flow-general・🌊

[sly cove]

Hey folks. I'm a first timer here. Watched the official how-to video and then tried my first Frames to Video creation using Veo3 Fast only to be met with "This feature isn't supported on Veo3 Fast - Switch to Veo 2 Fast." Same happens when I try Veo3 Quality. Is this expected behavior? Anything I might be doing wrong?

Seems surprising to be encountering this when support for Veo2 is ending soon.

[crisp basalt]

Portrait or landscape mode?

[sly cove]

Landscape

[crisp basalt]

Weird. Usually when it asks you to switch to Veo 2 Fast, it means you were using the start-to-end frame feature

since Veo 2 Fast’s the only one that still supports start-to-end frame and extend

they’re phasing out the veo 2 model soon btw

So expect veo 3 model only by next week

[sly cove]

That's what I was trying to do: first-frame/last-frame. But they demoed it in the how-to video so i didn't expect it would be limited to the older model

[crisp basalt]

Your mom

[crisp basalt]

You can refer to this page https://support.google.com/labs/answer/16352836?hl=en&ref_topic=16355775&sjid=162691888689150370-NC

[sly cove]

Helpful. Thanks. Counterintuitive that advanced features are limited to older models in many cases.

[crisp basalt]

The video you watched didn’t show or mention the model switch, so I get why you misunderstood it , it wasn’t that clear for me either @sly cove

[crisp basalt]

Yeah but thankfully they’re phasing out the older model (Veo 2 model) soon, so Flow will rely fully on Veo 3 for all kinds of video generation

Not sure when but

omar have u considered google it before you ask here

[dry frost]

Extra important given this is a Google service

[crisp basalt]

cuz first of all, Flow doesn’t have anything to do with “invitation codes.” the only ai video platform using those right now is Sora

[winter bear]

Ok thanks

[crisp basalt]

youre welcome

[keen quest]

hey bro can you help me with something whenever i am creating veo 3 videos, they are getting stuck at 99%

[calm sapphire]

Hi, in flow when I press on the "Generate" button nothing happens. I just purchased Pro. Please help. Thank you

[analog lava]

Hello unfortunately this is not a moderation issue and an issue that the Google Labs Staff is already aware of. Please be patient as they are working hard to address the numerous feedback and bug reports that have been submitted. However if you have something new please utilise using the #1319424187730755636 channel with as much detail as possible and they will do their best to get to it. It helps to keep the team organised and its also where all the technical issues are logged. As your post may get lost in this general channel so to ensure visibility please use that channel for all reports and feedback. Thank you!

[bleak flame]

anyone having problems with flow?

i never had at all, saw some people were, and then aparently my luck.. now all of a sudden I cant create videos

clicking create button just has no reaction

[untold vortex]

Hey there, when i use my flow lab and add photos in ingredients to video it will not let me to generate any video and it says this (this feature is not supported on portraits ) neither the photos and the video is not portraits i used my laptop and phon did not let me. ( i used other photos just to cheak it did not worked)

[calm sapphire]

i agree -it's very buggy

[calm sapphire]

i have the same problem

[bleak flame]

same

im 90% sure they used ai to make these websites

which was a terrible idea

[calm sapphire]

is there any way to get flow to generate voice?

[bleak flame]

ingredients to video doesnt work at all

use frames to video, just dont give a second frame

[vivid compass]

Ingredients work if you avail of a plan

My problem is how to download using my phone

[knotty gulch]

Prompt:
A terrified woman holding her child tightly hides behind a giant tree as a massive python slowly approaches them in a dense jungle. The woman looks up, eyes full of fear and tears, whispering desperately, “হে আল্লাহ, আপনি আমায় রক্ষা করুন!” (Oh Allah, please protect me!). The atmosphere is tense and cinematic, with soft sunlight filtering through the trees, highlighting the danger and emotion — ultra-realistic, dramatic lighting, high detail, 8K.

[kind bison]

So many "stuck on 99%" failed generations for me today - never had this problem to this extent before - would only occasionally happen . Today it's the majority of my generations.

[analog lava]

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, unless you are asking for advice or tips. Thank you!

[stuck adder]

Sora 2 is actually very good veo needs something quick

[pliant raft]

around 70-80% of my generations are locked at 99% then fail, i cant put the exact same prompt a second time and it will pass, i mean my productivity is so much affected that i cant relate on this tool anymore.

[kind bison]

VEO is way better than Sora 2, come on!

Sora doesn't even allow you to remove the watermark.

[frigid pier]

lets be honest here. overall sora 2 is much better.

super more realistic.

[mental mulch]

is the Veo 2 phase here yet

[stuck adder]

When it comes to realism sora 2 takes it, for professional cinematic work Veo carries it

[stuck adder]

Wave ai sora 2 water mark remover…easy lol

[kind bison]

Can't even upload images of photorealistic people on Sora, it's more censored the VEO, and even people who pay $200 a month are get the forced watermark. It does some types of actions better than VEO, but overall it's not a professional video platform.

[analog lava]

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, unless you are asking for advice or tips. Thank you!

[last thicket]

They did dirty marketing too, they are finally blocking all of the third party content.

[last thicket]

btw flow is stuck at 99% all of my generations and i have 5 simulatenous and can't submit more but they are stuck at 99%

[sterile flume]

Does Flow work for anyone? I keep getting a policy error no matter what I tell it.

[stuck adder]

You don’t neeed the code any more

Just sign up under a new email address and you’re in

Don’t use the one that prompted a code cause it triggered it but use a new email addresss

[jagged pelican]

Hello

[autumn night]

Hello how to improve audio? its sounds to robotic for elevenlabs the words are mispronounce

[spare portal]

Anyone can give me tips to get 60 secs video using flow? No matter how many ways I mention the time stamps of the video in the prompt, it's always giving me 8 sec of video

[analog lava]

Hello, a friendly reminder. As stated in #start-here・📍 , “Currently, our team can support engagement in English; the discussion in this server will largely remain in English for now.” Thank you.

[radiant bison]

hi guys, i have generated videos in projects, now most of projects are empty, some have thumbnail some dont, is there a waqy to acces them ?

[analog lava]

Hello if you're encountering errors and /or bugs or would like to leave a general feedback about Flow please don't hesitate in utilising the #1319424187730755636 channel with as much context as possible. It helps to keep the team organised and its also where all the technical issues are logged. Your post may get lost in this general channel so to ensure visibility please use that channel for all reports and feedbacks. Please be patient as the team works through the bugs and feedback requests

Thank you !!

[autumn vapor]

split the 60 second video prompts into multiple 8 second prompts and assable in scenebuilder

<@&1366831844502274180> when will we have ingredients to video and extend on veo3 ?

[analog lava]

Hello we can only help with moderation issues, please only ping us if there is an issue that goes against #rules-and-guidelines・📋. Thank you!

[autumn vapor]

aah my bad. is there anyone how can at least estimate?

[analog lava]

aah my bad. is there anyone how can at

[iron matrix]

is the problem still occuring?

Should be resolved now, Let us know if it is not!

[jagged pelican]

Hello mate

[last thicket]

I haven’t tested it yet, but this issue seemed to occur when using the image-to-video feature and the prompt gets rejected. It often retries up to three times. Basically, you use an image-to-video and a prompt, and if the prompt or image is rejected due to content guidelines, it retries around the 40-60% mark. I know this because I created a new project using only that prompt with the image, and when I refreshed the project page, I saw about three retries had piled up, you can easily see it monitoring your credits

Let's say 400 credits drop to 380 after submission, then return to 400 if rejected. Ideally, it should stop there, but instead, it keeps fluctuating. After a few seconds, it drops back to 380, and if the prompt gets rejected again, it returns to 400, getting stuck in this loop until you refresh or close the tab. (Example with veo 3 Fast - 20 credits used)

[valid mulch]

Hi there! I've been having a pretty big issue using 'Frames to Video' in order to add scenes to a video I'm generating, where the quality gets worse and worse with each one, making the image look more and more grainy. You can see how bad it is on the 3rd iteration in the images I attached. This happens with both VEO 3 Fast and Quality. I tried fixing this by using the 'Jump To' feature with Quality, but that resulted in the same degradation.
Does anyone know how I could fix this or if there even is a solution to the problem? Any help would be very much appreciated!🙏

[raw oak]

I have to say what is super annoying is when you generate videos think it's on FAST but it really is costing 100 credits a video. You need to remember to change the quality back to FAST. Does this also happen to other folks here or is it just me.

[potent swan]

you're definitely not alone, its a good reminder to double-check before starting a render

[spare portal]

Thank you for that idea

[raw oak]

yeah. i think it should revert back to the default of FAST. I mean once you are on a roll generating videos you just click and click

[crisp basalt]

that only happened during Flow’s day-one launch tho. most of us ran into that issue back then. they’ve already fixed it by adding a model tag indicator inside the prompt box and introducing the “last used model” tracker, which automatically keeps tabs on your video model usage in the background

[crisp basalt]

tho I’m not exactly sure how urs ended up defaulting to the Quality model , that definitely sounds like a bug. u should drop a report in #1319424187730755636 so the team can take a look

[coral elbow]

Experiencing a bug with 'Ingredients to Video':

I upload the recipes and add the prompt. It shows the generate button as white (i.e. ready to start).

But when I click on it, nothing happens.

Anybody know what's going on?

To add more context, this was with Veo 3 - Fast and Veo 3 - Quality.

Maybe it's not available for Veo 3 yet, but the UI should then be different so users don't get confused

[vital citrus]

It seems that the flow system is down, I can't generate videos

[coral elbow]

I'm able to generate on Veo-2. It's just the ingredients to video on Veo-3 that is not working for me

[raw oak]

It’s just defaults to the last setting. I think if it would go back to veo3 fast by default I would save a lot of credits 😅

[crisp basalt]

It’s not available yet

It’s not really a bug it was stated in faq page

[jovial sand]

99% issue notfixed yet? ;((

[analog lava]

Hello if you're encountering errors and /or bugs or would like to leave a general feedback about Flow please don't hesitate in utilising the #1319424187730755636 channel with as much context as possible. It helps to keep the team organised and it’s also where all the technical issues are logged. The problem you’re facing may have already been posted there so please also check the channel for updates.

We ask for your patience as the team works through the bugs and feedback requests

Thank you !!

[coral elbow]

Which aspect is not available? Just the ingredients to video on Veo-3?

Because I can generate text to video on Veo-3 right now.

Also can you link me to the FAQ page you are referencing?

[crisp basalt]

https://support.google.com/labs/answer/16352836?hl=en&ref_topic=16355775&sjid=162691888689150370-NC

[coral elbow]

Ahh ok. thank you

What are people doing as a workaround for now?

