original pic as reference to NBP and use the prompt to generate a new pic based on the reference pic

[small bane]

i cant download 1080p vid

any help?

[dim solar]

I've tried, a they always misidentify and describe many details in the images.

For example, this scene

[dim solar]

That's great. The problem I encountered is the incorrect restoration of characters' clothing, hairstyles, etc.

[little stone]

use chatgpt or gemini to do that

[dim solar]

Thank you, I'm still trying.

[little stone]

NBP on Flow seems recoverd but upscaling to 4k is still unstable

nano banana pro

[small bane]

what about 1080p?

[little stone]

using Google AI Ultra for Business

[little stone]

not stable

[meager edge]

Will it be stable with API @little stone

THE NBP

[little stone]

the same

[small bane]

eta ?

[little stone]

but API is quite different from Flow

[turbid umbra]

not sure if placebo but recently because no more credits lol, I started using the lower priority veo 3.1 fast and I'm getting amazing results haha.

[little stone]

After my test, in fact, this mode is similar to the 100-credit point model.

[analog lava]

Hello @little stone @dim solar , we wanted to remind you of our server guidelines, stated in #start-here・📍 : “Currently, our team can support engagement in English; the discussion in this server will largely remain in English for now.”

[dim solar]

ok

[tribal mirage]

what happened to flow ? i cant use nano banana. Cant generate image

[gloomy verge]

Is possible in Flow image generator set custom seed same as in Whisk?

[odd hinge]

is anyone having problems with generating videos from images in VEO3.1? When I generate video from a frame it doesnt work. The scene extension tool is also repeatedly denying my requests.

[analog radish]

Nothing is working today

[odd hinge]

I got onto the Ultra plan to start using Flow yesterday and it was working quite well. How often do situations like this happen?

[spiral rose]

same issue facing ? i dont no what time is good for use this

[odd hinge]

When I use the scene extension feature it tells me that I am going against their content policy, but my prompts have no nudity/racism/children, anything like that. It's a simple shot of a woman in pajamas talking to a camera.

When I generate video from frames, it gets up to 99% generated, then fails saying there was a problem with the audio.

And when I try to switch to Veo 2, it fully doesn't work at all.

I have tested this with Fast mode, low priority mode, all the different models

the prompt I entered was simple dialogue with no crude language of any type. and no visual changes. Google says "this might violate our content policy"

[tiny canyon]

@main saddle can you explain what happened to google flow

[spiral rose]

same issue again & again we change prompt but still not working some time error some to loading then start ZERO i think now move another tool

[viral folio]

Guys I think they are launching the VEO 3.2 and that might be the reason for these errors

[turbid umbra]

could be, there were interrupts last time they updated 3.1 fast

[exotic shell]

flow and website extremely slow?

[edgy holly]

Could be. Error rates also spike during peak use times. If you set up social media monitoring, you can see complaints rise and fall.

[viral folio]

Exactly!

[fiery panther]

Still down??

[uneven matrix]

their violate policy restrictions need to be let up!! it is insanity you can't create almost anything anymore with may violate policy. or god forbid now 3rd party restrictions!! what are we even paying for anymore.. i think everyone should honestly cancel their subscription maybe that will teach them a lesson

[edgy holly]

I've been using Veo for over a year, and almost never run into them. They don't have the legal option with the copyrighted characters, but it seems very hit or miss.

Some people seem to run into them all the time, and others almost never. I don't know what the pattern is, outside of anything related to kids, false positives with famous people and copyrighted characters.

[uneven matrix]

you never created venom or women with big busty breasts then or any other story telling stuff like i create

[edgy holly]

Yeah, I'm making original content, a drama TV series, and it's not "spicy", as the kids say.

[thick shadow]

I'm not trying to make spicy content, but I'm trying to make a music video for my surf punk song, and I can't even have a female character in a two piece bikini at the beach getting ready to surf...

[edgy holly]

Which error are you getting?

[small bane]

Is the 1080p rendering back up?

[thick shadow]

Basically "this can't be Generated because it might break policy"

[small bane]

Ask ChatGPT how to bypass with a similar prompt it will help

[thick shadow]

Tried that same result.

I had to create the image in grok,

And then bring it over as an ingredient

Still after wouldn't even animate

[unique abyss]

I, rarely and inconsistently, get them for no apparent reason.

I get them more often for potentially understandable, but still less than ideal reasons, like when showing a pregnant woman seems to, inconsistently, trigger "don't show children" safeguards.

It's not the end of the world, but it would save time and effort if things the policy guardrails could be more focused. Of course, it's much, much easier for me to type this than it is for Google to implement it.

[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

[thick shadow]

I wanted to show an age progression of a character going from child orphan to warrior and that didn't get generated either. I'm still sniffing out work arounds

[unique abyss]

At the end of the day, part of the challenge, and part of the fun, is that the devs, and we, are working on the cutting edge.

[fast tiger]

For me, it only works when I ask gpt or gemini to make the prompt for the video to be written in sanitized language (and completely safe for work), sometimes it's one or two words, sh** is weird

[thick shadow]

In the meantime I just need to figure out how to make action sequence that doesn't look too wobbly. But I also wanna test that gen4.5 engine.

[fast tiger]

Did u try any realism and stability settings?

[thick shadow]

I'm using a custom expander so everything is grounded in realism. but the results are hilariously wacky, but unfortunately unusable for what I need.

[fast tiger]

I see

[autumn vapor]

amm im trying to download a video from scenebuilder and it keeps giving me an error what to do ?

[unreal viper]

please use #1238621511401734254 to submit any bugs you find!

[haughty phoenix]

Hi guys, is Flow also giving an error when exporting?

[fast forge]

Same here, having problems with exporting the video in screenbuilder

[fickle tulip]

Yes. I am having too. Idk how to fix that

anyone know?

[desert linden]

Realistic 4K cinematic video, Indian village small old house exterior, soft morning light, emotional warm mood, slow camera pan from mother up to towering green giant Marvel Hulk. Elderly Indian mother in simple saree, tired sad face, sitting says in Hindi: "बेटा, ये पुराना घर अब टूटने लगा है... बारिश में छत टपकती है। काश नया घर हो जाता..." Hulk (green muscular giant, ripped purple pants, gentle caring eyes) replies in Hindi: "माँ, टेंशन मत लो। बेटा है ना... जो भी करना पड़ेगा, नया घर दिलाकर रहूँगा।" Gentle emotional background music, Hulk's voice deep but loving.

[spiral pollen]

flow is broken?

[clever sapphire]

😴 ...Something wrong with flow?

[muted blaze]

A high-quality 3D medical + Pixar-style vertical (9:16) animation set inside the human kidneys and urinary system, shown as a realistic medical environment with translucent kidney structures and flowing fluid. An anthropomorphic melon (fresh, bright green, juicy texture) with human eyes, expressive lips, hands, and legs stands confidently inside the kidney. The melon holds a medical cleaning brush/wiper tool and physically scrubs dark toxin particles and waste buildup from the kidney walls; the dirt breaks apart and completely dissolves/disappears into clean fluid as it’s removed. The melon looks friendly and energetic for kids and adults, with cinematic lighting and clean visuals.
The melon clearly uses its hands to clean, smiling as the kidney tissue becomes visibly healthier and glowing.
The melon speaks with accurate Hindi lip-sync, saying one short first-person line:
“Main melon hoon, main aapka kidney saaf karta hoon.”
Camera remains completely static (no pan, no zoom, no shake).
No on-screen text, no watermark, no logos.

[rough kernel]

Hi guys, is there any official discord server for Antigravity specifically? I don't see any section for it in this server

[edgy holly]

hqve

[edgy holly]

Have you checked the Google Devs server?

[rough kernel]

no, is that where I can find the antigravity chat?

[edgy holly]

Not sure, but that'd be my best guess.

[rough olive]

#flow-general・🌊 a mango

[glad isle]

Hello everyone, can we create a 60-second endless runner game background video using Flow AI, without a player or coins—only trains and obstacles?

[surreal phoenix]

hi guys, if you're looking for a cheaper alternative to google flow, for generating videos with Veo 3.1 fast, you can check this out: https://veogen.studio

It is still is development, but I'll launch it soon

[viral folio]

Is this your project?

[little stone]

can you guys download 4k images in Flow?

[edgy holly]

working for me, yeah

[fluid abyss]

is the nbp issue fixed?

[surreal phoenix]

yupp, I'm currently building it

[viral folio]

Crazy dude!!

[compact berry]

Cant even generate a mockup for a landing page with flow right now keeps saying it may violate policies. SMH

[versed phoenix]

stuck here is it down ? (antigravity)

[violet cave]

how to fix this ?

[zinc fog]

same

[versed phoenix]

um i tried using a vpn and changed logged in google acc it worked

[lunar moth]

is your country banned or what

[versed phoenix]

No , I used vpn of my country only

[ripe shale]

What the...

[chrome yarrow]

Hey guys if you're encountering failed generation over and over,try to change the video setting from FAST to BETA or from BETA to FAST. and also consider logging in to incognito. really works well for me

[worldly willow]

NBP not working again

[narrow sigil]

do u guys goes viral by using vo3

tiktok or youtube

my video is not getting views

[turbid umbra]

i got like almost 3 mil view on tiktok, veo3 is amazing you just gotta know what you generate

[narrow sigil]

well i made videos but not go viral. I create about anime videos and cat videos.

[digital yoke]

Swedish text mess up alot

[haughty phoenix]

flow downs guys?

[rocky forum]

is flow down?

[hard fossil]

Flow dont't exporting the files in scenebuilder

its happening with you too?

[dusty cobalt]

Flow is coming Up and Going Down again and again. PRO accounts are working fine. Ultra accounts affected according to my observation.

[balmy birch]

That's what I see too.

[versed garnet]

is it down again? I'm having "Something went wrong", ultra account

[thick oriole]

Hey all, yes, Flow (and other products) is having capacity issues again. We're working on as many mitigations as we can, and we should expect that free traffic will be impacted before/more-than paid traffic, but even paid traffic is experiencing some problems right now. We expect some, but not all, generations will succeed.

Separately, scenebuilder export has been somewhat unreliable recently. Sorry also for that. We've shipped a bunch of improvements over last week and this week, which should have helped, but we're working to make that fully-stable.

[restive rain]

sure buddy

[slate bloom]

we pay for flow，I am ultra account.I wish the system can work fluently.

[thick oriole]

Wait Discord is meant to be the friendly and supportive community. Reddit is where everyone sends me mean messages 😆 (but I get it, sucks when our product isn't working the way you expect)

[analog lava]

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

[shadow warren]

So new month, new credits, same issues and annoyances. Instead of compensation or reverting back to 0 credit ultra gens or fixing stuff, we just keep taking money and adding other services

[serene cipher]

Google Flow Veo Audio Suggestion: Enable Google Veo Flow to export separate audio tracks i.e. dialogue, sound effects and music. It would be very helpful for professional creators

[tropic breach]

I can't generate anything in FLOW. Does anyone have the same problem?

[fickle drum]

rn, i generate pictures. frames for videos

[fiery panther]

Its working but weird

Giving third party false postives or stuff like that

[narrow sigil]

How much is Ultra subscription?

[haughty phoenix]

Hi, my credits were reset to zero, I still had 30k... is it a bug?

[chrome yarrow]

same here

[haughty phoenix]

its a bug

[chrome yarrow]

what do you mean a bug?

tried logging in and out, still zero

[chrome yarrow]

I think the bug didn't take my credits 😆

[polar scroll]

I went back to the homepage and received this message: - You don't have access to Flow.
Review Flow access requirements at the Help Center or contact Google Support.

[twin mortar]

Can I ask if Flow has issues chacheing elements from previous generations into new generations?

I’ve noticed this a few times - for example, I created a number of 4 generations of a couple of waitresses high-fiving. It wasn’t working for me so I removed all references to high-fives and re-run the generations. The result was a couple of no high-fives but also gens with high-fives. This type of thing has happened on other generations.

So does Flow have some sort of cache ability or issue?

[wintry pine]

I've noticed today that I'm getting a message: "Flow is currently experiencing high demand. Things might not be working as expected, please try again later." a lot.

I think Flow is just hitting its internal server limits and it's showing up in different ways.

[craggy crane]

are you using shared account ??

