#flow-general・🌊
1 messages · Page 46 of 1
New perk from flow is that 20 images in 1 hour is "generating too fast" /.....
been waiting for over 2 hours and I'm still too fast.
why Extend/Jump to function is deactivated? i can not find them anywhere
O Flow é uma ferramenta de produção de vídeo com IA que permite criar clipes e cenas cinematográficas usando o modelo de vídeo generativo mais avançado do Google DeepMind, o Veo
Oh... now i see. The UX/UI changed.
You need to click on created video first to expand and see the menu bar
More likely it's just giving the wrong message once people run out of daily quota yeah.
Hi all, as some of you noticed the team is working hard to make the Flow experience better and are listening to you all.
More Aspect Ratios for Image Generation has been a common and repeated request. However, we are still working on the feature and want to make sure it's polished and in tip top shape for when we can get it in your hands.
At this time, I cannot comment on the rumors that the feature will be released tomorrow.
Go on..
It should be saying that you ran out of Nano Banana image generations for today, but instead it's saying you're generating too fast.
So more likely if you still keep getting the "You're generating too fast" message after like 5 mins yeah might as well just come back tomorrow.
Or pay up 5 bucks.
would make sense if I was trying to generate "pro" but I am on the regular, one and I have not used my account for 2 days...
its daily quota for everyone tho
daily quota is 20 images now ? I just said I have not touched my account in 2 days...
Just go create disposable acc , claim one month pro trial and create image if you urgently needed to create the image u want
Or try again in few 30 mins idk
Because you didn’t get the error like what did person said
So it’s not becuz you hit your daily quota
You generated too fast , as what the error says
I'm sorry it's not against you, but i'm frustrated as i loved using Whisk and since this flow announcement nothing is working as it did in my workflows.
?
You weren’t against me or anything , lol stop overthinking my dude
Whisk still available to use until April 30
🤔
no longer doing the model likeliness
subject, > Precise refrence, well not precise anymore
Yeah that’s nano banana
Post the issue here so we can help
Post the image , the prompt you tried to do
i'll try that but i doubt they will fix a tool they put on decomission tbh
i just said i'll do that, you intetionally want to missunderstand everything i can't help you
i learned my lesson thank you
Great idea! I created the code that can generate 1024x1024px using Flow.
Code for Tampermonkey, send me a Google coffee!
for google chrome
If you haven't touched it in two days maybe they're indeed having some peak hours problem or something.
I did it in 1 hour.
But it did happen on my 5 bucks account too once I ran out of free images so it might be just coincidental
Earth aint round, it's egg shaped. Fake composite earth images.
I don't think Flow would give me egg shaped earth images tho
I wouldn't want to live on top of an egg.
Yeah this just dropped recently too, who would've thunk?!
Hate to break it to you.
Damn Nano knows
Talk about warped
anyone having troube downloading upscaled prompts
My model of upscale in nordy: 6k image. https://nordy.ai/explore/69ae61a37ad210690ba031fd use workflow in nordy
Wheres the fancy walkthrough video at boss?
Is everything okay with flow? I was working on it like for 2-3 hours and now it wont load anything
doesnt load here too
same, my project suddenly all gone
Same here
it happened just few mins ago
login
getting 500s now
server down. fix it plz
flow is down?
do some thing dev.
fast run nordy 60s
now it shows this
500 http status code .... server error ... not good ... i.e. soemthing on Google's side
fix it plz
I thought I'd forgotten to pay my internet bill lol
NOOOO
Oh yeah i think thats a thing, there was a discussion about it, if you scroll more up the chat earlier.
who was spamming image requests?
me
same ngl my bad
lol
What's going on
i thought i reached some type of limit lmao
what happen, internal server error
500 internal server error
is google data center attacked now
do something dev plz
hope its not bombed
my all generations are there. my time and effort and money
Lets get a vote react to this message with an X if your seeing this right now.
❌
Do SomETHING
Sorry lol
update in sistem
that part.
veo4 update?
bro your profile card looks cool
in May yes
source?
Thanks 🙂
nice veo4 i test lol
where you test veo 4?
In his dreams
Where can I test it?
lmoa
Flow is updating or someting?
Hopefully it doesn't take too long to come back online
I was smashing NBP pretty hard haha
the technician forgot to turn on the cooling system
Oh my gawd😫
Don't you believe in Brazilians?