Generating in VEO-2 and then adding the audio over with like ElevenLabs or something?

[dire perch]

Hello,
I'm new I want to know why I can't generate things with veo 3? Veo 2 is working but not veo 3

[river delta]

Hey guys I get this error message but im already set to use landscape instead of portrait and whenver I click on switch to landscape it doesnt do anything.

[zenith cosmos]

Cód??

[spiral gyro]

hello, I recently subscribed to veo3, but I can’t find the start button in flow. Could you please let me know where it is

[wooden wolf]

can anyone give me some awesome prompt for flow to start as a beginnners?

[last thicket]

wrong server

[slate sphinx]

anyone know how to make that viral earth zoom in effect in veo 3

cuz that website is now charging money

[swift mantle]

Hi

[limber igloo]

hello everyone

[leaden mirage]

Hi guys, today flow has been rejecting my prompts for no reason. It says “this prompt may violate our guidleines”

And I’ve used this prompt before many times to generate plethora of other similar videos. Just today it keeps being stubborn and not letting me do my thing. How do I fix it?

[autumn night]

brother how did you get good quality audio from veo 3?

is there some trick?

[iron matrix]

o dang, this is a new bug I had not heard of. I'll flag to the team!!

[thick oriole]

Hi, sorry about that. Are you able to share the prompt you're using, or the specific error message you're seeing? We nearly always provide some detail in the error message on why a prompt is being blocked.

[leaden mirage]

Prompt: Three figures on the wall finish their food and suddenly step out from the wall and walk off from the view. Camera Stays still

Reason: it may violate guideliens.

Surprisingly I’ve used similar prompt before and similar frame to video style.

But today its always getting stuck at 99 percent. And i. The end returns supposed violation message

[granite sage]

Here too! It's a bug.

[main copper]

I confirm you need ULTRA

[timid niche]

Have people found a solution to make it work?

[granite sage]

Not yet. For me continues.

[ancient karma]

Anything I should adhere to when trying to create some TV advert style clips of people having a conversation? Currently struggling to get any lip syncing working or choose when the audio begins etc

[main saddle]

if you haven't already, would you mind sharing this in #1319424187730755636? want to ensure our crews can log + investigate! thank you 🙏

[kind bison]

Add timestamps to the action and dialog. You can say things like:
0-3 s | Static shot: the man looks at the woman. The man says (friendly voice): "can I help you?"
3-6 s | Static shot: the woman turns away from the man. The woman replies (rudely): "I already have a boyfriend."

[ancient karma]

This worked perfectly, thank you

[thick oriole]

Very sorry for this confusing error message. Veo 3 doesn't support Ingredients to Video, you need to choose Veo 2. We'll have a proper fix for that this week.

[thick oriole]

just replied above, this is a brief known issue we just figured out, sorry about that. tl;dr: Use Veo 2 for Ingredients to Video: https://support.google.com/labs/answer/16352836?hl=en&ref_topic=16355775&sjid=16194257996125180286-NA

and turns out we do have a bug thread for it, thank you: https://discord.com/channels/1172568727942860810/1319424187730755636/threads/1425158299585613984

[stuck adder]

Wonder what’s gonna be announced at the flow event Thursday 😮

[river delta]

But i was able to use ingredients to video with veo 3 quality

On veo 3 fast it didn't work, but with veo 3 quality it did it.

[late swallow]

How come , what region are you from

[near gate]

my flow keeps buffering and doesn't load my project, i can go to scenebuilder and see my videos that i attach there but on the main project page, nothing. any fix for this? i tried a different device and clearing my cookies and cache

[kind bison]

3.1 is dropping very soon

[stuck adder]

I see someone said it’s on the cloud internally?

What do you think the update it’ll bring? Think it’ll rival sora 2?

[kind bison]

I saw some details last week but I can't remember them or find the post on X. We will know soon enough.

[stuck adder]

After all it’s just a .1 update so idk

[turbid lintel]

veo 3.1 probably release this week

[dense onyx]

Sup anyone up for a quick project Collab

I m beginner btw

[elder walrus]

Just send in another prompt request. Your previous videos will show up.

[turbid lintel]

Sora 2 have great auto cut shot , and you can achieve the same thing on veo 3

the trick is within your prompt

[crisp basalt]

come to think of it, during the early-access phase of Flow (aka VideoFX), they rolled out the Veo 2.1 model , which actually turned out to be the “fast mode” version of Veo 2.0. now I’m curious what they’ll bring to the table with Veo 3.1.

[turbid lintel]

there is a leak of what coming in 3.1

someone post on X

[crisp basalt]

i saw it too, honestly, I’m not setting my expectations too high, but with Veo 2 being phased out, it’s clear they’re gearing up for a massive update.

[turbid lintel]

I am told "[Veo] 3.1 has a new base model with a stronger ability to generate rich audio and dialogue, deeper understanding of complex scenes and narrative, enhanced realism in outputs, and improvements in maintaining consistent characters across scenes"

from X

i just want consistent audio

so we dont need to use external tool

[crisp basalt]

the current veo 3/2 we have is already insanely powerful , it’s just that users really have to figure out for themselves how to get the exact video they want through constant iteration and prompt refinement. honestly, i’m kinda tired of having to keep tweaking over and over just to nail the specific style i’m aiming for 😅, when i saw some of sora videos, and i looked at the prompt they wrote, and its insanely short and simple but the video looks really good.

[turbid lintel]

Agree

Sora2 have great auto cut shot abilities , the model can cut a different shot and stick it like a pro video editor

Meanwhile in veo3 you need to do it manually

[crisp basalt]

hopefully when they mentioned “rich audio and dialogue,” it also means they’ve improved how dialogue gets assigned to each character cuz man, the number of times I’ve gotten a video that looks amazing but the dialogue audio just completely messes it up…

and hopefully the newer model will chill out with the random “background music” it keeps adding. like yeah, we can prompt “no music,” but honestly, itd be way better if the default was no music and we choose to add it when we actually want it

[turbid lintel]

Yup , i still use eleven lab for every shot.. just to get same voice

[turbid lintel]

Better if they implement no background music button

[crimson garden]

Is there any news about, when will be possible to generate longer videos? like couple of minute long?

[crisp basalt]

if that ever happens , thats gonna be expensive as hell lol

[stuck adder]

Ahhh make sense. Well it’s probably a good update that they need to release an increment update is it possible it’s to compete with sora??

[crisp basalt]

but yeah no theres no news about it even,

[stuck adder]

Otherwise they’d wait until Veo 4

[crisp basalt]

i’m pretty damn sure google does have a veo model capable of what sora 2 can do like, there’s no way they can’t compete. i mean, come on, veo’s training data is most likely built from youtube videos and similar sources… it’s hard to believe they don’t already have that level of capability

[crisp basalt]

i don’t know how they trained the sora 2 model or what dataset they used, but every video i’ve seen from it just feels so natural. like, “real” might be an overused word at this point, but still everything about sora 2’s output makes me look at veo 3 in a completely different way

most of those sora 2 videos i saw seemed like “brainrot” meme stuff or parodies of real-life stuff (like police bodycam or ring camera footage). so yeah, you can kinda tell who their target audience is. meanwhile, flow’s clearly positioned itself as an AI tool for filmmaking, which is a whole different focus

but yeah in conclusion,,,,,,

[stuck adder]

Lmao exactly lol

It better be

But knowing Google they don’t want to be behind the ball long so they might just freak us out

Crazy how Higgs field already have a page for Veo 3.1 lmao

[crisp basalt]

either they actually have it or they doing it for clickbait

wait let me see

omg

[crisp basalt]

https://higgsfield.ai/landing/veo3.1/showcase-multiple/sora-1.mp4 ???? the page showcase video of veo 3.1 but its actually not

sora-1.mp4

[stuck adder]

Im kinda skeptical

Based on the answers it gave lmao

[crisp basalt]

https://tenor.com/view/skeptical-gif-25366675

[stuck adder]

Exactly …BS

VEO 3.1 will probably have maybe longer seconds and HD and or better audio that’s about it I doubt it’ll be anything near sora 2 right now

[crisp basalt]

yeah right no

placeholder page ..

[stuck adder]

Lmao yeah place holder click bait

[prisma epoch]

Last work

[edgy mango]

Could someone tell me if Flow actually supports the Upload image feature? I've been facing an "Portrait is not supported" message, and when changing to landscape it says that "Landscape is not supported", vice versa

[coral elbow]

I was just charged 100+ credits, the video stayed stuck at 99% and then when I refreshed nothing is there.

What should I do?

[stuck adder]

Refresh bro 😎

[tacit aspen]

Whats interesting is I think there must be something going on under the hood because Veo 3 fast seams to be more cinematic. Maybe as a response to Sora?

[burnt oriole]

👀

thank god I didn't waste my month's credits on Flow yet

[tacit aspen]

They pushed a update this morning under release -- so maybe good things are coming

[keen radish]

veo 3.1 when

[stuck adder]

Did they? Should I try it out and see?

[tacit aspen]

Its a commit release -- so I don't think its 3.1 -- but I could tell the prompts were reacting a little differently?

One of the more expert people might have a better perspective on what is happening

[stuck adder]

Where did you see a committed released?

What did you feel was different more prompt adherence? Better quality?

What I did notice is a slower generation time for fast almost mirroring quality

[tacit aspen]

when i clicked on pro tab it shows credits, buy more credits and under privacy shows update

better quality

[leaden mirage]

Hello, my problem still persists. Need help please 😞 my credits will eventually expire if i wait out more

[stuck adder]

Wym?

[leaden mirage]

Prompt: “Oct 8, 2025
Characters on the wall drawing become animate and start moving in real life”

[stuck adder]

You’re on ultra?

[leaden mirage]

Im using pro. And I solved the issue thru trial and error. The problem is vertical video. It cant generate verticals for some reason. But horizontal videos gave zero problems.

[forest raptor]

Veo 3.1? Only if you get 30 free generations a day like with Sora 2.

[raw swan]

Veo 3.1 is coming tomorrow

[stuck adder]

We don’t know that

[kind bison]

Happened to me, too.

[kind bison]

Higgsfield already has it listed on the waitlist.

[stuck adder]

I noticed but it’s a placeholder without any info

What’d you think the update is?

[raw swan]

tomorrow

[patent slate]

Hi, Newbie here. I tried using Flow today but kept getting a messag esaying I didn't have access to the model for text to video. Is it blocked?

[blazing mulch]

yes. try not to create more Prons. jk

[tepid tundra]

Oh well, time to start prompting again

[spice moss]

Flow hasn't even released 3.1 yet. I don't believe the rumors are true. If 3.1 were truly released, they would have announced Flow's 3.1 announcement first in the channel announcements. Flow itself hasn't even announced the feature on its homepage. I don't believe a website that's a hodgepodge of AI features would release 3.1 before Flow. I think this website is just a rumor I saw somewhere, and the preview mode is a marketing ploy to entice people to buy.