[lone turret]

Flow is currently experiencing high demand + 0 credits again! Time to say goodbye flow

[haughty phoenix]

no bro

[pastel laurel]

I have a question. I live in France and I don't understand why I don't have the camera control option. Could someone please help me?

[thick oriole]

I don't believe there's any location-specific reason you wouldn't have access to that. Can you check you're following the steps in "Change a clip's camera position or motion" in https://support.google.com/labs/answer/16353334?hl=en&ref_topic=16355775&sjid=6165977999877809606-NA ?

[thick oriole]

Capacity problems are much better right now btw, we'll take the banner down and try and keep you all well-informed.

[exotic crystal]

@thick oriole - why am I consistently getting error messages for audio generation for the last 2 days? Is this part of the Veo capacity issue? And if so, when will it be fixed and is this going to be an ongoing issue? I need to find a more reliable tool if this keeps happening as I have an ultra account.

[winged plinth]

Hi @exotic crystal Could you try again? Ian mentioned capacity problems are much better at the moment, so audio generation should be more stable.

[exotic crystal]

Thanks @winged plinth - still happening

[winged plinth]

Could you share the prompt so I can check what might be causing the issue? Feel free to DM me your inputs if you’d prefer to keep them private.

[exotic crystal]

I don't think I can DM you unless you're a friend. Here is one of two prompts that are failing:

In a cozy kitchen scene, show a friendly chrome robot standing stationary on its wheeled legs with one bright blue light and a glowing “IDLE” display on its chest.
The robot remains unchanged throughout the clip.

Papa Sloth — FIRST VISUAL BEAT (PRIORITY)

Papa Sloth enters from the left doorway, wearing the same green sweater and blue jeans for the entire duration of the video.
His jeans must stay fully visible and must not blend into fur, change shape, or disappear.
He is holding a rectangular cake pan with a frosted cake.

As Papa Sloth crosses into the room, his eyes immediately land on the robot.
The instant he sees the robot, his expression snaps from calm contentedness to a clearly negative reaction of shock and alarm.

His eyes widen sharply, his mouth opens slightly, and his posture stiffens in visible disbelief.
He continues walking forward while staring directly at the robot, clearly unsettled by what he’s seeing.

His gaze remains locked on the robot and does not shift away.
Papa Sloth’s shocked reaction must be established clearly before Mama Sloth draws attention to the robot.

Mama Sloth — SECONDARY PRESENTATION BEAT

Mama Sloth is standing beside the robot, wearing her knitted dress.
She lifts both arms slightly outward in a cheerful presenting gesture and says, excitedly:

[exotic crystal]

Second part:

Mama Sloth (dialogue): “Ta-da!”

Immediately after finishing “Ta-da,” Mama Sloth closes her mouth and remains silent for the rest of the clip.
Her mouth stays fully closed with no additional speech or lip movement.
Her hands must remain three-fingered sloth claws/talons for the entire shot.

She looks at Papa Sloth expectantly, clearly waiting for his reaction.

Environment & Style Constraints
Maintain warm, soft kitchen lighting with natural shadows
Cozy, inviting atmosphere
No camera movement, no cuts, no zooms
No changes to environment, props, or character design
Soft, detailed fur textures
Expressive but natural facial emotions
Robot remains inactive and unchanged

[winged plinth]

Please let me test it for you.

[exotic crystal]

There's a first and last frame too, which I can't send unfortunately

[winged plinth]

I was able to generate it using text to video, but please note that server capacity still seems to be stabilizing. It may take a bit more time, so please try again after a little while.

[restive rain]

bro i wasent mean

[restive rain]

how is that mean

[crisp basalt]

Move on already lol

[edgy holly]

Stem separation would be super useful.

[edgy holly]

$250/mo

[edgy holly]

Sloths FTW

[crisp basalt]

audio failed on this part on my end, im pretty sure its veo model safety filter refuses to generate the audio due to the dialogue of the character being too short (ta-da)

without the safety filter, the audio generates but it start makin the character speaks gibberish for no reason

lmao

[exotic crystal]

Thanks for looking into this! I'll change the prompt and see right now

[exotic crystal]

Thanks for checking it out, I appreciate it

[crisp basalt]

since veo still refuses to properly generate the audio for the tada part... i improvised the whole thing

[latent matrix]

flow is sending error on 1% generating

[exotic crystal]

Hmm I took out the tada completely and it's still failing audio

[midnight cedar]

yesterday, we made some updates to improve multi-turn image editing quality. let me know if anyone spotted any difference 🙂

[crisp basalt]

oops i forgot to incude my workflow , I used Ingredients to Video. I generated the character sheet based on @exotic crystal written character description using Nano Banana Pro in Gemini. Then, I input the generated character image reference sheet into Ingredients to Video and prompted.: A cinematic, wide-angle shot, utilizing an anamorphic lens with a shallow depth of field and soft oval bokeh, captures a cozy, rustic kitchen scene bathed in warm, soft lighting and natural shadows. Papa Sloth, a highly detailed CGI sloth characterized by thick, textured fur and wearing a green knitted sweater and blue denim jeans, slowly walks into the frame from the left. He is clutching a rectangular silver cake pan containing a frosted cake, and his face is frozen in a highly detailed, shocked expression, his wide eyes conveying disbelief as he stares across the frame at the Friendly Robot. The Friendly Robot, a highly polished chrome and white mechanized figure with articulated limbs and two black wheeled legs, stands motionless on the right side of the frame; its chest features a glowing blue display showing the word "IDLE." Beside the robot, Mama Sloth, a fluffy CGI sloth wearing a textured beige knitted dress, stands in gentle volumetric light, cheerfully gesturing toward the robot with her three-fingered claws and saying, "Ta-da! Say hello to your new automated sous chef!" The camera executes a slow, smooth tracking shot to the right, slightly closing the distance and enhancing the dramatic tension between Papa Sloth and the two static figures (Mama Sloth and the Friendly Robot) in the well-lit kitchen environment.

[edgy holly]

Multi-turn?

Not familiar with that term. Does that mean NB Pro holds onto context longer when you're in a string of edits?

[crisp basalt]

I think the problem isn't the word "Ta-da." It seems like Veo has trouble with audio when the characters only say a few words 😓 its been an on-going problem since many months ago, i thought they would fix it by now.

[winged plinth]

Refine with shorter follow up prompts [e.g., adjust lighting, camera, or small design changes] instead of rewriting a huge prompt each time.

[crisp basalt]

oh?? a wild jaime appears and gave us this 👀

[exotic crystal]

Oh I mean I took them saying anything at all out of the prompt

[crisp basalt]

wait is that what it is? the multi turn?

[crisp basalt]

ah-

[exotic crystal]

But just tried again and now it went through

[edgy holly]

It'd be funny if it just meant that characters can just twirl around like whirling dervishes now.

[rugged arrow]

Anyone know why I would constantly be getting export failiures in Flow? I'm a newish user and still learning but am on the Google Pro plan.

[midnight cedar]

by multi-turn image editing I meant >= 1 edits in this UI

my follow up question is, on a scale of 1-10, how important to your workflow is it for nano banana to keep the previous images in this edit UI context to be used in future edits? currently it's only looking at the context of the current image you're editing

[turbid lintel]

what is multi turn image editing?

[crisp basalt]

OHHH

i get it now , so basically ... we can edit the image again and again without having to ... go back to base image (?)

actually nvm idk either

[exotic crystal]

@midnight cedar - before you get more feedback on extra features, can you focus on solving the problems for the ones you already have ? I'm an ultra user - can you fix the reason why some of my prompts are failing audio? (No NSFW content)

[midnight cedar]

hey, other team members are working on that 🙂

[turbid lintel]

very important , the first image is the base

[crisp basalt]

yes yes very important

[exotic crystal]

Which team members? Are they in this discord?

[midnight cedar]

thank you all. this is very helpful!

[edgy holly]

That and UI issues are far and beyond the biggest problems at the moment. The models are great.

[midnight cedar]

we're working on it! stayed tuned

[crisp basalt]

#1319424187730755636 well theyre mostly active keeping track the bug and respond to users bug thread so maybe u can check it out

oh sorry its #1238621511401734254

thank u for the heads up georgia

[edgy holly]

~8 (I often have to backtrack)

BUT: Ideally I'd have input influence be a setting, like a bunch of other apps have been doing for years. Playground AI had a nice slider like 2+ years ago I think. Leonardo has a limited range (see image 1), and InstantX InstantStyle goes deep on the settings (image 2). Extra points if you split it into Style (art style transfer) and Content (depth map/layout).

Ideally also with a masking option, so you can apply edits only to areas manually selected (Leonardo Canvas, etc. etc.)

Add this to my over-long feedback novella, @fierce nimbus haha

[digital yoke]

Please swedish language

Fix

[edgy holly]

Gemini är jättebra på att översätta, så du kan bara skriva det du vill på ditt eget språk och kopiera/klistra in det Gemini ger dig.

[digital yoke]

No the pronounce and simple word like skatt it make wrong on. That Common words

[wide ether]

Any chance there we will be able to download all of the videos for a given project in flow like we can do with whisk? I hate that you have to go through each video individually and download!

[golden shadow]

Must admit I hardly ever use the 'edit image' feature. I almost always use image reference and re-use prompt on the 'Images' tab. Main reason is that you can select up to four outputs and quickly roll lots of generations. Given the chance of any edit coming out as required first time is pretty low, working one image at a time just takes too long. The extra tools available in 'edit image' aren't enough to make that workflow worthwhile for me.

[earnest crown]

Hey guys did anyone know how we can have the same voice generation in the video while jump to another frame in scenebuilder

[midnight cedar]

whoa. thank you for this detailed feedback! it's extremely insightful.

[midnight cedar]

oh, that makes a lot of sense.

[turbid lintel]

i think dropdown menu for camera angle is a great tool for flow image generation , just type the prompt and then choose the angle

[spring oyster]

created a 40 second scene and it gives me export failed. any ideas on how to fix?

[quiet kindle]

I’d like to clarify something about Flow’s image-to-video process.
If an image larger than 4K is uploaded, does Flow downscale it internally?
And if it does, what resolution does it downscale to?

Ultimately, I’m trying to figure out the optimal image resolution to use for image-to-video generation in Flow @slate vapor

[edgy holly]

More in your DMs, if interested.

[quiet kindle]

this is very interesting, fren. You seem really knowledgeable about this, so I wanted to ask —
which platform is currently best for image style transfer?

I want to keep my original image’s details and structure, but apply different art styles.
Midjourney doesn’t seem to support this properly, and even if I get a good style there, editing it later in Nano Banana Pro often breaks the style.

So I’m looking for the best img2img platform. Any recommendations?

[edgy holly]

Specifically for style transfer, I've found Leonardo to be best from among the bigger apps, but unfortunately their new models only allow style transfer from one image applied to a text prompt (no other reference images), while with Nano Banana you can use as many as I think 14 input images with your prompt.

Nano Banana Pro has the others beat on pretty much everything except style transfer (Leonardo), artistic depth (MidJourney) and user interface (all the other high-end AI image apps have a bunch of useful settings, not such a strange over-reliance on the text box.

[hard fossil]

we need an update about the situation of exporting failing in scenebuilder

its already made 2 days that i can't export my films

[quiet kindle]

thank you, my fren. !🫶

[edgy holly]

Happy to help. I use Leonardo some still sometimes, but mostly Nano Banana for keyframes and ingredients. Works for my real-ish noir project.

[fierce charm]

its annoying as it takes 100 export fails before i can export a video from the flow. any tricks?

[stuck oasis]

Excuse me, I clicked to download, but I got an error message. What could be the reason?

[rocky forum]

Same here

[digital yoke]

Not True

[digital yoke]

We need context base(data extraher) say you do Comercial for new website or area and service generate local deals this is so crucial when can we get that. Specially shops or store or places that use Google places ( Google business profile) can you share that feedback?

[rare latch]

Same issue, i had to record my screen -_-

[alpine halo]

I've had the same problem since yesterday. Any ideas? Is there a problem? My credit balance has dropped to zero

[stuck oasis]

What has happened with Flow recently?

[surreal ruin]

Anyone here, especially from the USA, running a YouTube channel on spirituality, philosophy, psychology, metaphysics, etc.?

[past oxide]

Same po

[tawny brook]

Same

[frozen totem]

flow down ? everything i do in flow get "something went wrong!" in every my account.