Spilt coffee over the servers
Im am Brazilian, We are inevitable.
my time is wasting now. please fix it asap
or veo team should reward members for this waste of time.
gimme me more credits
I don't think it's intentional. Just have to be patient. No technology is fool proof
50 credit
Mine too, i need to proced my work
That's one 4K upscale lol
let's demand flow ultra plan for free 1 month instead
what a nice deal. you are genius
I am a necessary demon.
ICE raid google data center
200 credits is good enough
"Look at his eyes Chico, you can always tell by the eyes."
dafuq is up in dis bi, 2 mails of mine down, anyone know any workaround?
i paid for this service and i cannot use it now and my time is wasted.
my life is wasted on this issue. I deserve extra credit for this incident.
go touch grass
Im not that demanding
hell nah gotta scale
It only stopped because I need to hand in an assignment.
imagine if all our projects just go poof
wheres ur compassion lil parrot
Fortunately I save as I use lol
Bruh, WW3.
i did as well, but only half of them 🙁
Are the servers in the US? Prob sleeping lol
Tomorrow Google will build another data center for you.
the team sleeping
they busy playing with seedance 2.0
Hopefully there will be a VEO 3.2
Here in Brazil too. Its global.
I think so
they busy generating motion control for tiktok dance
don matter bro, still gotta put food on the table
🥶
Imagine if they are rolling out the next update and it broke. But when it's fixed, there's a shiny new model waiting (wishful thinking)
Is everyone having the same internal server error ? 😅
Yeah but the servers will be located elsewhere, likely US I imagine
yee, project is gone as well. for now.
I'm in NZ
It's necessary to release veo4 because it's falling far behind seedance 2.0.
The new model will be released in May.
Oh, yes, i remember now.
Im gonna stay hopeful for that since when everything* went poof on gemini it did come back afterwards 😌
My projects are gone, too. For now.
Actually, American AI video models are not doing well compared to Chinese models tbh
Bro, dream bigger! What we need is Veo 4.0
Sometimes they have released when not expected
Flow is down worldwide?
yes
is Veo 4.0 coming?
The new Kling is actually blowing my socks off lately
no
Less censored too
kling 3.0 is the best now
when will google put out MC in their platform
We need to question why American AI video models are not competing Chinese models..
If only they had an unlimited plan too
runway got unlimited kling 3.0 .. 100 usd permonth
But I think Kling's humans still look very fake
Oh really? Never heard of em
im using it now
Not if you use image references I feel
the generation slow compare to veo , but it unlimited
i hope there will be new information about VEO3.2or4 at the Google I/O 🥲
They are born creating AIs. There's no way to compete.
I am using an image reference, but it looks very unnatural, especially when speaking
I actually think Chinese Video models will surpass American models from 2026.
look at them.
I have been using VEO for speaking scenes, Kling for mostly action scenes that Veo like censoring or just get plain wrong lol so haven't tried the voice compenent
they already surpase with seedance 2.0
All I can say is that every model has its own strengths and weaknesses
100 video for subscription basic is number.
AAAAAAAH I NEED TO WOORK. Flow's GODS, HAVE MERCY ON US.
Is flow down today? I keep getting the “500 internal error”
Yup, it's down
Maybe Iranian hackers got through.
Bill Gates tripped over the cables.
Whats amazon doing right now? 😂
google have epstein files 😄
Their too quiet.
flow is down yes?
Yes
Bro flow is down?
I think the problem is with me. I was creating 150 videos in the same second and the server crashed at that exact moment due to the large number 😅😅
I've been spamming NBP today so lol
We found him, you know honesty goes along way.
@main saddle mam the server is down
Right might as well get some fresh air till it's back online
So, am I the problem? 😁
Maybe a 🐛 , never a problem sir.
Have fun y'all
It back now dude
is flow not working for you all
mine tooo
Nope not over here.
I can work again right now see y’all
mine still not working
Its back for me too 😇 i think we just needed to wait for a lil bit
flow not working for me
Yeah just relax and wait for it
Easy for you to say.
i get to the page but when i click on a project it says no project seems to be here
so its not back over here
Talvez estejam atualizando
sim, acredito que seja isso
is the server not working for all the regions
yooo what the bill gates is going on
Nah man I have a lot of work to do like y’all but am a chill guy and I don’t worry anything
veo 4 rumor is speculating
It came and left again im gonna shush from now on ahahaha 😂 all i know is that the projects were still there so thats already a nice thing
I'm sorry Google, I didn't mean to take down your server.