I don't see any mention of 3.1 pre-use features in Flow.

[raw swan]

you need to relax man. it's coming sooner than you think

[frigid pier]

scammers

[white forge]

Veo 4 coming after Veo 3.1 on nearly

Veo 3.1 is an gate first update for 30s expand duration, but the quality still look same with Veo 3 quality. it's open track way for next Veo 4

it will blowing up social media again.

[hidden sage]

Veo3.1 is API only atm and would be released tomorrow

[white forge]

If you spamming luck to generate in flow, you will possibly to get 3.1 edition generate quality but with 8s.

[white forge]

If you want to know anything leaks about Google products, be friend with indian & taiwan geektech. they're love to leaks anything to act like an superhero 😂

[stuck adder]

where did you hear this?

[hidden sage]

Its already available in 3rd party tools cause its available in API

Its just a 30second long video nothing special

[stuck adder]

show me

extending to only 30 seconds and thats all? smtt

honest dunno what we will need 30seconds for tbh

[spice moss]

For any news, I only look at the official news. I don't trust any news that has not been officially released.

[crisp basalt]

nothing’s been officially announced by Google yet , I really doubt those third-party tools already have access to Veo 3.1

[kind bison]

I need 30 seconds for my work - even if I don't use all of that 30 seconds. Would be amazing. I hate being limited to only 8 seconds.

[crisp basalt]

people actually thought the description Higgsfield wrote for Veo 3.1 on the Veo 3.1 page was real oops

[hidden sage]

Didnt want to promote that tool so i have hidden it.

[hidden sage]

Its not higgsfield oops

[crisp basalt]

i didnt ask anything from you lol , dont worry

[hidden sage]

It was for community not specifically you lol

[crisp basalt]

?

[stuck adder]

Why you hide it tho lol

Show the full thing

It says 30+ seconds that means up to a minute? That’ll be overkill

[crisp basalt]

gatekeeping a 3rd party tool is kinda crazy

[hidden sage]

Isnt it against tos?

[stuck adder]

Can you pm it to me then?

[hidden sage]

Done!

@stuck adder what do you say?

[stuck adder]

Yeah! I saw it

[hidden sage]

Well… 🤷‍♂️

[crisp basalt]

??

what do you want from me lmao

[hidden sage]

Hahaha chill!

[crisp basalt]

Okay? lol

[ocean river]

Hello friends

[white forge]

It's funny when every flop start-up company, begin to waiting Veo upcoming model possibility when they know it's Big Google behind. only company that can challenge Google on technology is Baidu x Bytedance x Tencent.

When Google joined AI Races, it's just an new playgrounds to Google.

[near gate]

i mean its obvious sora 2 is better than veo 3

[brave aspen]

I’m surprise Sora 2 had more google images then veo 3 😭

[stuck adder]

In what way?

[kind bison]

Nope.

Sora 2 is more censored, video quality is lower, you cannot upload images of realistic people for animation - it's completely nerfed.

The Sora hype will fade within weeks.

[raw swan]

i think veo 3.1 will be a nice update especially with the longer clip generation

[keen radish]

where u getting this info

[raw swan]

https://tenor.com/view/crystal-ball-crowley-david-tennant-good-omens-psychic-gif-3436909361193405344

[keen radish]

beside wavespeedai fake info

[raw swan]

they're not fake

[faint wyvern]

https://cdn.discordapp.com/attachments/1308160514278424707/1425523431142719499/Screenshot_2025-10-08_193957.png?ex=68e88e66&is=68e73ce6&hm=ddf5b7fbe619b767a9f1e9c178089ba8f9ce7a45c6e6c3151abffd1181c78394&

[raw swan]

no

[raw swan]

nope

[keen radish]

that guy says it still 8 sec

[raw swan]

relax, you will find out about veo 3.1 soon

[keen radish]

here

[raw swan]

dude, chill lol

you won't believe me? ok, then expect another model release one month from now

i wont say what is it

[turbid lintel]

is this true?

[keen radish]

its not true. its fake info.

[turbid lintel]

video extended to 1 minute .. too good to be true lol

[keen radish]

ya. fake 1 min

[turbid lintel]

maybe 15 second for veo 3.1

what we need is consistent voice for every generation

if google can pull this , bye2 to elevenlab

no more external tool

[keen radish]

i mean sora 2 already had that with ip character. consistent character/ voice.

[turbid lintel]

yeah , sora2 really set new bar

[keen radish]

hopefully, google raises the bar

[turbid lintel]

agree

[turbid lintel]

sora2 multi shot is great

[arctic cradle]

My existing Plus model subscription was canceled. Is this normal?

[autumn night]

Frame to video - always error

[spring breach]

change this image to video

[fast fern]

Would I call generation video from frame via API? Or any program for partnership?

[raven idol]

Broken right now ? Can't generate any videos.

[last thicket]

i lost 100 credits due to gen stuck at 99

2 hours passed still nothing

[upbeat pawn]

So do I.
It's stuck at 99%

[clever hull]

same, lots of gens stuck at 99%

[upbeat pawn]

😢

Should I keep waiting?
Will it be made if I wait?

[last thicket]

i dont think so

if you refresh page its gone forever and so your credits

[sterile junco]

Working for me ive generated well over 200 videos today no problemo

[rocky matrix]

Wonder if it goes live on Flow at the same time as these and higgsfield 👀

[keen radish]

where they getting the info from?

[rocky matrix]

Spoken to multiple sources and it's legit ✌️

[keen radish]

sure...

its 60 seconds not 30 lol

[stuck adder]

To be fair they did say 30+ seconds

[keen radish]

dont trust wavespeed, vodoo, higgsfield. they do scummy marketing

[autumn night]

still cant generate?

[crisp basalt]

hard to believe higgsfield would do that kind of marketing

[rocky matrix]

No it's confirmed today or tomorrow it's coming

[crisp basalt]

im gonna lower my expecto so that i dont have to expect anything in the coming week

i look forward to veo 3 extend/start-end frame feature anyway

[keen radish]

on higgs

[crisp basalt]

aight so basically... we will be getting =

veo 3 ingredients to Video.
Veo 3 Camera Control
outpainting/inpaiting using veo 3 model

[keen radish]

this post was created by chatgpt. "clean plates, be gone", em dash.

[crisp basalt]

all of those are literally what the DeepMind Veo page showed it’s just that those features were exclusive to Veo 2 back then. but now they’re making them available for Veo 3 soon, and honestly, I’m hyped

[crisp basalt]

lol probably

[keen radish]

random linkedin post, dont believe

[crisp basalt]

but if we’re talking actual confirmation from the Flow team or Google staff, the “ingredients to video” feature for Veo 3 is confirmed to roll out in the next update

[keen radish]

yes

[crisp basalt]

Veo 3 Start-End Frame is also ready (based on my internal checking 🤓 )

[stuck adder]

lol wow

They have an event 4pm today so

[autumn night]

3 videos wasted no audio

let me cry

[stuck adder]

What’s the outpaint and in paint like video-video???

[crisp basalt]

luka trying to hype up the “motion path drawing” feature is kinda funny to me it’s basically just the same “annotate image” thing the Flow team and others have been doing for months.

[crisp basalt]

yeah its a veo 2 capabilities

u can only use those outpaint/inpaint through vertex ai tho

https://colab.research.google.com/github/GoogleCloudPlatform/generative-ai/blob/main/vision/getting-started/veo2_editing.ipynb

[stuck adder]

Oh wow

[frigid pier]

sounds like they are scammers

[stuck adder]

It could be a place holder

[brave aspen]

True cus I went on flow and it says “support ending soon.” Meaning most likely a new model is coming and they need space

[burnt oriole]

ohh vow Veo 3.1??

When will we have it Floww

I cant wait!

[snow lion]

yeah

[burnt oriole]

Do we have any information if the limits of veo 3.1 are gonna be increased than Veo 3? Like it's so limited even for Pro users while Sora 2 and Grok Imagine are free for everyone lol

[keen radish]

lying so common on twitter

[green jetty]

no way google has a discord

[last thicket]

hopefuly they will add veo3.1 to flow

i rather use it in flow, its always cheaper and i cant live withouth the extra cloud space google offers in their plan

[crisp basalt]

+1

[crisp basalt]

https://tenor.com/view/finished-gif-8488721354669789943

[crisp basalt]

you know its bad when the lead product for google ai studio had to say something regarding about this whole "veo 3.1" shenanigans lol

lmao he straight up admitted the “video” they showcased wasn’t even from Veo 3.1

[keen radish]

fake acc trying to bait engagement

[stuck adder]

I knew for a fact that trailer video was. It Veo 3.1 lmao

The event ain’t even happened yet and they’re suppose to have Veo 3.1? Lmao

[quick rapids]

nothing in flow frontend for veo_3_1 yet but I did go digging for veo_3_0 and it looks like we're using 12-step distilled models, interesting to note, probably why they can offer it unlimited
model also shows up as veo_3_0_t2v_fast_ultra

[keen radish]

can u check on google cloud console of veo 3.1?

[last thicket]

ohh nice to know

[quick rapids]

from what I've seen on reddit plenty of posts confirming it shows up in vertex cloud, I'm just watching for signs of a flow rollout

[last thicket]

didnt know they were using distilled version

its like using wan 2.2 with lightx2v lora

[last thicket]

do you use webtools

[crisp basalt]

the ones you see in the screenshot are the Veo models mainly used by the Whisk tool to generate or animate images.
(e.g VEO_3_0_I2V_12STEP , VEO_3_0_I2V_DISTILLED ETC... )

Flow uses different kind of name for the video model

this is how Flow model key looks like

[last thicket]

how do you reach that?

[crisp basalt]

devtool -> network log

u should be able to see one of those log-

oh i see georgia in the event stage 👁️

[last thicket]

yea saw them

thanks!

so whats the difference between veo 3 fast and quality?

Sometimes i get better results with fast

[crisp basalt]

fast = quicker generation time, but lower quality
quality = slower generation time, usually gives better video output, though it’s kinda hit or miss sometimes

in other words… the fast model can handle long prompts, but since it prioritizes speed, it doesn’t always fully process or utilize every detail or “word” in the prompt

[last thicket]

Ohhh so its like uses a short prompt or like it modifies internally your prompt to fit in veo 3 fast?

[crisp basalt]

yeah i get good results mostly on fast too

[crisp basalt]

well technically yeah! , either way Veo 3 in general processes user prompts , short or long, and internally modifies (or rather, expands) them using Gemini LLM

[stray copper]

There example videos like Stable Video Diffusion 🥲 Waiting in Flow !