[brittle pelican]

Same

[turbid umbra]

it's working for me, and i'm working with the low prio one.

[alpine halo]

I have Google AI Ultra and it still doesn't work. It's crazy to pay so much for something that doesn't work.

[clever sapphire]

😭

[honest shore]

I feel like it’s getting worse this week because too many new users came in through trick versions just to get ultra cheap plans, and now the system’s overloaded

[clever sapphire]

hurmm

[crisp basalt]

This

[turbid umbra]

its just in your mind bros, I got ultra as well, it sucked from time to time but eventually worked again, I for one am really grateful it's an amazing tool.

[restive rain]

👍

[narrow sigil]

are you guys using flow. mine is not working

[surreal ruin]

If anyone need youtube video editor, you can hire me. I mostly edit short documentary type video with ai footages.

[crystal talon]

How is Animating this against the policy?

You people at google will destroy nanobanana and veo3 from within. You guys are crazy. So obviously an agenda going on here. On all ai discord channels, people are complaining about nanobananas policies. Nothing is allowed. Keep going. You will destroy yourselves.

[sharp orbit]

Why can't I download the video?

[rustic temple]

TRASH lol

[exotic stump]

flow down ? everything i do in flow get

[rustic temple]

Seems Google thinks any animated character is now 3rd party content

[urban wasp]

I can’t access Flow. Is that normal?

[quaint parcel]

How does everyone organize their projects within Flow? Our projects are starting to add up and its getting difficult locating older ai generations if we need to make a change to an older project. Project folders would be nice.

[sharp orbit]

Is anyone still able to use FLOW normally?

[edgy holly]

was last night.

[edgy holly]

Thats one of my suggestions aswell

[sharp orbit]

I am too.

[crystal talon]

Is nano sick? No matter what i prompt. He thinks the president is kevin spacey

[verbal jacinth]

I thought Kevin Spacey was the president..

[crystal talon]

Nano is terrible. Then it says no celebrities allow. Which mongo coded this?

[crystal talon]

It's terrible. Google products are just terrible.

[rustic temple]

I havent' used flow for a few weeks, I guess something has been updated? I guess Google is going down the route of Chatgpt where everything is against the guardrails lol

[thick oriole]

We're not! Did you see any more specific error messages pop up in the top right of the screen when generating? Or if you want to DM me your email address I can see why these generations failed.

[rustic temple]

the first few failed were because I used a disney character, but when I changed the character to someone not from disney, it still got caught up

[thick oriole]

At a guess from the small screenshots, that's Judy Hopps?

[rustic temple]

The first few, yes 🙂

I guess if I create a bunny that sorta looks like her, it still gets caught up

[thick oriole]

Who knew taking my kids to Zootopia 2 the day it opened would prove professionally useful. Let me see if Google will pay for those tickets 😅

[rustic temple]

Yes, you should absolutely get a reimbursement

[thick oriole]

Gotcha, that seems most likely then, false positives in the content filters if you have something that's close to that. I don't have any great advice there I'm afraid, we have a pretty robust process for finding those false positives and improving the filters but it's iterative and takes time, and will never be perfect

[rustic temple]

at least I can still do spongebob haha

[thick oriole]

Lol, no comment on specific IP, that's way above my pay grade, but keep letting us know whenever you see any unexpected or unexplained failures and we're happy to take a look, see what we're able to improve.

[thick oriole]

Yes! We're working on some revamps to how projects & assets can be organized. No specific timeline to share, but open to any suggestions there, and hoping we'll deliver something that'll make this a lot more useful, particularly for power users.

[thick oriole]

Could you give any details on what you mean here? What screen do you see when you try and access flow? A screenshot would help.

[thick oriole]

If you reach out to support via that button, they can help debug

[thick oriole]

Yes, we're still dealing with some issues in exporting from Scenebuilder, sorry for that. We've had some improvements this week, and hopefully (no promises) today's the day we get it fully stable. Will send an update here once that's out.

[thick oriole]

Me! Amongst others. The issue is that your videos aren't being returned because of audio errors? The prior state was that we would have returned them with silent audio and you wouldn't have been refunded credits. It seems some users prefer the silent video + credit cost, some prefer the no video + refund. (we can't offer silent video + refund because of various abuse scenarios). Most likely we'll just make that a toggle so you can choose which you prefer, working through that now.

The general fix is to make Veo not return very-low-quality audio. We're working on that on the model side, and my main advice on the prompting side is to make sure you're specifying a full 8 seconds of audio, not just a brief phrase e.g. "He says hello" where Veo needs to imagine the next 7 seconds and sometimes ends up generating gibberish or very weird noises.

[wide ether]

@thick oriole any idea if/when we'll be able to download all favorited videos in one operation in flow, similar to how you can do that in whisk? If that's not ever going to be a thing, is there a public API for flow projects that I could use to roll my own scripting to be able to download favorited videos in each project? I'm dredding the task of manually going through each video and downloading them all individually

[thick oriole]

As part of the general improvement to project/asset management I mentioned above, yeah, we'll have various multi-select and batch oepration views that should help here. No, we don't have a public API for Flow and I wouldn't expect one. Though, I'm hopeful we'll find ways to allow for customizable UI for cases where our default UI can't imagine/support every use-case. That latter part is probably not very soon though.

[alpine halo]

I can't do it

[thick oriole]

What do you mean? The button doesn't work? Did you pay via Google One or via Workspace?

[wide ether]

so multi-select and batch is on the roadmap, but not near term?

[alpine halo]

The button doesn't work when I'm asked to log in as an admin; it doesn't work when I select my Google account.

[thick oriole]

Multi-select & batch is fairly near term (no promises on dates). "User defined Flow UI" is not near-term but I'm hopeful we get there.

[wide ether]

ah, cool ... thanks for the update! Much appreciated!

[thick oriole]

I'm guessing this is a Workspace subscription then? If so, I'd suggest contacting your Workspace admin or, if you are the Workspace admin, then contacting Workspace support through their main site. Sorry, not really sure why the button/flow isn't working for you.

[alpine halo]

There's nothing to do except contact my administrator

[thick oriole]

My suggestion here is 'positive prompting' vs 'negative prompting'. You're telling Veo to generate the audio "Ta-da" and then no additional speech. But you're not telling Veo what audio to generate. And, in my experience, it doesn't do a good job respecting "be completely silent". So, although I can't inspect exactly what the model is doing, my guess/intuition is that it's imagining the next 7.5 seconds of audio, probably as meaningless gibberish/noise, and then our "this is very low quality audio" filter is triggering. I'd suggest specifying what audio you do want after 'ta-da', even if it's background noise or very light audio or similar.

[golden shadow]

@thick oriole I've posted about false positives on public figures in labs-bugs-feedback under 'Flow content blocks more restrictive'. I would post the link to the thread here, but when I try Discord blocks my post for some reason.

[edgy holly]

Funny how that happens, more often than people might think.

[little fulcrum]

I just number my projects individually to keep track, but it would be helpful to have folders and other ways to organize by subject

[edgy holly]

I had a section in detailed feedback I submitted recently devoted to this specifically. I called it sometihng like the 'Just Make the Damn Thing Work' team, haha

[raw oak]

Hi. the 720p version is actually better than the upscale 4k on this which cost me 50 credits. Any tips on getting better results in 4k?

Hi. Has anybody else noticed that 4k upscaling creates artifacts... or is it just me. thanks

[thick oriole]

Scenebuilder exports should be working reliably again now. Sorry for the inconsistency & errors this week with that. Turns out dealing with videos is hard! (but we're the video generation product team, so we should do better, thanks for the patience)

[raw oak]

Hi Ian. Thanks this wasn't dones in scenebuilder just regular first frame in home page of projects

[edgy holly]

Replacing the Library would be great. I submitted some ideas for ways to do that to Jaime and Nick, but since we get a TON of cloud storage, an easy workaround might be that any clip we heart/like would be added to a folder in our GDrive, with some form of metadata so any of the inputs (prompt, frames, ingredients, settings, etc.) would be indexed and searchable. Or just make a proper Library, of course.

[edgy holly]

I don't want to overwhelm the room with nitty gritty prompting chatter, but probably half of the text in that prompt is superfluous. Tightening it up would be a big help.

[turbid umbra]

Honestly, in my opinion straightforward prompting without a lot of detail works best, I make sure all the detail I want is in the nano banana generation that I use for frame to video, or ingredients to video.

[edgy holly]

100% - Good reference imagery with relatively streamlined prompts usually get the best outputs in my experience.

[wintry pine]

My rule of thumb: do not put anything in the prompt you do not expect to see in frame, literally

[edgy holly]

A bit of mood/feel stuff at the end can be a value add, but 90% agree.

[wintry pine]

I like to have a dividing phrase for these things like "Aesthetic: Neon Noir" or "Style: An Oil Colour Painting"

[edgy holly]

I usually don't do this at first if it's frame to video or ingredients to video, but if initial prompting doesn't work well enough, structured prompts are one of the next steps.

But with Nano Banana Pro, that's my go-to: structured but not JSON.

I think it was Nick who verified for us a few months back that JSON doesn't have any effect, but organized structured prompts do.

[wintry pine]

I agree, it's a different story if you're starting from a reference image. Unless, of course, you intend to change the style or aesthetic of said reference; though I would recommend Whisking the reference so Flow doesn't get confused

[edgy holly]

Whisk is neat, but its style and composition reference are too loose for me. Nano Banana, especially since Pro, is a lot better, especially on composition. Sometimes use Leonardo for style stuff.

[wintry pine]

For sure, you can do a lot of neat things - especially with photos - using Nano Banana Pro

[turbid umbra]

I think the main problem, and I caught myself doing this more than once, is that people think they want something but they don't have it written down in the prompt or in the refference then they get mad that VEO3 didn't read their mind. Once you get out of this mindset, miracles happen.

[edgy holly]

Yeah, and you see it in the way people ask for help on here as well.

Their questions are super vague. They don't share their prompt or inputs, but they ask people to tell them why their prompts aren't working...

[fiery panther]

Flow down again?

[hard oasis]

Is there a solution to fix this?

[turbid umbra]

yes: 💸 🫰

[brittle pelican]

Yep, same problem and question

[hard oasis]

?

[turbid umbra]

check country availability

[hard oasis]

if this were the case, it would be written that it is not available in the country

[quaint parcel]

I think even the 1080 upscaling looks worse than the 720. I'm not sure what the barrier is to generating natively at 1080 is anyway.

[wintry pine]

720p produces a smaller file, which is ideal for a lot of file transfer options

[edgy holly]

Mostly a cost thing. Cost scales, and from what I've been told most 'native 1080p' generations are actually 540 or 720 that was just upscaled before being shown to the user.

[wintry pine]

A fun little project for you if you do a lot of upscaling: open up Colab, create a Python environment, and run Real-ESRGAN

Within an hour or two you'll be able to upscale images to however big you want them

[noble sluice]

Is it major Difference between Veo 3.1 Fast and Quality option’s result?

[raw oak]

They are charging 50 credits for 4K but lately it’s like the old 1080p upscale but worst. I know they’re working hard to patch things up. Let’s hope it gets improved soon

[edgy holly]

The free/1080p is better than other free ones I've tried (hit or miss compared to what I can do in Davinci locally, depending on how much time I spend tweaking the settings).

Neither are anywhere near as good as Topaz, but 50 credits is waaaaay cheaper than Topaz, so apples to apples seems like a decent deal.

[tropic breach]

Can anyone generate videos in FLOW right now? I can't generate any videos; some error occurs immediately.

[icy drum]

Google is shooting itself in the foot by blocking advanced prompt engineering.

[edgy holly]

That's a new one. I haven't seen anything like that.

SMH

The spike in failed generations started yesterday. Has nothing to do with whatever you're making up there. There are no rules against deep prompting.

[icy drum]

This as been an ongoing issue for everyone? So am I entittled to any sort of refund?

[spice totem]

Opinions?

[hazy sedge]

Can't upscale, can't generate images, can't generate videos right now. Instant error. Am I the only one experiencing this right now?

[edgy holly]

Scroll up a bit.

[icy drum]

They cant even provide proper service to what they advertise. 🙁 Major letdown

They probably will say "did you read the terms of service"... did Napoleon read the terms of service before he failed? is what I will reply with

[sour sapphire]

Hi

[urban patio]

No buddy, you're not alone. I just wasted like 6,000 credits in 2 days, with 0 output from flow

[crystal talon]

Does flow have CPP ?