HAHA
They are the chosen one 😌
How the hell did it go down? What the heck?
it was bill gates
Why the rush? Go make a cup of coffee and take a break until it's fixed. Don't strain yourself in front of the screen. Relax.

When your brain goes Out of existence.
Same.
Yeah whisk is fried too.
I understand but some people do use the site for their jobs or have deadlines to stick to so it can be tricky sometimes 🙏🏾 😅
do you guys use Veo for social media?
is anyone from india?
i like curry
Hi guys is Flow working for you?
im not indian but i generated tunak tunak in veo
Yes, of course But what are the ways to deal with this? Can anyone recommend a new project I can profit from? 😁😁😅😅
Half of Google's programmers are from India.
For some it's back, but not for me.
I see I'm not the only one who's having flow ai problems?
fix it now and give us extra credits for reward
yeah I want 10000 credits
All the staffs are loggin in to fight good luck yall ! 🙌🏾
Everybody loves Potatoes
A potatoe patatoe
hey y’all, we’re investigating. plz hang tight
hello extra credits?
And one singular egg.
hurry.
Good luck guys ! 🙌🏾 ☺️
plz give me crredits and remove the restrictions so I can generate disney characters
double the monthly credit after back on 😄
it's back
its baaaaack
did they find out who broke it though
it's back!
Hey guys, is there a way I can make money from YouTube or TikTok, etc.? from ai
i see many people doing UGC, i guess you can try that
wait until people stop commenting "ai slop" on all the youtube videos and are forced to accept it
What is it sloppy joe day, extraaaaa sloopppyyy
maybe end of 2026
How and where do I begin?
Sorry everyone. Some technical difficlties. Were hoping to get flow back up 100% in the next 1h
AI Certified, tired of AI, AI B... 😄
yum
Time to get to work, boys!
I have 420 credits.
.
you can check youtube for that.
it's back guys
But where is my next month extra?
rich
do you guys get 1000 credits free i have just seen that a week ago
You guys are getting credits...
never seen free stuff 🙁
two
Am using flow since I have my job and I pay for ultra every month and I got nothing for free
Ok we should be back again, sorry for the wait!
Dont let it happen again. 😉
Also thank you.
goat
Brazilian salary: R$ 1,621.00
hey are you brazilian?
yes
Everything here that depends on the dollar is 5 or 6 times more expensive.
it's hard to follow
Is it cause the beef over there? Heard the steaks are 🔥 over there.
they video playback still lagging here
we should be back now y'all! thanks for your patience 🙏
thanks team
hello
Try using Gemini
would anyone like to see the app I made and give me feedback????
Go on... Depends better not be a hotlink.
nah
was thinking join the lab-lounge and i'll share screen innit
or should I just post some screen shots???
or I can send a hotlink with an internal virus
...
I'm not much of a talker atm, i'm on 3 projects right now, limiting my agentic uses by sharpening my actual skills rn.🙄
Where them screenshots at?
ight bet
Make sure its in the proper channel & etc..
Gallery for Screenshots fs.
@frosty niche any way you can join and not talk in the lounge
there is a few sections of my app but like
it's weird to explain but ill try to explain so it doesnt sound sketch
C'mon...we aint taking screenshots no more?
Wheres your MarkDown at?
You know, documentation is stilll important now-a-days..
so what do u say... 🙂
it'll be rq
u can mute ill show u the app on share screen on vc then you can leave and give me ur thoughts
np thanks for ur time tho
🐛 ⏰
This one is interesting, when i downloaded the top image, the download cut the top of the downloaded image off.
I downloaded it 3 times to make sure, & all 3 were cut off at the top.
All 3 were 2K.
#DoBetter
🐛 ⏰
2k On the left & 1K on the right.
2K is fried, but 1K is doing fine.
Anyone has try seedance 2 here?
Seems I picked a good time to go to sleep early, didn't even see all those problems, Flow's working fine on my end now. But I guess they are indeed having some peak hours problem. That would also explain why it was giving the wrong error messages yesterday.