[last thicket]

So this is why sometimes it does stuff that i did not ask for it

[crisp basalt]

hence the so-called “Veo 3 prompt enhancement” or “prompt rewriter” it’s built right into the Veo 3 model itself, and you can’t really turn it off

[crisp basalt]

yeah, that’s why it’s recommended to write your prompt in a very detailed and descriptive way ,

but i also dont recommend that either LOL idk its just my personal experience so far

[last thicket]

Thats intereseting

[crisp basalt]

reason I said that is because when the prompt gets super detailed or descriptive, Veo 3 sometimes starts to, uh… “hallucinate,” I guess you could call it.

[last thicket]

Yeaa 100%

I usually submited prompts that were 500 words long lol

i had to ask gemini to use 500 characters instead

It was also a pain to modify since they were toooo long

[vernal kernel]

Watched the figure 3 promo video and got a dark sketch idea one could do with veo3 + nanobanana.

First, create the humanoid robot with nanobanana. Then do a few pictures of it doing chores around the house, then a picture of it by the pool. And with veo, let's have the robot notice someone is drowning in the pool. Might need to work a creative way to do it if veo thinks it's too dark, like with bubbles coming to the surface until they don't. Anyway, instead of helping, the robot just starts thinking, and we have a close-up to its face where it is updating its to-do list, like "Clean the pool, added to work queue," "Prepare dinner for three, replaced with Prepare dinner for two."

Someone feel free to do it. I am once again out of credits😂

[crisp basalt]

if you want to see how the veo 3 enhanced your prompt, you can try generate a video using gemini.google.com,

open devtool and network log.

generate a video

then wait

then youll see the "expanded prompt" in the log, which is quiet long

[last thicket]

thanks!

just found my prompt but no cluue the expandex one

[crisp basalt]

try look in response section

actually no, refresh the page, and then do ctrl+f and search "veo"

the expanded prompt should be in batchexecute

[last thicket]

alrighty

found it

thanks! it was all in a long phrase i wanst able to find it lol

[crisp basalt]

yeah its very lengthy, just use gemini or chatgpt to help u take out the prompt properly from those ,, lengthy text

but yeah thats how the expanded prompt looks like

[last thicket]

i gotta say the prompt is large

from my 10 words prompt it made a long long prompt

also using character A/B

[crisp basalt]

yeah im surprised how lengthy it expands , but when you read it carefully , the enhancer actually tried to create a proper "8 second* sequence description

[last thicket]

So thats the inside model (veo3)expander not just gemini being tasked to expand user prompt right

[crisp basalt]

yep

[last thicket]

Now i wonder if other video models do the same

[crisp basalt]

i been askin the same thing especially sora 2 lol

becuz the model seems to, interpret user short prompt very well

[last thicket]

Yess sora 2 must use gpt 5 or smt

but kling/seedance not sure

Anyways! thanks for the info!

[crisp basalt]

no prob i hope the flow team doesnt boot me out from this server or somethign cuz i reveal the prompt expanded thingy-

jk

[quick rapids]

well it's not veo 3.1 but I did find a veo 2.1 in the frontend o_O
and a veo 1.5???

[crisp basalt]

those are old model lol

[last thicket]

So at 22:00 they will announce 3.1?

[brittle minnow]

yo i heard urban legends of 60 sec veo outputs. urban legend? truth? truth soon, time unknown?

[stuck adder]

Where u heard that

[last thicket]

at the event

[keen radish]

source?

[last thicket]

scroll up in this discord server

[keen radish]

where specifically are u seeing that

[analog lava]

There will be a Flow AMA later today
https://discord.com/events/1172568727942860810/1425191284137787464

[keen radish]

so u think veo 3.1 will be released during a discord ama session?

[stuck adder]

😂😂😂

[keen radish]

ikr

[stuck adder]

At 22 is not 4pm it’s 10pm

[keen radish]

i think he is at europe. cet time

[stuck adder]

Ahhhh make sense I forget about time zones 💀💀

They better drop something tho lol

[last thicket]

i'd love to see veo 3.1 soon

i feel like veo 3 its already old lol

[keen radish]

logan kilpatrick (ai studio) even said its fake

[stuck adder]

Not really you see it’s still between 1st and 2nd place

[last thicket]

I only use veo 3 for text to vid

i always get slow motion in image2vid

[wind spade]

have you guys tried google vids yet?

[crisp basalt]

To be fair veo 3.1 leak was real, you can check the models name in google cloud quota log page

[keen radish]

can you link the page

[stuck adder]

No one ever used the text to video prompt tho lol

Image to video alwaysssss

[stuck adder]

Did you check it yourself? Or is this that screen shot from the Twitter guy

[keen radish]

is that screenshot from twitter guy??

[last thicket]

thats vertex ai

[dry frost]

Text to video is great for starting a project and also for testing out certain outputs

[crisp basalt]

I checked it myself too

[dry frost]

Image to video is best used for consistency

[crisp basalt]

Which I wish you guys would have the tendency to do it yourself instead of having to wait for someone to feed it for you

[keen radish]

ok i will check.

[stuck adder]

Well not everyone spends time going through logs out of the blue

[crisp basalt]

Well yet you had the time to “clown” people out just becuz your lazy butt didn’t bother to do it

all of this over an ai video model is crazy lol ,

[keen radish]

i dont see it on mine

[crisp basalt]

Do base_model

[keen radish]

still nothing

[crisp basalt]

Base_model:veo

[stuck adder]

I wasn’t clowning bro I was genuinely asking if you got it from the Twitter guy cause I saw it there first

My bad if you took it that way

[crisp basalt]

lol go cool yourself down or something

[stuck adder]

It’s hard to do that I live in the Bahamas 😂

[crisp basalt]

Cuz im goin to do that too 🛌

[crisp basalt]

Honestly image to video would be my last resort if text to video prompt didn’t works well for me even after iterating many times

[dry frost]

That's fair. I often edit my text prompts after seeing the results several times myself.

[crisp basalt]

Veo 3.1 is real but it hasn’t been released nor announced officially by Google yet. The Flow AMA event might be where they make the announcement, especially since the description mentions a “latest update.” but don’t quote me on that

[keen radish]

it will probably be some additional feature. base model will still be the same.

[stuck adder]

I see it @keen radish

[keen radish]

so, veo 3.1 is real. but it wont be a major update.

[last thicket]

30 to 60 seconds videos i think its a major update (unless its fake and only the elements thing, combining 3 images for the video)

but the name its bad, like nanobanana

i still cant take nanobanana seriously

[keen radish]

30 to 60 second in one prompt, i dont believe that fact with my eyes closed. i think the information of 30 to 60 sec is overblown and fake.

[last thicket]

Yeah its crazy

[stuck adder]

Yeah i don’t believe that either lol

[last thicket]

What would you have to write in a text to video so it wont stay like idle for 40seconds

[stuck adder]

What df u need 30-60sec for?

For one straight away take? For film makers?

[last thicket]

1000 credits for veo 3.1 fast

lol

[stuck adder]

I’d say it might be between 200-300

[last thicket]

Hopefully it will be released today xd

[kind bison]

A .1 release is rarely going to be a major overhaul. I wouldn't expect any massive improvements until v4.

[stuck adder]

Same

Until then we are stuck with heavily censored sora

[kind bison]

One of Google's devs already posted on X about these new features rumors being repeated by everyone - he simple said "not true".

[kind bison]

Why would I be stuck with Sora? It sucks and I stopped using it. VEO 3 is fine for my purposes.

[stuck adder]

Veo 3 still still in the led for sure

[last thicket]

its actually kling 2.5 now

[stuck adder]

How so tho? It ain’t even that solid

[last thicket]

dunno

pixverse 5 is so bad tbh i dont know why its up there

[keen radish]

veo3 is at 12.

[kind bison]

Really depends on what you use these models for. Kling and Hailuo are better with fast action sequences than VEO 3 - but VEO 3 is better with character interactions and, of course, has dialog while the others are silent. Kling has sound effects as an option after you generate, but they usually suck - VEO 3 is much better with sound.

[restive raptor]

Pixverse lastest update is pretty good, it's actually outperform kling and Veo 3 in certain task, each model has unique features and talents.

[silk hill]

every time i see a hype post on X about kling or hailuo or ray 3 they show one single shot action scene where the camera flies around all over the place and everything explodes randomly and people shoot in random directions at nothing in particular and the shot descends into garbled nonsense and the poaster says OMG cinema is OVER

[remote crypt]

@silk hill hi bud 😄 from the old days of IRC #3dsmax

[raw summit]

Hey guys, just a quick question first time in this chat, but is it normal that I cant upscale to 1080p in Flow even with Ultra? It's said it's included but in my plan but have no way of upscaling. Before I had an option, but now it's gone?

[silk hill]

ahah hi

[stuck adder]

well ntn was announced lol

[supple pendant]

awe shoot - JUST missed the event

[remote crypt]

You were nominated for an Oscar for your animated short "50% Grey", and I remember you telling me that you were there when Russel Crowe was nominated for Gladiator or something? Good times 🙂

I look forward to next iterations of Veo.

I just want to be able to "extend" with audio, hope that comes as soon as Veo 2 is phased out.

[silk hill]

i think he was there for A Beautiful Mind, Gladiator was a couple years before if i recall

[restive raptor]

Anything good happened in the event?

[remote crypt]

ah yeah it was A Beautiful Mind

[silk hill]

bummer to phase out veo 2, its still better than veo 3 for a lot of things

[restive raptor]

Any update on veo 3.1 yet 🤔

[remote crypt]

@restive raptor It was pretty great, I came in a bit late, but I hope they can do longer ones next time. 🙂

[remote crypt]

I hear you.

I haven't used Veo 2 that much, as I didn't have access to Flow/Veo until very recently (July-August). I'm in Norway. What things do you think Veo 2 does better?

I'm a union rep for my colleagues at work - so this touched my heart strings. 😄 Subversive and interesting and very well made! What is your stance on AI in general for artistic endeavours?

[kind bison]

Well, I don't care about - or even pay attention to - stuff like that. I only know from my own hands-on experience that there are certain types of shots (especially involing complex actions) that VEO 3 simply struggles with, but which I can easily do with Kling and Hailuo. Just the way it is - all models have strengths and weaknesses.

[remote crypt]

Of course, Veo doesn't do well with gore and such, so you did as best you could with dummies/bodies being flung from an epicenter of explosion.