[little stone]

flow is down now?

I got "something went wrong instantly as I click the button"

[obtuse osprey]

is flow still down for ultra user? I got "you cannot access flow"

[long locust]

I'm a Pro version user. How is the daily quota for nano banana calculated? Is it based on Pacific Time? It notified me last night that I'd reached the daily image generation limit, and I still can't generate images now.

[analog lava]

Hi @scenic notch , 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!

[jovial tendon]

been up for me all day

[rotund helm]

Any idea why I'm unable to open flow all of a sudden? It was alright until yesterday. Not able to see the "Create with Flow" button that used to take me to the app. I'm a PRO user.

[brittle pelican]

In my case it resolve itself. (Само прошло спустя 2 дня, видимо у них сервера упали и ограничили доступ некоторым)

[edgy holly]

You have to go to the Flow app to run prompts.

[pallid violet]

@ruby oceanHello, as a friendly reminder we cannot generate images or videos in the server! Please do not post your prompts in the channels as it will not generate. If asking for advice or tips on your prompts do so in the appropriate channels.

[hasty cloud]

When will it start working? The eternal problems of such a powerful and expensive company, and nothing is working

[mighty rock]

Is free veo 3.1 fast real or is always heavy load? Just seem this in a forum "Thanks for your patience as we work through high demand for video and image generation. We’re now reintroducing a zero credit Veo 3.1 Fast option for Ultra users. Ultra users will see two Veo 3.1 Fast models in the model dropdown:

“Veo 3.1 Fast”, costing 10 credits. As always, credits are automatically refunded for any failed generations.
“Veo 3.1 Fast [Lower Priority]”, costing 0 credits; generations get deprioritized during heavy traffic and may have lower limits on concurrent generations. We’ll be experimenting with different settings over the coming days.

This gives Ultra users a way to generate for zero credits, while prioritizing paid generations under heavy traffic. As a thank-you for bearing with us, on November 25th Google AI Ultra subscribers received an additional 2,500 AI credits added to their account!
-The Flow Team"

[wintry pine]

Sounds like a way to compensate Ultra users, since many use Flow for commercial or other enterprise purposes

[terse crystal]

How to fix this, guys?

[wintry pine]

Sounds like Flow might not be available in your country

[crystal ledge]

how to ask for refund for failed generation, lost 400 credits today without noticing..

[brisk cloud]

same here

[wintry pine]

<@&1366831844502274180>

[wintry pine]

🫡

[winged plinth]

Don't worry, failed gens are auto refunded. You’ll see the credits back in your balance.

[timid rapids]

am i able to swap my face onto a video on flow?

[wintry pine]

You can try by uploading photos of yourself in "Ingredients to Video" mode

[timid rapids]

am i able to upload the video anywhere?

[wintry pine]

If you download the video from Flow you can then upload it somewhere, yeah

[timid rapids]

oh no lol, as in the video i have and i want to swap my face into the video. where can i upload that video into flow?

[vernal ridge]

When I log in to my account, I get the error "You don't have access to Flow
Review Flow access requirements at the Help Center or contact Google Support.Couldn't load your credits. Please try again later.", and when I try to log in to another account without a subscription, everything works fine. Can you help me?

[wintry pine]

I don't believe so, no

[hazy sedge]

It's been 18 hours, I still can't generate or upscale anything. I basically cant use flow with my "Google AI Ultra" plan. What a waste of money. Is it just me?

[raven siren]

I don't know if this is the place for a feauture request, but i'm going to do it anyway..... Please make drag and drop for image upload work. Comon guys, you are Google. Put some effort in your products. allegedly you've got all the best talents working for you and you're still letting payed customers work with a beta interface. It's anoying and disrespectful.

[fickle drum]

i worked with it all day, until 1h ago. now it s working again

well...worked only 1 time :)) now i get errors back

[twin mortar]

Is anyone having issues with face generation? I’m getting Will Smith Eating Spaghetti levels of nightmare fuel.

[toxic willow]

Hi, Im not sure how to ask or get advice, but im using flow the tool VEO 3.1 to generate a scene where i need it throw an object from a person hand and fly into the air of the screen front camera angle or the side, but i cant get the prompts any easier or clearer for the AI to read, any suggestions on like moving objects and the prompts on those?

[wintry pine]

Try making the object the subject of the prompt. Rather than "one person throws a ball to another", "a ball is thrown from one person to another", since the ball is what you want the AI to focus on

[toxic willow]

So i am using a image to video, and then a clear prompt explaining, but the object just doesnt leave the hand, then it adds a new object in the air its just not flowing with the prompts i use.

[wintry pine]

That's a very common problem with Veo. You may have to do a couple generations, and really emphasize you are referring to a specific object in frame

[wide ether]

Ugh, now comes the point where I have to perform the very daunting task of downloading all of my favorited videos in each project .... one by one (Because that's the only way currently 🤬), manually ..... point and click cramping here we come .... it's 2026, Google you can do better here!

[dim solar]

Feedback: All the partial images I generated earlier have disappeared. When I click into the project, it's blank. Why? These images are very important to me.

[wide ether]

that's why it's important to download resources every once in a while

[wide ether]

I just wish they'd add options to download favorited resoures in one operation instead of having to do them one by one!!!

[hazy sedge]

jesus flow is terrible. cant even use my ultra subscription.

[edgy holly]

You can do that in Nano Banana, with some cutting and pasting

[near gate]

Does anybody know how to fix this? When I go to use my uploaded images a bunch of them aren't showing up. Usually it comes back after trying a few times but this time I seem to have lost half my collection of pictures.

[edgy holly]

I've run into this several times as well. Sometimes they're there if you go to the same screen where you initially uploaded it, but sometimes I just have to re-upload.

[near gate]

Thank you

[ruby roost]

patience. with their recent sale and now huge electricity demand because of cold weather, systems are likely overwhelmed and underpowered.

[edgy holly]

I'm not so bothered by outages. My issues are with the app, not the models.

[ruby roost]

i see the two as related.

[wintry pine]

I think there always needs to be a level of allowance for unreliability with technology, especially new technology. Closing down at the first glitch is going to bar you from a lot of great stuff

[edgy holly]

Web apps like Flow don't take but a minescult fraction of the processing power, and run on normal hardware that isn't in any way bottlenecking, not the GPUs that are topping out that the models run on. Flow is very poorly built. There are issues with it that no other major competitor has.

[wintry pine]

Very true. I think that Flow is running into this problem where it is trying to do multiple things as different things at once, rather than a collection of things as one thing. If that makes sense

[edgy holly]

I use it because the models are great. I don't mind waiting a bit when the GPUs are topping out. The Flow web app is the problem here.

[wintry pine]

I'm still cautious about Flow for reasons beyond the web. Its audio reliability and accuracy among extended scenes is still really lacking

[ruby roost]

i'm not getting anything to generate at the moment. So i'd much rather have generation with a buggy app.

[wintry pine]

Yeah it's been very hard the last few days especially. Servers are crammed, consistently crammed

I hope Google is taking notes and does a rethink about how it deploys Flow as part of all of this

[edgy holly]

For work I'm serious about, I rarely keep any of the audio, from any of the models.

[wintry pine]

Sora is a bit better about audio and consistency, but overall video generation is primitive across the board

Flow capping out at 8 seconds and not being reliable in scene extensions is still a hurdle for long term viability

[edgy holly]

Even with this, most apps put generations in a queue instead of failing them out. That's a choice. Not a good one. They're the only major to do it this way.

[wintry pine]

The big issue with the web app is that it's still designed for the experimental stage. It's not "ready" as a consumer end product

[edgy holly]

I'm fine with 8 seconds, and usually can get what I want from Extend.

[wintry pine]

You'll have to show me sometime what the secret to good extensions is because I don't think I've cracked it

[edgy holly]

If you want it to be consistent, it has to be seen in the last second. Usually get what I want within 4 gens, and almost always within 8-12.

Then just copy/paste the previous prompt, and change as little as possible to get the outcome you want so the interpretation will be as similar as possible. Keep the prompts short and simple as you can.

[wintry pine]

So extend the final image rather than try to extend the whole clip you mean?

[edgy holly]

No. The way I was told Extend works is it's based on the last second of video, so if what you want to be consistent isn't there, the model doesn't have it in context.

The last SECOND.

This is why an 8 second clip extended = a 15 second clip, because there is a one second overlap.

[wintry pine]

Ahh

Interesting, I hadn't thought of that

[edgy holly]

I don't think that's in any of the docs, or it's buried. I only know because I asked.

[wintry pine]

It's something I will have to keep in mind. Because I'd tried both "do this to the clip", and putting the same prompt in the clip, and both led to different videos but consistently different types of different videos

So the first kind was a surreal mess, and the second was more of a parallel clip than an extension. My next thought was the image, but if it's the entire second that changes everything

[ruby roost]

If you do an extend and then drop both clips inside an editor you will see that you need to cut the last second (precisely) of the first clip or the first second of the second clip.

generations are working again ttyl

[edgy holly]

It's a shame they don't have proper docs, but none of them do. Hailuo has a nice Notion notebook, but most AI providers seem to not have a problem with people being confused about how to use their models/apps. Seems a big missed opportunity.

[wintry pine]

I'm of two minds. As a user I'd like more accessibility, but the more accessibility the lower the bar, and I'd like to change the public image of AI has soon as possible

Canvas in Gemini is an undersold feature, NotebookLM is a fine knowledge source module, and the workflow management potential to stratospheric, but adoption is - in my view - being held back by public image

[edgy holly]

That's outside of anyone's control. Sloppy stuff is a reflection of the ideas people have and want to use these tools to make much more than the state of the models or apps.

[wintry pine]

It's something we see in every new technology. People try to use the new thing in the old way. It takes time for people to find the new uses. TV didn't until the Sopranos, arguably

[edgy holly]

As they get easier to use, that slop ration will keep skyrocketing. Even worse than now, by multiples. Has its ups and downs.

[wintry pine]

One thing I like about Google is they're not stopping there at least. They're exploring the workflow and research tools and I think that's a good sign of things to come

[edgy holly]

Personally I think Sopranos is overrate, but very much agree on the point. Oz came before, and was better IMO. There were more classic TV shows, but that's roughly where this era prestige drama stuff started.

[wintry pine]

Yeah I am really more referring to using the medium for the medium's specialties. Sopranos is a good benchmark, though I'm sure there were earlier examples

Video games are still really struggling with it, but games like Cyberpunk are motioning vaguely in that direction

[edgy holly]

Google is behind on the workflow stuff. Kling had extend before Veo/Flow. Other apps are ahead on that, but with weaker base models. Etc. Etc. Etc. haha

[wintry pine]

I meant for professional workflows, not for generative AI

Again, stuff like Canvas and Deep Research, and vertical integration across tools

[edgy holly]

I kinda liked those... what were they called... Telltale games. Garbage mechanics, but story rich.

[wintry pine]

This is why games like Cyberpunk are watershed for gaming. Good mechanics, story rich, strong themes, high aesthetics

[edgy holly]

Not a gamer, though. Was into Total War Rome II, but the last RPG tpe game I actually finished was Ryse: Son of Rome. They all get boring after a few hours.

Suno is a lot more fun, haha

[wintry pine]

My main point is really that technology takes time for its medium to meet its function

Sopranos and Cyberpunk were really just made to be illustrations of that in other media

[edgy holly]

Eh... I don't think that's why. People just aren't choosing to use tools in that way. Exponentially more people want to "clout chase" than make anything with substance, so we get tidal waves of slop... a thousand shows that come and go for every 1883 or Shogun.

[wintry pine]

Early television was just a continuation of stage media people were already familiar with: people playing musical instruments or doing plays. They were using the new medium the old way

People are doing that today with AI. They are doing the old stuff with the new tech

[wintry pine]

Let's not sit here and pretend we were in a golden age of high art and creativity before AI.

[edgy holly]

Nothing about the Sopranos or other prestige dramas is unique in how they use the medium of television. The difference was the seriousness of the writing. Storytelling fundamentals.

[wintry pine]

Writing, use of lighting, staging, camera work. There were a lot of technical aspects that had to go into making prestige television that had to be earned.

[edgy holly]

None of that was new to television.