Been having fun trying Nano 2 for making photorealistic cosplayer images, by default it makes them a little chunkier than Whisk's nano, but it looks good enough
Also it was giving them plastic hair, had to write "well taken-care of real cosplayer wig hair quality. "
Cause it wouldn't understand by itself.
Are human heads really that big. I guess only when the person's small. Gonna have to try a bit more with heights and scales.
Friends, yesterday I was using Flow for images and I came across a bug or a possible update issue. It had several image aspect ratios that are usually only available in other paid tools, and also the image generation was failing a lot with simple things. Has anyone else seen this and had problems generating images?
yes but now we can use the tools
Yeah they seemed to have some server problems yesterday should be fine atm. And the other aspect ratios are gone now, seems the feature wasn't ready yet, I imagine they'll bring it back when it's ready.
Why just ultra plan without watermark 
what watermark
It was horrible last night dawg.
But we got through it.
Veo
I'm more bothered by that hand, prompts can help hands be better but there'd always be at least a couple weird frames in some vids.
Nope, just tried that pose and almost snapped a finger
@wet copper @main saddle @tacit halo
I think we should make it so extend also have feature for start and end points.
Making the models taking few frames from before hand of the original generate video then generate more frames to make it suitable for the start point of what user want then generate till the end point.
I don't know if this idea aren't clear enough, but i just comeup with this
.... they are back 👀👀👀 sorry for the wait!
Ah just in time, thank so much.... but could you please also add 2:3 and 3:2 when possible? That one's useful for more visibility in art website galleries.
I don't think many more others are needed but people would say
1:1 sure is must atm and it's there now
We will as soon as we are able!
Can I ask, why other creators have good videos and when I created video about dancing female, it created it with bad dancing rotation? Why I have bad video and other have good videos? Thank you for help.
Characters do have problems when twirling or spinning in whatever ai video generators not just VEO, so you either include a bunch of prompts for the characters to have consistent morphology and anatomy and fluid unhurried coherent movement, or just tell in the prompt that character should only face front and never twirl.
Also if it does need to turn around, ensure you tell it to what side exactly, needs to be either left or right in a 360° for example, or just turn left or right to backside.
Otherwise if you just say "Turn" it might just constantly flip left and right guessing, needs to be specific
I written video prompt: "Female is dancing." It is bad?
No, that should do, but VEO does need some more help sometimes there, that or pray to sometimes get the right video. You can only know after trying tho, if you see mistakes in the video just add the corrections in the prompt for the next one.
finally flow can generate 1:1 images
my life is complete!
Veo 4 when
I managed to create a system that generates 4k (creative upscale) in alpha, but it depends on the Pixelcut (Pixa) API. Google needs to buy Pixa for this new upgrade.
test
Aspect ratio back
In fact, the image aspect ratios are there, but it's very unstable; it's not generating anything for me.
Anyway about the dancing topic cause talk is cheap, here's my attempt, first auto generated with a dude without reference image. Because equality.
Took me a while to find a safe enough woman image ffs
More often the videos come out fine like, 4 outta five, gotta be speficif about it.
[-Pointers: Make the animation coherent and unhurried with way more than enough frames to be fluid but don't repeat any frame and keep all frames at the same speed. Keep body's anatomy accurate and morphologically correct to any body's sides or in relation to poses.
When rotating the body can only do it until a 360° finishes, the whole body can only move at the same speed as the backside and can't unlink/flip/pose anatomically/morphologically unnaturally.]
Scene: The character dances to a popular existing k-pop intricate choreographed non-simple acrobatic routine with related music. must include at least one twirl towards the same side.
Identity of character: True to reference image, same facial structure, same face, same props.
Camera angle/Composition: Locked to the same one from reference image, character must remain full body in focus.
Background: Coherently related to scene. No transitions.
[Video Quality: All professional. Remastered 4K HD ultra high resolution, sharpest definition, denoised, unblurry, clean lines, High contrast, fix any jpg artifacts.]
Can also include camera panning and other things in the routine, it's kinda obvious it can be locked to body too.
But yeah as long as one's even trying for it to work they can twirl just fine, maybe can be simpler than that idk, ran out of credits already.
Also I should probably add hand morphollogy next time, too quick a test.