[silk hill]

i think it makes some people with no talent very scared and other people with no talent think they have talent lol, i have mixed feelings but generally positive

[remote crypt]

I have made some things that Hailuo does better than Veo, and some things that Midjourney did better than all of them. I tend to experiment with them all. But recently been mostly Veo due to the audio.

[remote crypt]

right on

I don't know which I am - I just have a f-ed up imagination that needs an outlet, and not enough spare time to do it (like I used to with 3dsmax etc).

[kind bison]

Yeah - VEO is always my go-to for dialog shots, even with Sora 2 and WAN 2.5 now doing dialog. Of course, this is likely to change in the coming months as all the models start offering dialog - hopefully VEO 4 can keep the platform on top.

[silk hill]

i think its the closest thing to being able to bring whats in your imagination into reality with the fewest steps

[remote crypt]

and I LOVE that.

[silk hill]

i did a project recently where i tried all of these tools and kling succesfully did like, 1 shot out of 220, people swear by it, and i'm happy it worked for them its mostly failed more or less everything i tried to do with it, in the end i used mostly a mix of veo and wan (because you can train it so its good for character/vehicle consistency) and some seedance pro

[remote crypt]

kling has a lot of weird hallucinations

I've never touched it, but I've seen enough about it on youtube to make an opinion.

what is wan?

[silk hill]

wan is an open source video model that you can train

[remote crypt]

ah ok, if you have enough VRAM probably

[silk hill]

yeah so like i have 2 5090s and 2 4090s at home, and also rented a load of 96gb rtx pros on vast

[remote crypt]

I got 1 x 3080 Ti

[silk hill]

yeah gonna be rough to use it on that

[remote crypt]

ye

I was in the early beta of Stable Diffusion, august 2022

[last thicket]

Hey

[remote crypt]

hi Miwwa

[last thicket]

Do you know if its possible to use 2 gpus for 1 process

hello

there was a node called "raylight" but its barely any speedup

[remote crypt]

Tensor or CUDA related? Probably Tensor cores.

[silk hill]

not really atm there was one multi gpu extension in comfyui but not tested and i think it was mainly about offloading models to vram

unfortunatly multi gpu setups are not common enough for it to have taken hold as standard

[last thicket]

thats bad, because i'd like to do more steps rather than 2 videos simultaneously

ohhh i see

[remote crypt]

even back 20 years ago SLI (2 GPUs in tandem) was not as great as they made it out to be.

[last thicket]

Rn its nvlink right?

[remote crypt]

yea

and cross something

[dry frost]

Crossfire

[silk hill]

i mean they were in the rtx 1080 ti days but nvidia killed nvlink for the cheaper cards and then PSU power use went crazy so less of an option now

[last thicket]

NVlink is only available on data center gpus anyways

[silk hill]

so like a 5090 is 600 watts or whatever

and no nvlink

[last thicket]

:((

[silk hill]

so i have a machine that HAD 4x rtx 6000s in it nvlinked - and i sold them and it now has 2x 4090s currently but the second 4090 is mostly idle

[remote crypt]

The great "promise" of DX 12 was that we / developers would be able to exploit more resources more efficiently.. allowing for using more of every resource, but if not every part of the pipeline is "in on it" - they aren't gonna spend time and money on something that only affects like 1% of the user pool.

[silk hill]

but its handy in a sense that the first one drives the display so the second one is headless and is used for compute so driving the display doesnt eat any vram you need for gens

or slow down the machine working

[last thicket]

Thats great man... i used to rent the pro 6000 in vast

[remote crypt]

yeah it seems the high-end "promise" of the DX12 revolution that was promised never really came to pass. Focus has always been on the average mean, because that is where the money is.

I keep mentioning DX12, but any API really.

[last thicket]

I heard some people instead of purchasing expensive gpus buy ton of ram

To use the native models

[silk hill]

kinda need both

the models are all too big to fit in vram even on a 5090 and need to be scaled down to fp8 and use tricks like block swapping to offload to ram

[kind bison]

No 3.1, but at least some minor updates.

[stuck adder]

Meh…

[stuck adder]

Exactly

[fading dirge]

https://x.com/heysirio_ai/status/1976162949195588033 even the owner of enhancor AI did it. idk whats tha catch, whats the gian from doing this

Sirio Berati (@heysirio_ai)

I am playing with Google VEO 3.1. This is what to expect:

1. It still doesn’t surpass Sora 2. If Sora’s image-to-video feature wasn't restricted, there’d be no competition right now.
2. Lip-syncing and facial expressions are very accurate.
3. High-fidelity video with full 1080p

X · Oct 9, 2025, 5:48 AM

[stuck adder]

Now sora pro on wave speed is giving me a error smh

[kind bison]

AI X is being taken over by grifters. Reminds me of Crypto Twitter after the big bull run of 2017 - the lying grifters took over and ruined everything.

[fading dirge]

Some of the guys arent grifter though, like the owner of Enhancor AI, so idk what he has to gain from it

[crisp basalt]

It’s more crazy that Google themselves never actually announced anything about veo 3.1 yet these people seems so confident with talking about it lmao

[sudden condor]

Very random question will flow ever get a mobile app?

[crisp basalt]

The only thing that’s kinda “official” from Google would be the sighting of veo 3.1 model that appears in Google cloud quota log

[fading dirge]

yeha one guy on twitter saying 1080p 1 minute videos, and i commented you sure about 1 minute, hes like absolutely... just making things up, he doesnt know

[keen radish]

lots of bots on twitter

[opaque mist]

Yes, they anounced, resolution 1080 and duration of 30+ seconds

introducing Veo 3.1— the next-generation AI video model designed to turn simple prompts into cinematic experiences. Veo 3.1 doesn’t just generate visuals; it creates moving stories with synchronized audio, expressive motion, and seamless narrative flow. Whether starting from text-to-video or image-to-video prompts, Veo 3.1 ensures characters stay consistent across every frame, environments transition naturally, and camera angles shift with the precision of a real production studio.

Every Veo 3.1 generation comes with automatic synchronization — lip-sync, voiceovers, music, and effects are perfectly aligned in a single video. With extended support for clips up to ten seconds and multiple aspect ratios, Veo 3.1 goes beyond the eight-second limit of Google Veo 3 while remaining faster and more cost-efficient.

Veo3 achieves breathtaking results using a groundbreaking training method that drastically reduces costs. Join the waitlist and build the future of video.

@fading dirge@crisp basalt

[uneven drift]

How I can use it for free

[opaque mist]

Coming soon, not available right now

[keen radish]

source?

[stuck adder]

Source?

[crisp basalt]

https://tenor.com/view/well-oh-well-whatever-shrug-no-words-gif-17168293

[keen radish]

its fake info.

[kind bison]

Fake

[crisp basalt]

sorry @opaque mist when i said official announce from google , i mean from google themselves. not from 3rd party

so yeah higgsfield doesnt count,

[opaque mist]

Here is the source, remember that higgsfield is colaborating with Google

https://higgsfield.ai/veo3.1
@crisp basalt@keen radish

[crisp basalt]

no

[opaque mist]

Higgsfield is colab with Google, means that have that info sooner

[crisp basalt]

no

[opaque mist]

Yes

[crisp basalt]

https://tenor.com/view/laughing-lol-cracking-up-embarrassed-funny-gif-11455762

whatever floats ur boat dun dun

[keen radish]

they are callobrating how? all they do is use their api. any person can vibe code such sites.

[opaque mist]

https://cloud.google.com/transform/3-lessons-for-gen-ai-startups-higgsfield-ai-infrastructure-talent-models
https://cloud.google.com/blog/topics/startups/google-for-startups-program-fuels-more-ai-innovation

In short, Higgsfield's information about Veo 3.1 does not mean it has been officially released to the general public by Google. Instead, they likely gained access as a priority partner.

@crisp basalt@keen radish

[crisp basalt]

https://tenor.com/view/bugs-life-a-bugs-life-bruh-not-smiling-ants-not-smiling-gif-8815283262918161887

[keen radish]

you using a chatbot(prone to hallucination) to get ur info, rather than ur brain 🤡

[frigid pier]

all these veo 3.1 is fake news

get them out

[keen radish]

this fake?

[kind bison]

Partners have non-disclosure agreements and are embargoed from speaking about details until release date - if they break those agreements, they won't be partners anymore.

[restive raptor]

It's all fake, it's all about marketing and ads....3.1 is real but the real update are like consistency and first and 2nd frame with minor UI adjustments. (Forget 1min generation)

[keen radish]

where did u find this info "consistency and first and 2nd frame with minor UI adjustments"

[fading dirge]

@opaque mist https://x.com/koltregaskes/status/1976342575071322196 this is literally what higgs had until someone pointed it out then they changed it lol

Kol Tregaskes (@koltregaskes)

This is from Higgsfield's Veo 3.1 website (this will get me kicked off their CPP!).

Under what is Veo 3.1, they say:
"Veo 3.1 is the next major version of Open Sora text-to-video model,"

Under when it will be released they sa,y:
"Users can start generate Veo 3.1 starting on

X · Oct 9, 2025, 5:42 PM

[remote crypt]

This is ridiculous 😄

Does anyone know what Veo 3 Quality does better than Fast? I've so far not experienced it doing any better job than Fast. Yet it's 5x more expensive.

[fading dirge]

They literally copied what the had for sora 2 and changed it to veo 3 lol

[crisp basalt]

The audacity of some people here ,, they’re really convincing themselves that the whole Veo 3.1 × Higgsfield thing was real. Atp, I feel really bad for them

[crisp basalt]

fast = quicker generation time, but lower quality result (it really depends on.. your luck , cuz sometimes it will give a really good results for a fast model)
quality = slower generation time, usually gives better video output, though it’s kinda always hit or miss sometimes

[meager wolf]

Mine is stuck at 99% when generating, whats wrong?

[jagged trout]

anyone creating ads with flow?

[shell oyster]

I'm doing it, great results

[jagged trout]

check dm's

[remote crypt]

happens to me quite a lot too

[sterile junco]

Yep

[shell oyster]

Do you use prompt optimization or simple prompt? Since I set up my own optimized expanders I haven't had any problems.

[ashen harness]

Brothers when we will see portrait scenebuilder ??? Thanks

[iron sonnet]

me! Works pretty good

[jagged trout]

How is it going so far

[jagged trout]

What type?

[sterile junco]

Actually not bad, I'm trying out things that could work tho, since it's mostly slop that it generates

So it takes time

But I found ways around

[thick oriole]

https://discord.com/channels/1172568727942860810/1422203209065300019 covers this issue, we'll keep updates there (sorry! working on it)

[thick oriole]

We're working on it! No specific timeline to share but hopefully more like weeks, vs months

[thick oriole]

Hi all, re our latest announcement (#flow-announcements・📣 message) I just wanted to let you know that all those fixes were prioritized due to the feedback you're leaving here (and in the bugs & suggestions channels).