[wintry pine]

You don't wake up and obtain nuanced understanding of technology. You deconstruct it, you experiment

[wintry pine]

It was primarily the province of cinema. Finding how to make that work for television, specifically in the constraints of television, was certainly new.

[edgy holly]

Not sure what your point is here, or with the bit about "before AI".

[edgy holly]

There are shows that are more cinematic, but Sopranos is not one of them.

[wintry pine]

Slop predates AI generation. It is not a problem created by AI, only worsened by it

[ruby roost]

anyone getting anything to generate??

flow specifically

[wintry pine]

I think you are really getting hung up on the illustration and not on the point

[edgy holly]

Obviously. And? Nobody said anything to the contrary.

[wintry pine]

What I'm saying is pretty straightforward. Let's not muddy it up.

• New technology is made
• People still have the ideas of the old technology
• Time passes and people learn to understand the new technology
• New ideas rise out of this new understanding

This is exactly where we are with AI today. People haven't gotten to a point where we understand the creative nuances of the technology, so we're still making the old technology worth of stuff

[edgy holly]

And? Nobody is arguing against any of that.

[wintry pine]

It wasn't meant to argue. It was meant as optimism

[edgy holly]

That's not a problem. Quality isn't what makes something slop. Better understanding of tools doesn't make spam any less spam.

Quite the opposite, actually. Some of the most technically proficient and leading edge users of AI media models are making some spectactularly awful slop.

[wintry pine]

Sure but better products are possible, even if public perception is currently that AI is a slop machine

I'm optimistic quality will come when understanding improves

[edgy holly]

Dor Brothers are probably the best example. All of their content is miles deep into slop territory, but it's technically spectacular.

[wintry pine]

And early cinema had basically no narrative but was visually spectacular

Early video games were incredibly basic but, for their time, technical marvels

You don't get masterpieces this early in a technology's lifecycle. The problem isn't the technology. It's time.

[edgy holly]

Haha, I don't know what conversation you're in, but it's not this one.

[wintry pine]

I beg your pardon?

[ruby roost]

frontier it's pretty obvious you are trying to pick a fight

[edgy holly]

You seem to be having an argument against "the meta".

[wintry pine]

No I am arguing against your dismissive attitude towards the creative potential of the technology

[edgy holly]

Quite the opposite. He's arguing against a bunch of stuff I haven't said anything about, and I've been playing along, and just laughing at the strangeness instead of fighting.

[ruby roost]

you're engaging with him and then dismissing his replies

[wintry pine]

Right you're above it all. That's totally not the last refuge of the defeated.

[edgy holly]

Case in point. I haven't said anything even a lilttle bit like that. You're lying about what I said to start a fight. Not interested.

[ruby roost]

he's a troll at this point.

[wintry pine]

I'm thinking so.

[edgy holly]

What are you even talking about?

I'm sitting here, eating snacks, sipping coffee, making music in Suno and having a random chat online and you're all over the place arguing with things I haven't said to bicker about something I don't even think, much less said here.

[hasty cloud]

Have you fixed it yet? You're giving me a powerful...

[edgy holly]

We're not even in an argument, much less a fight. You're on tilt, making things up to fight about, for no reason I can discern.

[wintry pine]

You should pay attention to what you are writing, because your concept of chat seems to be to halfheartedly dismiss what other people say then pretend you have made a point in doing so.

I think that's a very disrespectful way to approach good faith attempts at discussion

[hasty cloud]

why hasn't it been working for 24 hours???? what's the deal with this?

[edgy holly]

...what in the world are you even talking about?

This is a random chat on the internet. Just miscellaneous opinions about ai apps and television.

[ruby roost]

okay you too. It's failing for me as well. I had good luck last night and earlier today, but now it's failing on every attempt.

[ruby roost]

I believe it has to do with the severe weather event in the united states but i'm not sure.

[edgy holly]

You're acting like some ideologue on cable news. I haven't said anything even a little bit like you're saying. You're reading some kind of intent and tone that isn't there.

[wintry pine]

I'm not going to sit here and buy into your roleplay of you being above it all when you have been engaging in an entirely different way until you decided to change the script when you ran out of points.

[edgy holly]

I don't have any animus toward you at all, and don't have a clue why you're trying to start a fight/

[hasty cloud]

it's not a fact because they've been doing this for about three months. It's always crashing. It's not a serious tool.

[edgy holly]

Again... you're in a completely different conversation than I am. I haven't said anything even a little bit like that.

[wintry pine]

I'm not engaging in this anymore. This conversation is over.

[edgy holly]

This is you right now: https://www.youtube.com/watch?v=naleynXS7yo

[ruby roost]

This is the first time i've encountered failings like this, and i've used it daily or at least 3-4 times per week for last 30 days. I noticed a big spike in failures right after their 50% off anual sale, but nothing this bad.

[ruby roost]

this isn't a good look for you. Hope you get some rest.

[wintry pine]

Usage has been high the last couple of days. I think they have reached some sort of limit

[hasty cloud]

it's always like this. even when they sell it through the api to others, you have to wait 15 minutes and it crashes... It's just not good... It's a pity that this has caused the price of RAM to increase so much, and we need to deploy our own resources locally to avoid relying on monopolies.

[wintry pine]

I agree that Flow needs a lot of work before it becomes a good consumer product

[ruby roost]

I've generated about a 1,000 clips, and i've not had this experience until today.

[hasty cloud]

and the most fun thing is that when veo 3 came out, I generated 1080P native video of excellent quality in flow, but now that veo 3.1 has been out for so long, it produces 720p native video of lower quality than it did back then.... ))

[wintry pine]

My problems with it are that it's not very reliable

[wide ether]

okay just spent a couple of hours manually downloading all 1000+ of my videos, one by one 🙄🤦🏻

[edgy holly]

Flow has never made 1080p natively. You can, sort of, in the API, but like other models that's a smaller generation that's auto-upscaled before it's brought to you.

[wintry pine]

What are you going to do now with them

[hasty cloud]

to be honest, I'm already confused with all the fuss from Google))) but I definitely remember that I received 1080p, and I have the first videos in native quality. I promise that no one will say that this hat is like the one on your head or mine)))

[wide ether]

just tuck them away .... all favorited videos ... why do you care?

[wintry pine]

Just curious

[edgy holly]

They had an upscaler. It wasn't very good, and the new ones are a lot better, but it was there. In some clips where the amped highlighting didn't look bad it worked okay.

[ruby roost]

YOOO - it's working on a different browser

[ruby roost]

I got flow to work by using a different browser.

[hasty cloud]

t's like I have thousands of videos ))) I was even a cco creator of kling when it first came out... all this hype around it... but in reality, it's just funny... but it's all raw and just sucks our money, and at such a fast rate that I realized I'd chosen an expensive hobby, even though photography and videos have always been expensive )) but come on

[wintry pine]

Which browsers are working for you

[edgy holly]

Nobody has cracked an upscaler that's both cheap and good. Topaz is by far the best, but it's expensive. All the cheap ones make at minimum some alterations to fill the extra space, trying to assume what would be there.

[wintry pine]

It's true with videos for sure. The best tools I've seen are text based. Still lots of issues with video.

[ruby roost]

Firefox was not working . Edge worked.

[hasty cloud]

I've already abandoned blender.... but when I opened it and spent about 4 hours exploring it, I realized how great it is to have control over everything))) I'm thinking about taking a break until the new neural network updates are released, and I'm going to focus on normal hobbies like photography and traveling. There's too much AI in our lives. Good night. I've been waiting for this video to be created, but it's been a day and it hasn't produced any videos. Oh well.

[wintry pine]

Edge is the top choice for something in 2026? It must really be the end times

[ruby roost]

first time in my life i've used edge

[wintry pine]

I think that's a good choice if your focus is video. I still think tools like NotebookLM are great as is, but their usage is much different

[hasty cloud]

edge top))

[wintry pine]

Is this the year we Edge on main, boys?

[edgy holly]

Blender's great. I use a Blender-connected plugin called Character Creator to make 3D models of recurring characters. Super useful.

I'm terrible at animating and making complex alterations in Blender, though.

[hasty cloud]

in the old-fashioned way, a notebook with good ballpoint pens, a desk with a warm table lamp, and all of this by the window, where there is frost and a snowstorm. That's a song!)) creative opportunities and a warm mood in your soul

[wintry pine]

To you as well

[edgy holly]

Hybrid is the way. When I need to work something out, I go back to my old default of large notecards and mechanical pencils. A stack of them are always within reach.

[elfin oriole]

Hey guys, did your flow video generations working? I've been failed to generate a video for all these week except 1 video...

[wintry pine]

Yes, it's been very common the past few days

[elfin oriole]

what is happening?
nano banana still working by the way

[wintry pine]

My theory is that it has to do with servers reaching their capacity limits. Though @ruby roost got it working switching browsers

[ruby roost]

My generations failed all morning, but I just got them all to work again by switching to a new browser. Firefox was not working. Edge is working flawlessly

[wintry pine]

I think if we see more people get it working on Edge, then we'll confirm it. Still the best theory so far.

[elfin oriole]

ah right, thank you guys for your solution, I'll try it...

[ruby roost]

let us know

[wintry pine]

Let us know if it works

[ruby roost]

my guess is it works lol

if we haven't heard back from him

[tender sleet]

does anyone knows what happend with veo 3.1? I tried to create videos but all of them fail

worst part my credits are not returning

[edgy holly]

If they don't get refunded soon, contact support.

[crisp basalt]

A

It’s back ,

[obtuse wyvern]

Hi

[neat rivet]

hi

[wintry pine]

I like it, looks cool

[versed garnet]

What does this even mean? It was working 2 days ago... "You cannot generate a response to this prompt due to Google's guardrails related to third-party content"

[jolly halo]

anyone knows to free voice over site?

[crystal talon]

Is anybody working at google flow actually in here?

Why does my stupid actor always turns head when talking. I tell it not to move head, only to look outside window. But she always turn and reveals her face. Tried everything possible even negative prompts. Why cant veo do the simplest things?

[fickle drum]

in negative prompt, explicit this action

[crystal talon]

I did

[fickle drum]

give me ur prompt and reference, i ll try it too

[crystal talon]

She always turns head, no exception.

[fickle drum]

try with start and end frames

[crystal talon]

@fickle drum i dm you not to lose your chat. Am outside right now

[viscid junco]

Can we do Motion control with Veo3.1 ?

or Any google's Ai Tool can we do Motion control

[full flicker]

Hello, as a friendly reminder we cannot generate images or videos in the server! Please do not post your prompts in the channels as it will not generate. If asking for advice or tips on your prompts do so in the appropriate channels.

[viscid junco]

Hey can You help me with my Doubt

[full flicker]

I'm sorry I don't understand the question.

[viscid junco]

Can we do Motion control with Veo3.1 ?
or Any google's Ai Tool can we do Motion control

Like give a Reference video of Motion and simulate it on a Given Anime Image

[full flicker]

That may be a better question to ask in #flow-tips-n-tricks・✏️ . The moderators are here to help with server related issues.

[quartz echo]

Pixar-style 3D render of a smartphone wedged inside a tight pocket.Eyes: tired, slightly bloodshot,

[hasty ermine]

is anyone’s flow working ?

[blissful seal]

https://cdn.discordapp.com/attachments/1461043776523862301/1464977670239555709/I_want_a_202601251337.gif

i love flow: gives me free veo

[frosty comet]

Has anyone noticed that nanobananapro is different in flow than in the geminiapp?

It seems smarter in the app.

[rotund elm]

how do you prompt to eliminate slow motion and use regular speed?

[edgy holly]

Definitely. Has some kind of prompt expander that's quite bad.

[edgy holly]

I've never gotten slow mo to work in Veo. Usually the issue is the other way arond.

[frosty comet]

why doesn't google implemenet whatever works best everywhere

[clever axle]

good evening everyone

flow award me with a new msg that dont allow me to generate :/

"You cannot generate a response to this prompt due to Google's guardrails related to third-party content"

[tawny sandal]

Hi everyone, I need help with something. I'm making timelapse videos, ASMR-style videos. Out of 20 videos, only 2-3 of them have perfect sound effects, while the remaining 16-17 use background music. I haven't been able to solve this problem. Normally, I can find a solution, but it takes an incredible amount of credits to reach it.