Adding prompts against intersection, clipping and overlapping sometimes is a good idea too, that way limbs clothes and other things can't go through each other, depends on what the character wears. That goes in fixes.
Fixes: No unnatural clipping/intersecting/overlapping.
That way the AI thinks every outline has volume and profoundity.
bruh flow blocking everythign with nano now
did they censro it more?
like you cant use real faces as ref anymore?
Same thing here, no matter the prompt.
Hiii
I'm a large content creator trying to create movies using flow and I want to add elements to my video as the video is being played. How come there isn't a video to video generation?
Please, can ask something? It is important for me.
I've gotta say that I've hit flag output so many times on the types of videos I've been trying to achieve, I think the message has finally been received and the generations are working now and absolutely AMAZING! WOW are they amazing!! Kudos to those who have corrected the model input etc. to make it happen! I can now see content that 80s filmaking couldn't achieve due to tech not being available lol!! And you never saw it when the tech came about because it's definitely a rare/odd niche 🤣
i have a girl and guy in swimsuits in a hot tub, im just trying to get a shot of the camera dolly in and back out why does it leep saying its "sexual content"
Great now I can't stop doing this
People more often add their own music tho, what do I know
Can I ask you something? I don´t understand one thing in images.
Shoot
I use as reference basic shirt and I write prompt: "Female has worn this shirt". Left image is by Nano Banana 2, right is Imagen4. And my question is, why Imagen4 do always images so very close, no same as left image - NB2? I tried it with other images, but in Imagen are images this so very close, without faces/heads.
Why?
Imagen 4 is crazy with the artistry, so it does what it wants but it more often goes better than that, I think your prompt is centering on the shirt tho, that's why it does that. Maybe if you try
Composition: The female must be in full body, cowboy shot.
Or
Composition: The female must be in precise cowboy shot from hair to waist.
That unless you do want full body, then you have to say her footwear must be included. These can be very picky about it
I add to previous my prompt: "Female has worn this shirt" your prompt: Composition: The female must be in full body, cowboy shot. and it created this:
Damn it's got hands. Are you using imagen 4? I'm not good at that one
By Imagen 4 is this image without head
bro nano in flow is literally block everything no matter the prompt
I created this image with Imagen4.
I think it's because of the first part, it still focuses on the shirt- Maybe reverse it, instead of "Female has worn shirt" "There's a girl seen from head to waist, she wears this shirt"
The first line of the prompt always gets the most priority. I gotta eat ravioli brb.
No, same.
Guess I'll go find a shirt
We just need to do something totally different instead of trying to fix that, let's see now
Yeah I think you have some other line of prompt conflicting there somewhere, might just have to rewrite it something like that
Imagen 4's working nice really
But, you can see, shirt image has closed collar, but on girl is opened
I think you can add that to scene, it should catch on fine, idk I can't see the rest of that prompt but yeah more likely you just need to remake it seems
Okay, thank you for tips, I tried it by your prompt used on image.
I need practice with imagen 4 anyway, damn makes nice blondes
I have one question, you have propnt splitted to sections, scene, compostion... It is better than all write in one long sentence?
It depends what you're aiming for, if you write a single line easy to understand in the correct order it will catch on just fine too
I divide it so it gets things easier by separate, but yeah if you do that you better be doing the encasing well
But it's not necessary really
Okay, thank you, I try it by your splitted prompt
These imagen 4 realistic images are nicer than in Nano. I get it.
I don't think it's gonna give me all the same accuracy tho, but for random stuff it's awesome
Also idk what kind of censorship stumps are people hitting with Nano 2 and VEO atm, I did just post two generated vids of an anime girl twerking up there so what gives.
And they worked first attempt
You don' t like it, or what?😀
no, its just since nano pro i never looked back lol
Yeah but I guess those realistic imagen 4 generations do show their fanbase.
I am not Pro, so I use only NB2 or Imagen
Can't judge
imagen is blocking me too lol
Oh yeah imagen 4 is surely more censored than the nanos, so no wonder
posted a face as ref, the prompt is- walking
We wanted to come on here to clear the air and confirm that the rumors are true...
Cinematic Video Overviews are officially rolled out to 100% of Pro users in English!