So, please keep it coming, we know it can be frustrating sometimes using a new & rapidly-evolving product, but we want to make it a great experience for you all, and when you let us know what needs improving, we'll listen and make it a priority.

I've been at Google for many years, and worked on some much larger products, and this is one of my favourite projects to date, largely because I get to actually just chat with the people using it, instead of just looking at dashboards and graphs. Though we have plenty of those too. Thank you!

[fast gust]

I have access to the API, but where do you see Veo 3.1?

[fringe horizon]

does someone has news about veo 3.1?

[rocky matrix]

looks like next week

[stuck adder]

There was none yesterday

[undone jolt]

Riddle me this. If Veo2 support is ending soon, why tf can I still not do an "extend" with Veo3

[kind bison]

Thanks, Ian.

[last thicket]

This was interesting

[quick rapids]

hey look someone else looking through the frontend. audit-brothers! lmao
I still want to know what veo 1.5 and 2.1 are

[stuck adder]

The earlier builds I think

[quick rapids]

where did you find that, anyway? capturing a response during a failed gen?
I've seen a ton of frontends where basic moderation layers are visible (like categories) or a flag is visible "output_moderation: true" or something.. but to find exact classifier thresholds is rare

[last thicket]

At gemini after a sucessfull generation, inside network look for batchexecution in response

[haughty oyster]

hello Is there a JSON prompt for brands?

@haughty oyster

[analog lava]

Hello, a friendly reminder. As stated in #start-here・📍 , “Currently, our team can support engagement in English; the discussion in this server will largely remain in English for now.” Thank you.

[hushed sandal]

Does anyone know if there is a problem with Veo 3 generating video with audio??? I can't generate anything with speech, it keeps giving me an error.

[orchid fossil]

HI Friends

Why Veo 3 IS EXPENSIVE?

Cuz i cant afford it ? Or i did not met with a person who sell it for 30 USD ?

[kind bison]

VEO 2.1 was one of the updates from March of this year.

[keen radish]

veo 3.1 leaked in google vids

[fading dirge]

Google vids?

[stuck adder]

Wym??

[crisp basalt]

https://tenor.com/view/patrick-stupid-stupid-town-hey-buddy-spongebob-gif-6574682199670721484

Is this you

[stuck adder]

I found it

[finite fiber]

Anyone have sora invite

[stuck adder]

You don’t need one now

[quick rapids]

they wont release it til the moment veo 2 is retired lol

[crisp basalt]

veo 3.1 video samples is out

[quick rapids]

where? I saw a couple on x that were clearly fake, didnt even bother to use nano banana for consistency just stitched 2 unrelated shots

[fading dirge]

Why does veo 2 fast still cost credits lol

[quick rapids]

they're punishing you for using it lol

[keen radish]

well veo 3.1, veo 3 will look the same.

[crisp basalt]

exactly 😅

[quick rapids]

I was hoping it would contain storyboard-style generations, it's the one thing sora turbo had that was an incredible concept but did not execute at all

[crisp basalt]

but yea take this video with grain of salt btw , this vid were generated using veo 3.1 using Google Vids website (it was internal config editing stuff ) (credit to testingcatalogue

[quick rapids]

still cutting at the 8s mark :\

[crisp basalt]

always has been , i think user can just use Extend functionality in Flow which in the end it will use Veo 3 model

[quick rapids]

I just mean out of all the features veo 3.1 could possibly have, I was hoping for veo 3 quality with slightly longer clips

[crisp basalt]

nah google will always sticks to 8 seconds

if anything user can always do extend with veo 3 in the future

[quick rapids]

true enough

[kind bison]

3.1 is not going to be a major update - gotta wait for v4 in December for that.

At best we can hope for one cool new feature plus some minor improvements. Happy to be proven wrong, though.

[crisp basalt]

i think veo 4 would be their starting point where they can do vid more than 8 seconds

im gonna assume veo 3.1 is nothing but,
-improved prompt adherence
-improved generation speed
-improved audio dialogue stuff

and thats it

[keen radish]

they will probably release ingredients to video for veo 3 alongside

[quick rapids]

sora 2's advantage with consistency is that it's running a storyboard behind the scenes, 10/15s long duration with multiple blocks containing a bunch of individual prompts that can introduce cuts and actions and so forth, that's what more models need. not just a single all-encompassing prompt but the ability to reprompt the model to do something specific at like 4s then 6s and end on a final concept at 8s

[crisp basalt]

sora still amazes me lol idk how they did it but i hope google can bring something better like sora

im talking about the feature rather than video generation itself

[quick rapids]

scene for scene I find veo 3 fast is still better quality, but sora's got some serious magic with the multiple cuts

[crisp basalt]

fun fact, before flow , video-fx used to have storyboard style feature,

[quick rapids]

lol I never made it past the wait list on video fx

[crisp basalt]

but im not sure what happened to that but i think they were just testing something

i was thankfully get to try video-fx back when Georgia openeed the registration for all discord peeps on this server

[quick rapids]

I'm still trying to figure out how to do storyboard type consistency but nano banana ignores my commands lol

I recognize a lot of latent power in flow and I'm trying to figure out how to actually wield all of the tools together, but image editing models confuse me after years using stable diffusion clip and flux t5xxl style prompting

more than extend I look forward to veo 3 ingredients if that isn't retired

sora turbo had a cool feature called style presets, kinda similar to flow's new prompt expander except instead of expanding your prompt it would let you upload videos and images and supply 1500char of directions on how to interpret your prompts. that was another big one. it was like insta-loras with text guidance

[crisp basalt]

wait when u said multiple cuts , you mean like you been trying to generate a video that features multiple cut of same character right?

[crisp basalt]

or same context of the scene but diff. angle etc

[kind bison]

You can do multiple cuts with VEO 3 - you just have to prompt for them - helps to give timestamps and directions for each shot

[quick rapids]

mostly cars, trying to keep the car consistent across cuts by using nano banana to swap angles and create new backdrops that look like they're from teh same rally course

[crisp basalt]

https://youtu.be/d-cGj3kAnsQ is it something like this? (thx to Dr. Wafae Bakkali )

Prompt: A person sitting in the same position but wearing different outfits, with sharp jump cuts between each outfit change. The background should stay static and the person should reappear instantly in the new outfit, creating a fast-paced, rhythmic jump cut effect. The lighting and framing should remain consistent to emphasize the sudden changes

i think this one is suitable for you, tho let me experiment it with car

[crisp basalt]

A high-energy, fast-paced rally video with sharp jump cuts between various dynamic shots of a single, consistent, bright red sports car. The camera is constantly moving, employing quick pans, tracking shots, and low-angle close-ups, capturing the car kicking up dirt and sliding around tight corners on a sun-drenched, rugged rally course. Each jump cut instantly swaps the camera angle and the immediate backdrop (e.g., from a wide shot showing mountains to a low-angle shot showing only gravel), while maintaining a cohesive visual style and the consistent appearance of the red sports car in full motion, emphasizing the rhythmic, adrenaline-pumping nature of the rally. The aesthetic is bright, high-contrast, and action-oriented, with motion blur used to accentuate speed. gonna try now , brb

[quick rapids]

my issue is more that you can describe the car model and color with perfect details and just generate t2v shots that seem plausibly from the same track run, but there are very minor details different with the car in each generation, so they can't be clipped together cuz it's obviously not the same vehicle. nano banana seems like a way to get different angles, taking individual frames and then continuing the shot with vehicle consistency using i2v. just having trouble figuring out how nano banana is prompted.

[crisp basalt]

actually nvm dont mind this , let me refine this

[crisp basalt]

hmmm do you have example backdrop and angle of the shot you wanted?

Veo 3 Fast
A dynamic, low-angle tracking shot follows a sleek, dark-blue rally car with white racing stripes as it speeds around a dusty, winding dirt course. The car, consistent across the scene, powerslides through a tight turn, kicking up a dramatic plume of brown dust and gravel. The camera quickly jump-cuts to a new angle, now a high-angle drone shot, showing the same car accelerating down a straight stretch of the course, flanked by tall, sun-drenched green pine trees. The aesthetic is high-contrast with deep shadows and vibrant colors, simulating the high-speed action of a professional rally race. A second, sharp jump-cut transitions the scene to a wide-angle shot from the side of the track, capturing the car briefly airborne over a small crest. The backdrop remains visually consistent with the first two cuts, featuring rugged terrain and the same pine forest, maintaining the illusion of a continuous rally stage with sudden, rhythmic changes in perspective.

[crisp basalt]

boy this looks like a game render lol

[quick rapids]

my prompts are super condensed, find it helps with consistency, the outputs look great, it's just making the vehicle consistent beyond 8s in additional cuts that's janky

Rear three-quarter low from the marshal’s side, 120 mm, handheld with a short recoil as the vehicle launches. A bright red Lancia Stratos explodes off the line. Forest pines flicker in hard noon sun. Palette warm earth with slight grain and gentle desaturation, rendered in hyper-real texture. 2.39:1 scope. Audio: violent gravel spray, intake howl through the roof scoop, echoing bark ricocheting between trunks, marshal yelling “Go!”

also veo 3 fast

its easy enough to get a plausible continuation but the car is always slightly different lol

[crisp basalt]

i think i kinda know the workaround..

[kind bison]

Create a grid of images - 3-4 images representing your different shots - and above them put a label that says "shot 1", "shot 2", etc, and add a timestamp in seconds after the shot name. Upload that in Frames to Video and in your prompt tell it to immediately jump/cut after frame one and to use the upload image as a reference for each shot. Describe each shot, with the timestamps, in the prompt.

You might need to make shot 1 slightly longer than the others to give veo time to remove the first frame image (it rarely follows instructions and makes it instantly go away - usually it stays for a second or so).

[raw swan]

@keen radish check your dm

[turbid lintel]

Veo3 recognise multi shot , but the generation is hit and miss..

Start your prompt with Multi Shot of blablabla

Then generate

Multi shot prompt work better if the prompt is short

Remember one action per prompt , more than one then the video going to be sloppy

[turbid lintel]

a racing car in the desert , then hardcut to cockpit insert shot , then to the driver face , then to the back side mirror

short and direct prompt .. far better than multi shot prompt

rally car race in the jungle , then hardcut to the drive face , then to the side mirro , then to top view

[last thicket]

Is only US site?

[crisp basalt]

Someone deadahh dm’d telling me about this wavespeed proof that they have 3.1 and im like ??? Like ok if you wanna get scammed then be my guest.

[fading dirge]

and the crazy thing is, they are charging people 3 dollars per run

[last thicket]

That's terrible expensive.