[winged plinth]

To get the best results, you’ll need to pair an audio prompt with your motion prompt. The model responds directly to the level of detail you provide.
For professional grade output, ensure your audio prompt covers these four pillars: Dialogue, sfx, ambience, and mix.

[tawny sandal]

I make timelapse repair videos, showing things like workers hitting metal, digging, cutting wood, and the final result is really satisfying, but it's full of music. Could you give me an example?

[winged plinth]

ASMR "Soap Cutting" audio prompt:

PRIMARY ACTION: Crystal clear, "Macro-Lens" audio. The sound of dry, chalky soap being sliced into tiny cubes.
COMPLEX SOUND DESIGN LAYERING:
Layer 1: A hollow, dull thud as the knife hits the cutting board.
Layer 2: High-frequency crunching and snapping as the dried soap breaks apart. Crisp and brittle.
Layer 3: The sound of cubes falling and tumbling over each other like dice.
AMBIENCE & MIX: "Dead Studio" silence. Ultra-Close Proximity. The slicing is Center Channel. Boost high frequencies for crispness.```

[edgy holly]

Why do people you don't know do the things they do?

[tawny sandal]

Thank you, but I produce the video frame-by-frame from text instead of video. For example, in the first image there are disassembled wooden pieces, in the second image there's a repaired table. I write a prompt saying 2-3 workers quickly repairing the table [timelapse], I add ASMR sound, one worker hits with a hammer, another uses a drill, I write it in a timelapse style with no music, I do 15-20 of these and then I get the desired result.

Thank you for your help and interest. I've created 8 more videos with these prompts, and added music again.

added music to 3 of them, and 5 of them were completely silent.

[winged plinth]

Would you mind sharing the input images and the motion prompt you used? I'll DM you now.

[wide ether]

What's the story with videos now being generated in fast forward speed? The same types of prompts I've used for quite a while now seem to generate videos where the playback speed is like fast forward. How do I stop that?

[hallow gate]

Hello, good day everyone. I just want to ask where I can message support. I’m having a problem with Flow, it keeps saying “something went wrong.” This has been happening since yesterday, from when I bought it until it expired, and I couldn’t generate anything.

[pale wolf]

Video creation on my Beo 3.1 is currently not working. I only get a pop-up message saying "Something went wrong." This issue has persisted since last Friday. I have an Ultra account and have close to 20,000 credits remaining. Is there a solution?

[pale wolf]

My internet speed is close to 1 Gbps. It's not an internet issue at all.

[mossy iron]

Is it possible to make a character speak with the same voice or a similar one in each generation? I'm using Flow Veo 3.1.

[gray sluice]

???

[shell heron]

What happen to the server?

[minor badger]

your daily dose of Flow struggle.

[minor badger]

best is to use a third party tool to replace the voice so you can keep consistency. cf. elevenlabs

[minor badger]

been a daily issue for everyone, no matter the plan, for weeks now, if not already months.

[pale wolf]

Is there a workaround that effectively resolves this issue? This issue has been around for a long time, but it's gotten much worse this year. It used to be resolved in an hour or two, but now it's completely unusable.

[chilly epoch]

Help me guys 🙁

[serene cradle]

https://tenor.com/view/hang-first-time-smiles-gif-14991372

[warped wind]

Again and again flow is not working :/

[pallid violet]

@forest bearHello, as a friendly reminder we cannot generate images or videos in the server! Please do not post your prompts in the channels as it will not generate. If asking for advice or tips on your prompts do so in the appropriate channels.

[turbid umbra]

anyone here hear about Clawdbot? I'm curious about some integration with Flow or Gemini as a productivity tool in helping to do research for tiktok and video generation.

[trail flame]

Am I only one here with this bad boy 😁

[turbid umbra]

showoff

I got ultra as well but but mine looks like this for now:
:)), waiting for 31st ❤️

[honest shore]

Did you purchase your plan using a VPN?
Usually users without profile pictures gmail tend to be on cheaper plans, and they often cause server overload. hehehe just asking😋

[trail flame]

Bro I got it for way less then the original. I still paid for it But I don't know how they are doing this for so cheap.

[trail flame]

I think you use Flow heavily how much it costs you?

[turbid umbra]

I do, I used up all of my monthly 25k credits, I got the ultra offer for the 3 months.

[trail flame]

I got 45k credits though..

[trail flame]

Are you somekind of youtuber

[turbid umbra]

I'm making tiktoks

[trail flame]

cause 25k are alot if you are not using Veo3 Quality

[trail flame]

We are on the same pitch

[turbid umbra]

I used mostly the fast model, maybe 20% quality

[trail flame]

Is your tiktok working fine?

[turbid umbra]

However I usually generally generate 4 videos at a time, until I get what I want.

[trail flame]

CRP dashboard all are gone in my account

[turbid umbra]

I've done about 3.4 million views since the 29th of december I started, It's pretty cool, I've created a persona, I'm not posting like ai slop without any thought.

I want to develop this persona, people love it already, I post about 2 times a day, respond to comments etc, it's kind of a full time job, but I want to see how far can I take it.

[trail flame]

Check the DM

[tough yacht]

Can you tell me more about what's happening?

[tough yacht]

If you refresh, do the videos show up as generated now, or still stuck at 99%?

[tough yacht]

I can look into this for you - can you send me some specific prompts that are consistently getting sped up?

[edgy holly]

Flow doesnt have an API, but Veo and Nano Banana do.

[warm knoll]

Hi, someone told me today that the videos I create on Flow may end up on FlowTV. Is this right? I would like to confirm if what I create on Flow stays private to my account.
Thank you for your help

I have a Google Ultra subscription.

[stuck adder]

Any updates to flow that’ll make me
Get ultra again?

Veo 4 should’ve been out by now

[trail flame]

Is someone else also facing the same issue with Google Nano Banana Pro Model on Ultra account while the same prompt can generate images on Gemini with same Nano banana or is it just me ?

[wicked sapphire]

would be great to control the length of a shot. If i have beginning and end frame and a quick move to get from A to B, Flow currently wants to fill all 8 seconds. SO if you have any quick shots they don't work. if you could choose a 2 second shot, it might create a better output.

[fiery panther]

Yeah just adjust some content of the prompt cuz it gives a false positive

[trail flame]

But the same prompt is being generated in Gemini App but not in flow I am soo frustrated @potent shell

[stiff ledge]

Any plans to be able to distribute AI credits to other accounts? I might partner up with a guy to start teaching troubled youths computer skills and how to create AI videos but I'm not sure how I can get them credits that are donated. We're in Fort Wayne, IN too so Google seems to be getting a lot of negative press for the data center they're building here so this might help with that.

It would be awesome if I could use an API to let people donate to specific kids.

[wintry pine]

I think making transferable credits might be a massive legal labyrinth, and based on Google's history it is unlikely they would take such a risky path

[subtle isle]

Is there any update regarding Veo 4 launch

[edgy holly]

I believe they ask before they do that.

[edgy holly]

That's the idea!

[edgy holly]

If you don't need that many, you can start with Pro and pay for credits only, or use one of the third party services that accesses Veo via API (that costs more, though, per generation).

[edgy holly]

They rarely pre-announce things like that.

[fiery panther]

flow is down??

[warped birch]

flow is down

[quaint olive]

just created an image in Whisk of a Sales representative. I saved it out and brought it in to Flow. It says I can't use the image because it prohibits the policy of using "prominent images", thus ending my job progress. Is there something in an image that might cause this?

[edgy holly]

Maybe it looks like some popular stock image that's in it's database of IP?

[full flicker]

Hi, here is the link to the FlowTV FAQ and the link to the terms of service. The important part that you are concerned with would be the Content in Google services. I hope these help. https://labs.google/flow/tv/faq https://policies.google.com/terms#toc-permission

[warm knoll]

Thank you! So my data stays local to my account

[restive needle]

In Flow, all output generated using Veo and Imagen will be embedded with a SynthID invisible watermark. We directly embed the SynthID digital watermark to help identify AI-generated content.
Google One Pro users' generated videos will have an additional visible watermark indicating that the content was created by Veo using AI, making it easier for others to identify.
Google One Ultra users' generated videos will not have this visible watermark.
Will Google One Ultra users also have SynthID embedded?
No wonder all the videos uploaded to YouTube have zero views! Is it because you've all added watermarks and are preventing them from getting any traffic?

[worldly willow]

NBP and flow is not working properly

[iron burrow]

Video Generation Prompt: "Close-up shot of a hand on a mouse frantically copying and pasting text between windows on a messy desktop, screen glare, fast-paced, tech-dystopian aesthetic, glitch transition at the end."
· Voice-over: "If you’re still copying and pasting text in 2026, you are wasting your time—and your money."

Video Generation Prompt: "Close-up shot of a hand on a mouse frantically copying and pasting text between windows on a messy desktop, screen glare, fast-paced, tech-dystopian aesthetic, glitch transition at the end."
· Voice-over: "If you’re still copying and pasting text in 2026, you are wasting your time—and your money."

[pallid violet]

@visual karmaHello, as a friendly reminder we cannot generate images or videos in the server! Please do not post your prompts in the channels as it will not generate. If asking for advice or tips on your prompts do so in the appropriate channels.

[wild snow]

woow i like it men

hello guys i wanna know how i can upgrade flow for more credits

[tough yacht]

Hello! If you tap/click your profile picture in the top right, you'll see the button to get more AI credits

[wild snow]

Yeah but I don't See How much credits I Get

[tough yacht]

The videos you create will not end up on Flow TV; it's a curated selection and all content has been approved by its creator.

[wild snow]

i don"t see how much credit i get

@tough yacht

[tough yacht]

Plus is 200, Pro is 1,000, Ultra is 25,000

[wild snow]

oky 1 000 And I can get Credits daily ??How much credits I get If I get pro

??

[tough yacht]

Pro is 1,000 credits monthly

[wild snow]

yeah Understand But I get Daily Credits Right

[tough yacht]

Just monthly - you get them all at once at the beginning of your billing cycle. You can also buy top-up credits as needed.

[wild snow]

AHE oky

but also you get 10 credits Daily Right ??

[wild snow]

@tough yacht

[tough yacht]

Only the monthly credits

[wild snow]

@tough yacht hey Can I Ask you

[foggy radish]

Hello with the cheapest AI subscription on google, I can use flow without watermark?

[halcyon drum]

What's wrong with veo? It is not generating any video with any dialogues. I get error, "the content goes against policies This generation might violate our policies. Please try a different prompt or send feedback"

[sick sparrow]

downloading 2k / 4k images via flow is throwing an error (ultra plan)

[crystal talon]

Whats goin on with nbp?

Is everything forbidden now?

[edgy holly]

...scroll down.

[edgy holly]

Monthly, that apply to Whisk and Flow, plus some daily gens in Gemini and Vids I think.

[edgy holly]

Monthly. 1000.

[edgy holly]

You seem to be making numbers up out of thin air.

[edgy holly]

No.

[edgy holly]

"the content goes against policies This generation might violate our policies. Please try a different prompt or send feedback"

[crystal talon]

If this goes on veo3 will be unusable

[unreal viper]

please see our #rules-and-guidelines・📋 for more information on advertising.

[sick sparrow]

??

[edgy holly]

Try tech support, maybe?

[sick sparrow]

is it working for you?

[edgy holly]

One sec...

[edgy holly]

Nope. Broke for me as well.

[sick sparrow]

🙁

[edgy holly]

Mods are very punch drunk today. I'll DM you with a workaround option that should work if 2K is good enough.

[honest pumice]

have the same issue...

[edgy holly]

DM me if 2K is good enough.

[honest pumice]

using other service for that (the same as you I guess 😉)
but thanks you for offer

[edgy holly]

Same service. Different app. Not an offer, just trying to avoid trolls.

[halcyon drum]

woman talking to camera: Welcome to the podcast, We will be talking about the real estate
This violates community guidline WTH

[edgy holly]

...so you know what you need to do.

[halcyon drum]

what?

[edgy holly]

To get what you want, it says right there what you need to do ("try a different prompt").

[halcyon drum]

I tried alot of different prompt. How does that violate any policy? or any other prompt? or dialgue?

[dry frost]

Keep trying until you get a result

[edgy holly]

You're acting like this is a human being, following clear rules. It's a bot. They do a lot of weird things. All you can do is change words around until it works.

[onyx eagle]

I can't download 2k or 4k image from flow rn. Is there any maintanance?