Please respect our privacy during this time by flooding our replies with your favorite creations.
nano and it is saying everything i do is flagged
hello, I'm trying to upload my 2K visuals to Flow, but it's showing an "enhancement failed" message. Did I do something wrong? Or is it a technical issue?
thank
I did notice some more censorship in Nano 2 and Pro since yesterday too. So idk you either do it 12 images at a time and pray one works, or rewrite the prompt.
Are you uploading them one by one or all at once? Maybe it'd take it easier 1 by 1. I've uploaded 2K images and they work.
I downloaded them one by one, it worked, then I got this error... How can I download them all at once? I'm interested.
I don't think that's available yet, however downloading 100 images or so 1 by 1 doesn't take as much time as one would think, I've done it in 10 mins or so from PC.
You can keep the left mouse button pressed and drag+select all in your gallery then right click select download to download all at once, but that will only give 1k images, for 2K direct upscale you need to download them one by one.
And the upscale from Flow is strangely better than any other uspcaler I tried, it keeps very small details
Getting a lot of "interests of third-party content creators" failures today even on videos with prompts and Nano-generated images that have no resemblance to any sort of copyrighted content. Not sure what that filter is based on, but it definitely seems overly sensitive.
Yeah Nano 2 is just able to detect and recognize most popular stuff from any franchise, for example Zootopia stuff is more often a no-no unless you're very patient and ambiguous about it.
Okay, thank you so much for all this information, it's very interesting! Indeed, the hover function works and I can download several files, but there's only one image in the zip file. And for now, downloading them one by one isn't working. So I don't know if it's a technical problem or not...
Nah it's not just available yet, but you can ask for it at "labs-bugs-feedback" so we can gang on them about it, it's been needed for a while ago.
ok thanks I'll do it 🙂
Going back to this, even tho they fixed it a lot in the past months cause it used to be much worse, I also still have probs with it as in just sometimes recognizes the character as if it were from some existing canonical thing, and gives that character's face.
I know image models aren't supposed to work like that, but besides the image generation ability it's like Nano 2 itself would also be connected to some sort of powerful search engine or something, or at least it updates itself about new content each two-three days or so. And that's good when you want to make an image of a canonical character but yeah the other thing could also happen.
I've seen this somewhat helps on it tho.
[Character's identity: True to reference image, no creativity freedom, no external sources should be used to create this character. This character is NOT from any existing series or franchise.]
But funny enough you can also do the opposite and put something canonical there, and if it's known enough it will also get accurate about it if you want. Put character name, franchise name, who owns it, who draws it, etc, all data you can and it will know.
Much better than Nano 1 can.
In this video, i wanted only Banana to speak but kiwi kept on speaking. Which video generator is best to avoid this speaker attribution error?
How do i fix this issue on flow?
Provide the prompt you used for it.
Banana — banana fruit head, muscular athletic build, wearing an open green tropical floral shirt showing his chest, black cargo shorts, jungle boots. Slightly taller and more confident.
Kiwi — kiwi fruit head with fuzzy texture, average build, wearing a tan safari shirt and camouflage shorts. Kiwi does not speak in this scene and only reacts with facial expressions.
Banana leans toward Kiwi with a mischievous grin.
Banana says quietly:
“Tonight during the games… I’m kissing Cherrita.”
Kiwi turns toward him with a shocked expression but does not say anything.
Banana smirks confidently.
Banana continues:
“Hell yeah.”
He pauses and adds with a sly smile:
“Just wait for it.”
I was gonna say pretty much the same yeah. Maybe try giving the model more visul cues and pointers so it can be sure about things, even tho they might be obvious to you.
Like "The yellow banana in hawaiian shirt says" then "The round head kiwi in brown shirt says"
BRB
I’m thinking if the model can’t pick up fruits then will they pick up outfit colors
Can you try that from your end
Yeah lemme go see what I get, I need to learn about this anyway. I've done with other things... not fruits tho
What's with this gen and bananas
Hope this helps.
https://docs.google.com/document/d/1evv29jTcIhDVIe6fZCi9gA0sWKIs_Qawj9X8ONaS1bo/edit?usp=sharing
@steady bobcat
INTERNAL ADVISORY - FLOW-BASED CHARACTER LOCK TO: confidentgeek (Banana & Kiwi Prompting) FROM: Gaston / GasMan SUBJECT: Implementing "Flow" to Solve Speaker Attribution Errors The "Flow" Implementation Chart The Problem The Flow Solution Technical Implementation Dialogue Bleed Use Layere...