I will never get why use this kind of websites for one specific model its always more expensive. Unless you want to use other models

[fading dirge]

Well now that i see they are scaming, i will never use this website again. its not just expensive, its not veo 3.1 and they are charging that for whats more than likelt veo 3

[last thicket]

Anyone knows the difference between veo 3 and 3.1?

[snow lion]

is it worth it to buy ai ultra ?

[livid silo]

No, they throttle you

[turbid lintel]

Ultra give you free veo3 fast

[snow lion]

but i read this channel and other channel, everyone like very angry 😂 😂 .. free veo3 fast but always stuck at 99% and error

wanted to buy, but.....

[turbid lintel]

well for me free generation is worth it , just spam veo3 fast until you get what you want

totally free

[snow lion]

thank you

[turbid lintel]

4 x 5 = 20 generation

enjoy ultra bro , best plan for veo3

[turbid lintel]

how to apply?

i want to test

[autumn night]

how you make the dialogue clear?

[kindred osprey]

architecture is built around the principles essential for serial production:

Character and World Consistency (Continuity): The key challenge in a series is maintaining the appearance of characters and locations. For this, we have the Assets system. Using the /add_asset command, we create a "passport" for each character or important object. From that point on, the assistant will use this exact description in every shot, ensuring visual unity throughout the entire series.

Narrative Structure: A series is a story broken down into scenes and shots. The /sequence command is the primary tool for this. It allows you to describe an entire episode, which the [EDITOR] agent then methodically breaks down into a sequence of shots, working on each one consecutively.

State Management: The world of a series changes from episode to episode (and even from shot to shot). A character might get injured, or an object could break. The internal State Tracker monitors these changes. When you approve a shot (/approve), the system records the new state of the world and carries it over to the next shot.

Iterative Workflow: Working on a series requires constant adjustments. Our workflow is structured like a film set: we "shoot" a take, you (the director) review the result and provide feedback (/refine), we "reshoot," and this continues until you are satisfied and give the /approve command.

[crisp basalt]

https://tenor.com/view/thats-a-lot-of-text-lots-of-reading-too-much-reading-corl-corl-horl-gif-14051903

[fresh iris]

how to set language for tts

[sonic cedar]

uhh did veo 2 went from 10 credits to 20 credits on flow?

[agile flume]

@ashen basin Would it be possible to arrange access to Flow via Google Skill Booost?

I know it's not possible right now, but do you have anything planned?

please

[agile flume]

nice

[tall oar]

guys my audio failes all the time, is it also for u ? And i get requesting generations too quickly bug also... frustrating

[analog lava]

Hello, as a friendly reminder we cannot generate images in the server! Please do not post your prompts in the channels as it will not generate, unless you are asking for advice or tips

[upbeat arch]

Guys im having issues with flow, i hava google ai pro subscribtion but i cant use flow even though I live in belgium and I am over 18 years old

[near gate]

Google Veo 3.1 is soon 💚

[stuck adder]

Said everyone ever lol

[turbid lintel]

we just got mini update , probably next month lol

[keen radish]

so when veo 4?

[turbid lintel]

probably disember or 2026

[vivid nova]

Frame-to-video / Veo 3 Fast

[last thicket]

I still think veo 3 image to video feels slowmo

[zinc swallow]

Yeah even youtube comments tell me that

[zinc swallow]

🔥 wow this is so good. the fast motion is actually believable.

[agile flume]

Would it be possible to arrange access to Flow via Google Skill Booost?
I know it's not possible right now, but do you have anything planned?
please @ashen basin

[stuck adder]

In some cases yes, but most cases I think it depends on exact what the image is

That determines if it’ll feel natural or not

[quick rapids]

anyone else dealing with problems generating videos from the images made with flow?
every image generated has to be downloaded and reuploaded or it renders with a black bar covering the right half of the shot

[kind bison]

There was probably a thin line of black on the side of the image that VEO then expanded. Always download your images and open them in an image editor to check for stuff such as this before animating.

[quick rapids]

the images are fine, simply screencapping and reuploading fixes it

just can't generate -> save -> use as first frame without a black bar appearing. need to download asset, reupload, then use as first frame

size of black bar also changes between generations. no idea

if you compare those top/bottom the top image is shifted left so its like its pushing half the image off the screen

though nothing seems cut off the left side

[flint summit]

Seems like its been a long time since any useful new features were added to Veo. Would it kill them to give us extend in Veo3 fast? Or end frames? C'mon Google. The thrill is gone. -- Not that Sora is any better, it's so nerfed I don't even want to waste my time there.

[marble turtle]

what does this pertain to?

[turbid lintel]

Veo 3.1 soon , not big upgrade i bet .. at least we got something

[mental mulch]

Has Veo 3 got the Veo 2 stuff yet

[keen radish]

nah, google abandoned flow

[mental mulch]

https://tenor.com/view/cmpunk-wwe-yeah-gif-4928889

[grim palm]

Opal

[woeful beacon]

Has anyone experienced this? failed to upload image

[cinder geyser]

veo3 more realistic than sora but sora more accurate with prompt. i can't wait next update of veo

[deft burrow]

Hi I'm having problems with creating a video using the "ingredients to video" feature. I upload the text and 2 images but I when I press the create button it does not even start the video creation part.

Is anyone else having an issue similar to this?

[sterile junco]

eh?

[proven onyx]

Does anybody here having their video gets stuck at 99%?

[mortal storm]

did you try F5? reload the page?

[proven onyx]

Yeah. Sometimes it says failed. Other times it just falls into oblivion

[pure quail]

Maybe this is mundane, but I like the expanders in the TTV option. If I want a specific style I create an expander and enhance it. I then cut out the audio and text references in the prompt and use that to generate an image. I copy the whole prompt and then use it as a frame to video prompt. Most of you probably know this already

[proven onyx]

They are very good! I agree
👏

[woven escarp]

I've created a TV ad for a client but they have questions around copyright issues. The videos generated through Flow are references from the Google network but they don't want to fall into a legal trap of reproducing content that is linked to existing work/actors/locations. How can I put them at-ease? Maybe someone from the @tacit halo team can help?

[slow widget]

Hi guys! Do you know if there is any issue with the Flow page? I'm not being able to open the page to generate videos.
Is there any system outage? @tacit halo

[hot estuary]

This might be a newbie question. But for some reason everytime I go to make a new scene flow pops up with this error message. Can anyone help me figure this out?

[slow widget]

I saw the problem is maybe related to a cookies or a session issue.... I wasn't able to fix it but had to open Flow using Firefox. 🤦🏻‍♂️

[woven escarp]

Some guy sent a support ticket link to my query on here, but then he deleted the comment then the support chat disapeared. Is this normal or was it a phishing attempt?

[potent swan]

veo2 also not working?

[hot estuary]

veo 2 gives me very very weak results

[slow shore]

#flow-general・🌊

[stuck adder]

Start middle to end frame would be a solid feature

[gentle nebula]

does anyone know why i can't access to flow? says that is isn't available and I really need a google subscription? I'm more than 18 y.o. and residing in Spain, so I think i would be able to access flow..

[opaque sand]

Help me conquer the scene.
I'll simplify the prompt so you can understand my hatred for everything alive at the moment. It's been an hour of all kinds of prompts and all kinds of generation choices - fast and quality...
The train is going to the right.
The train is going to the left.
The landscape outside the window is moving from left to right.
The landscape outside the window is moving from right to left.

NOTHING helps to make this train move to the RIGHT — it always goes to the left... AAAAA

[stuck adder]

I can access it

[gentle nebula]

should the problem is the browser? i'm using edge rn

[quick rapids]

I use edge and flow works fine aside from a few ui quirks

[gentle nebula]

bruh so i dont know the issue

i will try later

[kind bison]

Try uploading a version of the image with arrows and instructions and tell it to immediately jump/cut after the first frame.

[vague wigeon]

Is the Flow video generation server down? I've been trying to generate a video for a while now and it keeps showing an error. It doesn't say what the error is, it just says there was an error generating the video.

[wind saffron]

with Ultra does V3 work with start and end frames in quality mode

[wind saffron]

This on the Higgsfield website

[keen radish]

fake info

[upbeat yarrow]

hi guys!

[stuck adder]

You can’t do start to end frame with Veo 3 yet

[upbeat yarrow]

I have an issue on Flow.!
I cannot generate videos.
It stops in 99%

[stuck adder]

F5

[upbeat yarrow]

I tried that for many times.
Inven in mobile it doesn´t work.

It worked just a little bit .

And then I cannot do anything

[upbeat yarrow]

May somebody help?

[drifting torrent]

is flow not working, i click the white send button but it does nothing when i click it

[drifting torrent]

<@&1366831844502274180> any help or ideas?

[unreal viper]

mods are only here for discord related help and to moderate the server. please ping the labs team or open a thread in #1319424187730755636 for flow related help

please only ping us if there is a moderation issue :)

[drifting torrent]

sorry 🙂

[grave orbit]

Is Flow the primary way to work with Veo right now, outside of pay-as-you-go access? Is aistudio.google.com an 'old' gui and Flow the new?

[crisp basalt]

Cinematic medium shot of a sophisticated silver-haired man and a blonde woman in an elegant, vintage train dining car. The man writes on a document as the train moves to the RIGHT, the stunning scenery of snow-capped mountains and a turquoise lake gliding past the window. Suddenly, a powerful earthquake strikes. The entire train car shakes violently, the camera trembles, objects on the table rattle, and the couple looks up in alarm, their serene moment shattered.

[hexed socket]

help why I am seing this. I only generated 1 text to video and this always comes up. its been 2 days now I cannot generate anything..

[analog lava]

Hello, as a friendly reminder we cannot generate images in the server! Please do not post your prompts in the channels as it will not generate, unless you are asking for advice or tips

[red finch]

Awesome

[oblique patio]

how can i make my flow video longer than8 sec when i input a script for 25 secs?

[upbeat yarrow]

I have got the same issue bro.

Flow is great but as a costumer I have no answers. I bought credits and I wanna buy more ... But I really fear it, because I have no idea if the team is working on it or not.

I just sent a print here, hopefully I have some answers.

[woven escarp]

Hey, this issue was causing me headaches last week. There's some filters at play here, if you have any reference in your prompt to a child/boy/girl, it won't make it through the filter. Be specific with the age of the person instead - Young male aged 6, for example. I was trying to build a family scene and it took me a long time to figure this out.

[upbeat yarrow]

Thanks mate. I am trying making an anchor journalist giving the crypto crash news.

I did 2 or 3 scenes and then it doesn't work anymore....

I described all characters of the journalist, I out several informations, but I am gonna try out the age .

Thanks for your support.

[jagged trout]

anyone charging clients to create AI ads for them??