[dapper edge]

Hi everyone! Does anyone know why subtitles appear when using Veo3 fast?

[edgy holly]

Usually doesn't. Seems to happen more often when you put quotes around it, but it's hit or miss.

[iron sleet]

I can't even download 1080p! Are you serious?! Lower the price, this is becoming a dump.

[dapper edge]

okay im gonna try without quotes thanks

[edgy holly]

Quality is usually better on that issue.

[dapper edge]

yes it works thanks !

[fading dirge]

bruh yall got to do something about veo wanting to add music to clips man, sometimes i have to geterate something 20 times before i finally get no music

slows down what im trying to do by so much time just trying to get it to not have music

[edgy holly]

I get this problem a LOT. I usually just have to delete all of the sound entirely and replace it. Not worth burning generations to try and get it to listen.

[fading dirge]

its so freaking annoying. can almost be a deal breaker for me

[edgy holly]

It's pretty new still. None of the AI video models do it well. Usually better to cut it and replace it.

[wary sparrow]

Thank you guys for the updates! That is all!

[fading dirge]

anyone sle cant download 2k and above images in flow? only letting me download 1k, when i try higher, it just says somethign went wrong

[tepid sail]

I am getting the same Error

[frozen totem]

2k/4k error for me

[cinder saffron]

Me too!!

[swift lynx]

Yep can’t download 4k or 2k

[gaunt gazelle]

Me too

[unreal viper]

please use #1238621511401734254 to provide feedback to the team!

these threads are actively monitored by the team!

[edgy holly]

I shared a workaround tagging you folks above asking about this, but the mods deleted it for some reason. There's an easy workaround if 2K is good enough for you. DM me.

[iron burrow]

Video Generation Prompt: "Split-screen animation of ChatGPT and Gemini interfaces, a question being typed happily, clean and bright modern office background, optimistic tone, motion graphics."

[kind bison]

This doesn't work all of the time, but try adding to your prompt... Music/Sound Effects: Ambient sound and dialog only

[brittle rapids]

“I want to buy a Flow account. I live in Bangladesh. Can anyone help me?”

[agile geode]

Just creating a video with a person's name, even if I used the name Laozi, a fictional Chinese deity, would still get me penalized by the policy. This policy is so stupid :))))

[viral folio]

Hey Guys, I started a Yt channel where I post songs, but I'm being very inconsistant because of the time taking for generating visuals, if anyone interested in collaborating with me for this project, lets make a partnership deal and work together, ping me if you are interested. Thanks🤝

[edgy holly]

That's odd. I haven't had a problem with historical people.

[gritty badger]

Oh yeah, that is definitely my experience I have been doing recreations of biblical stories and have generated unique likenesses for all these characters so they’re not getting prominent people warnings but I have to describe them completely generally in the prompt not using their names or it gives me the prominent person warning. Sometimes I get busted just because of what they say! I agree, the prominent people filter is one of the most broken and frustrating features right now.

[agile geode]

I completely agree. I can use Banana Pro to create anyone, any character from Caesar, Otavian, Leopard Tra... etc., but just using that image and character name will trigger a policy flag. Even if you remove the character name, you have to try several times with that prompt before you can successfully create it. The policy rules are the most annoying thing.

[pine panther]

Subject: Need help with prompts for AI Interior Design/Room Transformation videos.

I'm really fascinated by the 'AI Room Transformation' trend (like the videos showing a small space being optimized with smart furniture and creative layouts).

I’m trying to achieve this specific look: a clean, modern aesthetic with smooth transitions and 'smart' storage solutions.

Does anyone have experience with the best prompts for this? I'm specifically looking for:
1- Style prompts to get that realistic 3D render look (Lumen/Raytracing style).
2- Structural prompts to maintain consistency between the 'empty' room and the 'furnished' version.
3- Which Video AI tools do you recommend for the smoothest furniture 'pop-in' effects? (Runway Gen-2, Luma Dream Machine, or others?).

Thanks a lot for your help! 🙏

[frozen sand]

I am in Vietnam. I want to buy or borrow temporarily to use VEO3. Can anyone help me?

[pulsar quarry]

Hey! Regarding getting Gemini Plus/Pro while you're in Vietnam:

Check Availability: First, go to the official Google Gemini or Google One site. In Vietnam, direct access to 'Pro' features sometimes depends on your Google Workspace or Google One region settings.

Payment Method: Make sure your Google Play/Pay account is linked to an international credit card or a local payment method accepted by Google in SEA.

VPN (If needed): If the 'Upgrade' button isn't showing up, the feature might be rolling out slowly in your region. Using a VPN set to the US or UK usually unlocks the subscription option immediately.

Google One AI Premium: The easiest way to get it is by subscribing to the Google One AI Premium plan. This covers the Gemini Advanced/Pro features across all your devices.

Let me know if you need more specific steps!

[frozen sand]

It's very complicated. Can I use an account registered in another region? I really want to try it out.

[cinder saffron]

Yesterday I was having trouble downloading the 2K/4K image, and today, in addition to the problem persisting, I can't even generate the image. Is anyone else having this problem?

[pulsar quarry]

Yes, you definitely can, but here is the catch:

The Account: You can use a Google account registered in another region (like the US or UK). However, Google checks your IP address and your Payment Method region.

The Fix: * VPN: You must use a VPN set to the region of that account (e.g., US) whenever you try to access or subscribe to Gemini Pro.

Gift Cards: If you don't have a US credit card, buy a Google Play US Gift Card online. Redeem it on that US account to add balance, and use that balance to pay for the subscription.

The Risk: Sometimes Google blocks 'foreign' payments on accounts if the phone number used for 2FA is from a different country.

Pro-tip: It's often easier to just create a fresh 'Burner' Google Account while connected to a US VPN, and then use the Gift Card method. That way, there’s no regional conflict with your main account.

[winged plinth]

The video you attached is a classic example of 3d animation [likely Blender] rather than pure gen ai, imo.
The spatial consistency of the "expanding" walls and the precise measurement of the storage units is difficult for pure ai to do without hallucinations.

[pine panther]

thank you very much, it’s not easy with blender, he will take a lot of time to make 1 video, what do you suggest my friend

[winged plinth]

But, you can utilize a 1st & last frame technique to anchor the start and end points of the generation. This ensures the model follows a specific path.

[pine panther]

okey, is there any specific prompts for that, I am a new beginner

[winged plinth]

I'd suggest this workflow for you:
Generate the Full Room image first, make it your last frame.
Upload it to nano banana, ask it to erase the bed, the desk, and the shelves. Leave the floor, walls, and window untouched, and make it your first frame.
and enter this prompt:
A satisfying architectural timelapse of a tiny room being furnished.- **Action:** Galvanized square steel beams rapidly assemble to form a bed frame. Wood veneer panels slide in to cover the steel. The mattress pops onto the bed. The desk unfolds from the wall.- **Motion:** Fast-paced, snappy "pop-in" animation style.- **Camera:** Completely static camera (locked angle). No zooming, no panning.- **Atmosphere:** Smooth, viral home design aesthetic.

[pine panther]

I can generate this with google Flow ?

[winged plinth]

Yes, this is all flow in action.

[pine panther]

okay I will try and I will show you the results, thank you I appreciate your gesture my friend

[winged plinth]

You're very welcome! Happy to help. Can't wait to see what you create with it.

[tough yacht]

Hello! Flow is available in Bangladesh; what problems are you having?

[tough yacht]

Is this still happening to you?

[tough yacht]

I'm not able to reproduce this; what prompt are you trying, with which orientation, and which model?

[cursive thistle]

Hi where do you guys make yours scripts

[pine panther]

thank you

[shell heron]

It keeps failing to generate all day… every single time I try today, it says “generation failed.” Because of that, I don’t think I even managed to generate 20 video clips today. Does that make any sense?

[edgy holly]

Notion

[regal raptor]

Anyone else getting "This generation might violate our policies. Please try a different prompt or send feedback" on everything regardless of prompt. I use frame to video and have my own character and it keeps giving me this, tried changing up character doing 20 dif things, this thing is super broken

[bright heron]

For some reason in Scenebuilder I am not given the option to use the final frame as a start image/ingredient for the next generation. My friend has an ultra account as well and he has this ability

[edgy holly]

Are you on mobile?

[bright heron]

Mac desktop

[edgy holly]

Try leaving Scenebuilder, then coming back in. (it's real buggy - sometimes this helps)

*especially when you have a bunch of clips in there - there seems to be some kind of memory leak, like there is in the Library

[bright heron]

ah that would make sense. I've been refreshing regularly since things will get choppy after awhile

I'll try switching browsers too - currently working in Safari

[edgy holly]

That can help, too, yeah. I can't for the life of me understand why they'd build it this way, but I have a whole browser (standard Chrome, with only a few extensions) I basically only use for Flow because it's so finnicky.

[bright heron]

FIXED: by switching to chrome haha

[edgy holly]

Sweet

[sonic crown]

i've gotten so many failed generations with ingredient to video, why is this? it works for me on ladnscape but doesnt for portrait burt ththing is i need portrait for a video.

[hoary stag]

is anyone else having issues with Veo today? literally EVERY generation has failed for me

[tardy ridge]

can someone tell me how this goddamn prompt is harmful content: {
"scene_description": "0-2s: Extreme Low Angle Close-Up looking up at sams face against the towering, pristine surface of the Obsidian Monolith and the sterile white background.r. \n\n2-7s: The Fury.sam screams at the machine WHAT DID YOU DO TO HER?! ... WHAT ARE YOU??. \n\n4-7s:s.",
"visual_style":; High Contrast Void; Gritty Physicality; Acoustic Resonance; Raw Emotion vs. Cold Machine.",
"camera": {
"angle": "Extreme Low Angle Close-Up (Looking up toward Monolith).",
"lens": "85mm Anamorphic (Shallow depth of field on face).",
"lighting": "Cold Rim Light from Monolith (#FFFFFF) vs. Deep Shadow."
},
"main_subject": {
"character": "sam .",
"action": "Screaming up at Monolith, reacting to sound wave.",
},
"audio_cue": "Sams Raw Scream ( NO MUSIC.",
}

this is insane cenosring

its taken me 2 hours to do 2 shots because of this, crazy @tough yacht

[tardy ridge]

this is happening to me, its crazy

[broken mirage]

what's the difference exactly between extend and jump to...

[analog radish]

Try again, I can't see the issue on my end

[shell heron]

ahahahahahahahahhhhhaaaaaaaaaaaahhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh insame failing!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

I can't work just! how many days so far... I just unsubscribed

I keep running into failures, so I can’t use my credits at all. I’m going to have a bunch left this month, and they don’t even roll over. This is seriously pissing me off.

[tardy ridge]

its mainly happening with ingredients, seems to be much more sensitive

[analog radish]

So than the issue is in the ingredients

[tardy ridge]

yes

[analog radish]

Take a look at the images

[tardy ridge]

images are benign, monoloth dude standing in front of a greenscreen

no gore, no expressions

and im working in 9:16 so thats different too

[edgy holly]

Extend flows from the end of the first clip smoothly, Jump to will show mostly the same things in the first clip, but from a different angle.

[slate olive]

All my generations of people that are supposed to be speaking end up singing why does this keep happening

[lost lodge]

bro why cant i download the ai i created
with it saying i dont got credit but i got over 45k

[sick sparrow]

are there any plans to upgrade the 200 limit per day for NBP 4k downloads?

[tough yacht]

How often are you running into that limit?

[tough yacht]

I'm so sorry to hear that. I'll try to reproduce - are you using any image ingredients?

[near vigil]

Hey
Did you find a fix for this?

I'm having the same issue.

Press upscale 4k - it says its downloaded. charged me - but there is no download in my files

@tough yacht Hey, any idea on the above?

[tough yacht]

I'm not able to reproduce this - it just downloads directly into my downloads folder. Would you mind DMing me the email you sign in with and we'll investigate what's happening here?

[toxic shale]

im getting violations for simple things like having someone with a handsome face? Has the policies for NBP changed and become super soft?

[near vigil]

thanks will do

done

[lost lodge]

bro fix your flow

ive been trying to generate for hours now

[tardy ridge]

Yes I can DM images if you like

[acoustic ocean]

hi

[tough yacht]

Yes please!

[lost lodge]

@tough yacht i dm u something

[trail flame]

@potent shell

Google doesn’t sell accounts for $15 — scammers do.

[wintry pine]

Don't make racist remarks on this server

[analog radish]

Do you think there have been changes to the prominent people filter? On my side, it has started flagging my generated characters, which I use regularly for non-prominent roles.

[wintry pine]

Noticed that myself too

[sick sparrow]

Only once. But it was kind of frustrating that I had no ability to upgrade or do something about it. There was also not some sort of timer telling me when I was able to use it again

[edgy holly]

Do you have accounts that run on regular users' context (regular computer, not-stock Chrome browser, with a regular user account) to run these tests on? It seems like a lot of the widespread complaints I see on here and social media (I have a column in Tweetdeck/X Pro watching what people say about Veo) are very common, but when you folks test them, you rarely seem to be able to replicate them. Proper troubleshooting requires replicating the context, not just running it in yours.

[cold veldt]

Hey
I’m trying to generate 9:16 videos from 16:9 photo references, but when I crop horizontal images to vertical, too much important information gets lost.

I’ve tried rotating the horizontal photo before uploading it, but the results are inconsistent.

Is there a better workflow or prompt-based approach to preserve the full scene?
Maybe padding, outpainting, camera reframing, or a specific prompting technique?

Any advice or best practices would be appreciated

[edgy holly]

Before you use those images for image to video, upload them into whatever image model you use (Nano Banana Pro is good at this, unless it's something artsy, then MidJourney probably) and outpaint or alter it so the things you want are in the frame.

[ivory zenith]

Also is it only me who experiences how laggy and chopped the ui of flow is?
It runs at 10fps max and it sucks to work there.

[sick sparrow]

hey guys, how would you guys go about making the bedroom in the bottom left corner come to life? I've also tried adding an arrow to a separate closed up shot of the room, but Nano Banana Pro is giving me weird and inconsitent results. Do any of you have tips for this?

[edgy holly]

Wait, the app runs at 10fps for you, or the videos?

The app is super buggy, but if video is playing that choppy, something is off.

[ivory zenith]

App and the videos both.
It's almost impossible to review the videos from the app.
I have to download em and view always.

[edgy holly]

The app is a mess, but something is probably conflicting with your browser if the videos won't play.

I have a seperate browser I use only for this. Just a stock Chrome install with no changed settings or extra extensions. Flow works much smoother in there for some reason.

[ivory zenith]

I shall try it out.
Thanks aton @edgy holly !

[edgy holly]

Happy to help.

[surreal ruin]

It's very bad, Banana Pro doesn't create image in 2k resolution natively inside Flow. It has option to upscale, but upscale is not same as native generation.

[tough yacht]

Totally fair point - we are indeed working on a way to upgrade beyond the free quota. We'll get that out asap, though it won't be in the very short term.

[tough yacht]

We do QA on a broader set of conditions than just the Flow team's personal set ups, but agreed - there is a lot we can't reproduce. I'll look into how we can widen those conditions, to your point.

[edgy holly]

There are tools that theoretically could serve the same purpose for testing websites, where you can select an array of browsers and screen resolutions. Can't say I recall ones where you can select extensions as well... but wouldn't surprise me if that was out there somewhere.

[spare lynx]

hi guys, I'm using flow to generate videos and I want to include it into my n8n workflow, does anyone know how to do it?

[edgy holly]

Flow is a consumer web app, not an API. To use Veo and/or Nano Banana in a third party service like that, you need to get set up on Vertex.

[sharp wadi]

I am using flow and each time i am trying to generate a videoc its telling me audio generation failed. is someone else facing this issue?

[warm gull]

GN. Theres a zero credit model available right

[frozen sand]

When will we in Vietnam be able to register and pay for a VEO3 account in the most convenient way without having to change VPN?

[pallid summit]

[frosty comet]

Geniunely why is nanobananapro much worse in flow much than in the gemini app? I thought it's just bad luck, but i tried 100+ same input for both, and the results in flow has been so bad compared to the gemini app!

[edgy holly]

Different prompt interpreters. I can usually get what I want in Flow, but it when it's not working is usually does in Gemini. Not sure why they do it that way either.

[frosty comet]

It's not just " a bit better", it's miles better

Feels like an older worse model

[unborn viper]

Why frame to video stuck at 99%

[edgy holly]

Seems to ebb and flow in quality. Maybe 1 in 10 generations in Flow just gives me the image I gave it back to me without listening to the prompt. Very strange sort of failure I don't see with any other model.

(just happened now)

[rough halo]

Hello. Has anyone lost a big chunk of their uploaded images on flow suddenly?

they just disappeared

[edgy holly]

There seems to be some siloing with the image stuff in Flow, where an inage will be in the area you make it, but when you leave that window and go back to the main stream inside a projwct, its gone.

[proud dagger]

https://youtube.com/shorts/pRHeed_wg4Y?feature=share .. new video for pippit ai videos.

[tough yacht]

Can you give me a few examples that are performing better in GeminiApp than Flow?

[half cypress]

Have they removed the option to top-up credits on Flow? Using from India

[tough yacht]

Hello! Top-up credits are still available for Pro and Ultra subscribers. They are not available for Plus subscribers yet (Plus subscribers just need to upgrade to Pro).

[tardy ridge]

@tough yacht Curios ive youve been able to peek at the prompt / image examples triggering flags for ingredients

[tough yacht]

Yes, I was able to reproduce the same issue. The filter is overblocking. I filed a bug.

[tardy ridge]

thank you

i sent a very silly question over 🥺 lol

[tough yacht]

🙂 I'll respond in DM!

[winged plinth]

Hi @crimson pine, please head over to the website to start generating your video.
https://labs.google/fx/tools/flow

[full flicker]

Hi videos aren't created in the discord server. In order to make a video you'll need to go to https://labs.google/ pick the tool you want to work with then go from there. If you need help with prompts please ask in the appropriate channels such as #flow-tips-n-tricks・✏️ for advice. Thank you!

[stuck adder]

Kinda sucks that we can’t try genie 3 lol

[near gate]

Hmmm....all my videos I made on flow disappeared, how do I fix this?

Its loading the images but not the videos

[edgy holly]

Try hard reset, or different browser

[near gate]

I tried a different browser and still nothing. All my other project shows up but just the one project I need didnt. The scene builder shows some video but not all the video I was making because i didnt add them to scene builder yet. How do I do a hard reset?

[edgy holly]

Ctrl-F5

(clears cache)

[near gate]

🥹 doesn't work....rip, cleared cache, cookie, history, signed out

[crisp basalt]

You guys seem to have forgotten to redirect users to the actual pages of the respective Google Labs tools when they were browsing the ProjectGenie page

When a user is on the ProjectGenie page and clicks My Library, it leads to a 404 error if the user doesn’t have access to ProjectGenie tools (for example, because access is limited to Ultra users or the US region).

On that same 404 page, the user can still use the top-left navigation to switch to other tools (such as Flow or ImageFX), and those pages load correctly. However, once on the Flow page, clicking My Library again still results in a 404 instead of opening the library for that tool-

I only noticed this after getting confused about why image generation in Flow wasn’t working 😅. When I checked the link, it turned out i was still using the projectgenie page

Looks like removing /projectgenie/ from the link should fix the image failed to generate issue 🙂

[near gate]

Hello

[edgy holly]

"Hmmm....all my videos I made on flow disappeared, how do I fix this?"

[near gate]

So images are still there, scene builder videos are there but no videos when I go back to video. Tried clearing cache, different browser. Signing in and out. It is just this project. My other projects are working fine.

[analog lava]

Hi @near gate to report any product issues please use our #1238621511401734254 channel. When creating your thread please post as much as possible so that the team can properly investigate.

Thank you ☺️

[midnight cedar]

hey all, quick pulse check, do you all have any feedback to share around the Veo3 1080 or 4k upscalers?

[edgy holly]

Naner Banananer, or Vayo?

[midnight cedar]

Veo

[edgy holly]

4K looks good, especially for how cheap it is. 1080p is bad. I swear it was better when it first launched...

[edgy holly]

Testudo dos
Yeah, this current 1080p upscaler adds a lot of visual noise. Top one is 1080p. Bottom 720 original.

Looks like it just does some kind of bad edge detect and crams pixels. Again, 4K did great.

[edgy holly]

Very noticeable in this one, in the sky. The 720 gen has that crosshatching pattern that has, on and off, been an issue since Veo 2, but the 1080p upscale makes the sky look like it's got a bunch of cracks in it, while the 4K one smooths the issue out fairly well (I'll probably soften it up a bit in Davinci).

[midnight cedar]

thanks for sharing these!

[edgy holly]

Can do more, but the basic pattern is clear. 1080p no bueno. I know I've seen issues with 4K before, but these were good, and for only 50 credits its a steal.

At the moment the main thing for me personally is I need this to work with Scenebuilder, upscaling strings of clips together, if I'm going to use it for any of my projects.

[zealous tusk]

Upscalling is not very good in flow

and the crosshatching patterns make a lot of images unusable, Ive noticed them a lot

I´ve posted images with those crosshatching problems here with no answer of what they are or if they can be avoided somehow

Its like an AI Moire

Its one of the reasons i downgraded from ultra

[edgy holly]

In images? Or do you mean visuals in Veo?

[zealous tusk]

I mean videos in veo 3.1

[edgy holly]

Okay, yeah. I see that sometimes too. Haven't in Nano Banana.

[zealous tusk]

no, it comes with veo 3.1 only, its a crooshatched pattern that show in the videos

[lost lodge]

anyone having problems generating images

[austere zodiac]

me

[lost lodge]

bro everyday this happens

[elfin oriole]

I have problem with video generation, it's been 2 weeks it's not working anymore

do any of you guys have the same problem as me?

[lost lodge]

paying for ultra and this is how its working

[elfin oriole]

I just paying for pro, my credit still full

[lost lodge]

when will they even fix the problem

[elfin oriole]

not sure, hopefully they are fix it immediately

[lost lodge]

it’s been going like this for days i’m not believing theyll fix it immediately

[elfin oriole]

I think they are doing some updates to the website, I seen some update on homescreen of flow website which is we can scroll down right now, before this, we cannot scroll down

[ripe yacht]

Hi @cunning mango, videos aren't generated in the Labs discord server. In order to make a video you'll need to go to https://labs.google/, pick the tool you want to work with, and go from there. If you need help with prompts please ask in the appropriate channels for advice. Thank you!

[tight ravine]

how to remove watermark?

[surreal ruin]

Again and again "Failed Generation". No upscaling or downloading of previous videos. Tired of this. This is happening because google server is being hacked by hackers.

[kind ivy]

Hi, does anyone know the FLOW limits in the Pro plan?

[tall mica]

whi this piece of crap let me generate anything for six hours and keeps saying “failed generation”?

[granite minnow]

Paid for Pro and extra credits and this shittt doesn't even work. I keep getting failed generation error

[tall mica]

why this peace of crap still doesnt work??

[winged plinth]

Mind sharing the inputs? [prompt/image]

[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

[uncut valley]

How do I access genie 3?

[full flicker]

For all things Genie related please visit the #project-genie・🌎 channel and check the pinned posts. A lot of the general questions are answered there.

[real oasis]

still facing 404

just after login

[hushed abyss]

Is it possible to upload an audio file and have it sync to the character? I need the character to speak.

[edgy holly]

Not yet. There are a bunch of AI "lip sync model" tools you can search for.

[cerulean creek]

What clip editor do ya'll use?

[autumn wigeon]

I can't upscale 720p to 4K when exporting; I get the error "quality cannot be improved." And I have the Ultra plan? Is this a temporary issue, or is something wrong?

[agile geode]

How soon will Veo be capable of producing this level of quality?

[edgy holly]

Around... 8-9 months ago?

[agile geode]

Do you mean Veo3 could already do this 8–9 months ago, or that it will take 8–9 months from now?

[dry frost]

It could do that 8-9 months ago.

[edgy holly]

8-9 months ago.

[turbid lintel]

grok imagine already beat veo

[wooden jackal]

H

[agile geode]

The first video was generated by Veo 3. The second video was made with another AI. They use the same source prompt and the same reference images, but Veo 3’s action looks less smooth. Is there any way to improve it?

[lost lodge]

flow having some problems again

when will flow be stable??

[lost lodge]

cant even generate peacefully for hours