[opaque sand]

Are they changing something in VEO-3 today? I haven't seen so many “subtitles” and sound output problems in a month as I have today...

[opaque sand]

Couldn't beat it. Redid the initial frame.
By the way. Try to make the REAR door of the sedan open—that's impossible too. It only draws the front door open.

[faint wyvern]

According to Google Gemini, Veo 3.1 is getting ready to roll out via API to partners. Why not for general consumers though? And why isn't this being announced on Flow Announcements?

[snow lion]

depend on your prompt i think...

[blazing turtle]

I can't generate images using reference characters... Does this happen to anyone else? I was working on a project and everything was fine until now.

[agile flume]

Use nano banana

[raw oak]

If I can make a suggestion- I am an AI creator and an ultra client...it's not cheap. When I signed up for the ultra plan the mention by Google was that we would get early access to the new stuff. In the words of @faint wyvern can we get some comms on the release. I keep seeing stuff from third parties and it would be nice to hear it from you. Thanks!

[opal jackal]

#flow-general・🌊 from yesterday i am trying to generate videos and it get's stuck at 99% most of the time is it me or everyone else having the same issue

[tribal elm]

SCAM

[dry frost]

Yeah... At least twice a week.

[tribal elm]

🫠

[kind bison]

What you are seeing is fake news from people using hallucinated replies from chatbots. 🤣 People should know better than to use chatbots as sources for reliable information (no less info on upcoming releases!), but clicks matter to them more than being accurate.

[raw oak]

There are some folks working in big companies announcing they are getting access to Veo 3.1 right now . I don’t think it’s very fair.

[kind bison]

Nonsense

There is a small group of people who have partnerships with Google and have been given details (or possibly advanced access) to 3.1 and all of them are staying quiet due to NDAs. Go ask Tim from Theoretically Media - he is part of that group.

[raw oak]

There are companies too that will integrate with their api. I’m just saying would be nice to have early access. The way it’s being done it’s a leaky ship.

[kind bison]

You are responding to people - and a few very shady platforms - spreading disinformation.

[raw oak]

These are folks in tier 1 companies. Very reliable. I think it’s the ‘buddy’ buzz strategy.

[kind bison]

LOL - they have literally been called out by Google employees for spreading disinfo. Please keep up.

See for example...

[raw oak]

That’s old it’s dates October 8. The question is will Ultra users get early access or not.

[kind bison]

LOL, nothing has changed since then - people are still repeating the same nonsense from last week as if it's news, and now Gemini has picked up on it and is regurgitating it - so the same people are saying "see, we told you so!" Even though Gemini is just repeating their BS posts.

[kind bison]

Gemini will even admit it's nonsense.

[proven onyx]

Is anybody here having trouble with flow? Seriously? I keep hitting rate limits after, like literally one generation. I have an ultra plan.

[quick rapids]

you can literally search the frontend, guy. they've added a ui element to gemini for "try now" on a veo 3.1 popup, not live yet, but that's as official as it gets

[keen radish]

gemini says its true

even gemini admit its true

[kind bison]

See my post above - when you remind Gemini that it is not using reliable sources, it apologizes and admits that this all comes from online rumors, not official sources.

[keen radish]

see here

confirmed

[kind bison]

Literally no one is saying that 3.1 isn't coming - only that the details about it are still unknown

[keen radish]

you inputted "not a single one of the sources" and "blatant falsehood", gemini 100% will be sycophantic towards u. u are directing it to say it is fake.

[kind bison]

There are a few people who don't work for Google who know, and all are under embargo not to release details

[crisp basalt]

Is no one gonna talk about the edit button or am I the only one

[keen radish]

here u go

[crisp basalt]

Oh we are busy discussing veo 3.1

[kind bison]

LOL - go ahead and check the sources that Gemini is giving for for the specs - it's literally all rumors and false posts from shady video generation sites

Gemini is just repeating erroneous info from the web - and now you guys are citing it like it's official, LOL! Gemini hallucinates and lies all the time.

[keen radish]

so u think this fake?

[crisp basalt]

It’s 2025 and people still relying on LLM for news ..

[crisp basalt]

They never said that’s fake, it’s the Gemini screenshot you posted that’s not true

[keen radish]

thats what i thought!

[crisp basalt]

Either way ..

Is no one gonna talk about edit button in Flow

[keen radish]

why u calling me a dumbass. chill down. 👍

[crisp basalt]

This thing

Oh we’re still discussing veo 3.1 alright cool

[keen radish]

thats not nice.

[crisp basalt]

Flow team added Inpaint feature 😭 thank you!!

Wait this is basically nano banana but for video

[tacit aspen]

Does it cost credits to use this feature? Very cool BTW

[crisp basalt]

Apparently it cost 20 credit

You can access it by clicking on edit button on any of our generated videos

[stuck adder]

lol calling names over Veo 3.1 is crazy

I did still did not justify tho

However that’s dead, I can’t wait for an update

😂😂😂 ok

[digital yoke]

Can anyone generating some thing with Swedish language and name kihlströms se if it still broken?

[crisp basalt]

Wait is this insertion feature uses veo 3 model only or can I change to veo 2

[stuck adder]

The the in paint feature? I think Veo 2 handles most of those type stuff

[crisp basalt]

Did you try it yet

Scroll up

[stuck adder]

Yeah not all that tbh

Wait hold

[crisp basalt]

Lmao

Too busy with hoping for veo 3.1 I see

[stuck adder]

Ohhhhh that’s a new feature 👀👀

Somewhat video to video in a way

[crisp basalt]

It is video to video frm the looks of it

[brisk wigeon]

why does Flow misspell words

[misty nova]

Is 150 words still good for the best results?

[quick rapids]

based on the built in expanders turning them into like 5 paragraphs idk

[misty nova]

hmm ill try to reprompt around 5 paragraphs and see if the results are better refined and focused

[quick rapids]

lol no I mean the built in "enhance prompt" button all makes it super verbose, and breaks it into categories and paragraphs, dont just aim for 5 paragraphs lol

[misty nova]

ill run my prompt through gpt again and let it expand pass 150 words and compare and contrast

[burnt oriole]

Logan said "veo" 👀

[raw swan]

Tomorrow is the official announcement of veo 3.1

[burnt oriole]

it's in 16 hours

[raw swan]

13-16hrs

[burnt oriole]

most of the time it's 11 am pst

[keen radish]

why logan hyping veo 3.1? its a minor upgrade.

[burnt oriole]

How do you know?

[stuck adder]

Says who lol

[keen radish]

logan kilpatrick

[keen radish]

veo 3.1 video already leaked, looks same.

[stuck adder]

Where is that?

Let me see

[keen radish]

testingcatalog on x

[burnt oriole]

...

[keen radish]

so, what u expect from veo 3.1? it will be the same, just slight improvement in quality

[burnt oriole]

i hope we get better persistancy so to compete with Sora 2

Image to Video and better prompt understanding

[stuck adder]

That wasn’t real I doubt it

No one has been able to generate with 3.1 yet

[keen radish]

i think its real in google vids, but we will see tomorrow morning

[stuck adder]

He updated his status and said Veo so idk lol

37 min ago

[keen radish]

he said gemini previously, the day before a version of gemini 2.5 dropped

[burnt oriole]

i saw a pop up in higgsfield and I think it said it can generate 1 min long videos

but i think it's impossible lol

[keen radish]

why u trust higgsfield, scam marketing over there.

[burnt oriole]

i never said I trust it

[stuck adder]

@crisp basalt Insert does absolutely nothing lol

[keen radish]

prob a placeholder for tomorrow

[crisp basalt]

it works for me tho

idk im not gonna bother ask what did u prompt etc,, cuz those features are self explainatory lol

[stuck adder]

im in a car with a dog and i ask it to add a random person in the back of the car and nothing happened

gave me the same video

[crisp basalt]

youre in a car?

sorry what

[stuck adder]

inside view front on

[crisp basalt]

...

just uplaod the video here lol

[stuck adder]

😮

[stuck adder]

interesting

[raw swan]

told u 🙂

[stuck adder]

lets see what happens 🙂

[sleek vale]

Where is this from?

[turbid lintel]

Legit?

[peak rain]

Would be nice if we can revert back to older versions

[devout jungle]

Can we choose if we want Flow to generate 2 videos or 1

[copper mica]

Hello, 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.” Please keep in mind that non-English discussions and content may be removed.

[frank dome]

when will Veo3.1 release on Flow? I see some website already can use it

don't want to pay again

[opaque sand]

Is this already VEO 3.1, or did I miss something?

[pallid violet]

@tacit condorAs stated in #start-here・📍. Currently, our team can support engagement in English; the discussion in this server will largely remain in English for now. This community also doesn't have option to generate videos.

[late skiff]

Just saw veo 3.1 come up on my flow app like 5 mins ago as i was using it. Does anyone know the new things that comes with it ? i came here expecting an update or announcement but nothing :o

[zinc elm]

More info coming soon 🙂

[stuck adder]

I see itttt

Woahhhh Veo 3.1 supports start to end frames now

But it’s without sound lol

Never mind

[stuck adder]

I didn’t find any other features other than Veo 3 now supports start to end

[turbid lintel]

im testing the veo 3.1

lets see

[torpid cedar]

Nothing is happening at the moment till it's actual release with information coming soon.

[turbid lintel]

veo 3.1 stuck at 99%

[torpid cedar]

Like I mentioned till it get fully announced

[stuck adder]

What do you mean? It’s out on flow

Announcements wouldn’t just magically add more features that it have now

[torpid cedar]

They will give an announcement on the new update not just release blindly like here you go...

[crisp basalt]

true. well they also released veo inpainting feature in flow earlier today but no announcement was made either , i assume they either rushed it out or forgot the exact time to release it

also hi Nick

you know what, i shall commemorate or unoffically welcome the new veo model to the flow fam

[turbid lintel]

release the option without announcement lol google is noob

[stuck adder]

I’m not disagreeing I’m just saying when you said.Nothing is happening until they announce

[torpid cedar]

Either way...I expect nothing less than greatness with this update. Getting me some breakfast in the meantime

[stuck adder]

I didn’t see any other things other than “insert” yesterday

And start to end frame is finally with Veo 3

[crisp basalt]

theres another upcoming features apparently based on what i saw internally

something something reshoot feature

[stuck adder]

Do well 👀

[crisp basalt]

veo 3 start-end frame is here too!

[torpid cedar]

I have the pro subscription so I have enough to work around comparison

[crisp basalt]

oh my god wait etxtend veo 3 is here too

[turbid lintel]

where you get that info lol

share here

[stuck adder]

Okay that’s the first I don’t try that

[crisp basalt]

i checked in scenebuilder