Can i dm? This isn’t clear
Ehh no dms sorry..
what you got on your mind @steady bobcat ?
Practically, can you tell me how this works
I have a prompt and photo. What do i do next?
work on your prompt structuring and use some elbow grease to fill in the confusing parts..
There’s no confusing part which is why i use it. If you feel there is, please point it out and I’ll correct it
Since most outputs are 8 seconds long, id say something like this usually helps with prompt structuring, especially when you need to zone in on a specific sequence, like that dialogue you want.
[0:01]
[0:02]
[0:03]
you got what you need the prompt, and photo, just the structuring is all that needs work.
Still doesn’t work. Timing doesn’t change who speaks at said time
Okay what structuring?
Here
[0:01]
[0:02]
[0:03] <- pretty simple.
Timing does not work still.
yeah the new revised version with the timing sequences added to it, you can redraft it use and then check.
Timing changes when something is said, doesn’t change who says it
is there anyone that can answer my question?
I'm a large content creator on youtube. I'm wondering why there is no video to video generation? I'm trying to add elements to my video as its moving.
I have
Banana — banana fruit head, muscular athletic build, wearing an open green tropical floral shirt showing his chest, black cargo shorts, jungle boots. Slightly taller and more confident.
Kiwi — kiwi fruit head with fuzzy texture, average build, wearing a tan safari shirt and camouflage shorts. Kiwi does not speak in this scene and only reacts with facial expressions.
Banana leans toward Kiwi with a mischievous grin.
Banana says quietly:
“Tonight during the games… I’m kissing Cherrita.”
Kiwi turns toward him with a shocked expression but does not say anything.
Banana smirks confidently.
Banana continues:
“Hell yeah.”
He pauses and adds with a sly smile:
“Just wait for it.”
Yeah your still not getting it, but hey i tried.
I have added the timing and it’s still wrong
Also this is not the first time im using timing so i know it doesn’t make a difference
So you mean the essence of this document is for me to add timing?
You gave the same prompt twice. Its cool if you didnt fix it, its your video.
Oh let me share the new one
Anyone know of a program that can do ai video to video generating?
Sorry video to ai video generating with real footage as i want to add elemnts of a zombies attack my house and I'm trying to escape from them.
Exactly, this was the direction.
Can anyone answer this or we just ignoring me lol
only seedance had this ability to do video to video and it was removed from the US market is this true?
I added a prompt example in the G-Doc, that should give you a decent starting point.
Thx.
@steady bobcat
Banana — banana fruit head, muscular athletic build, wearing an open green tropical floral shirt showing his chest, black cargo shorts, jungle boots. Slightly taller, confident posture.
Kiwi — kiwi fruit head with fuzzy texture, average build, wearing a tan safari shirt and camouflage shorts. Kiwi NEVER speaks in this scene. Only reacts with facial expressions.
⸻
TIMING + ACTION SEQUENCE
[0:00–0:02]
Banana leans closer toward Kiwi with a mischievous grin.
Kiwi looks relaxed.
[0:02–0:04] (Banana speaks — quiet tone)
Banana leans in and says softly:
“Tonight during the games… I’m kissing Cherrita.”
[0:04–0:06]
Kiwi turns sharply toward Banana, eyes wide, shocked expression.
Kiwi says NOTHING.
[0:06–0:07]
Banana pulls back slightly, smirking confidently.
[0:07–0:08] (Banana speaks)
Banana says:
“Hell yeah.”
[0:08–0:10]
Banana pauses, maintaining eye contact with Kiwi, sly smile growing.
[0:10–0:12] (Banana speaks)
Banana adds quietly:
“Just wait for it.”
There we goo!!!!
Thats the jist of it, but how does it look?
The video result loading 😭
Even worse 😭
Yeah, i used that thank you
This is the one you used?
Yea
BRB
bytedance that owns seedance had the video to video generation they removed it for now copyright issues.
Give me a summary of how bad that output sucks really quick @steady bobcat
😭😭
Honestly the team needs a real quick fix for that
Got a big page and that is the MOST discouraging reason to keep using flow
Mind you not one of the most but the most
I updated the G-Doc, review it & give that 1 a try.
Everything is a work in progress. 🙌
lol, nah for real though i need that summary from you so it refines better.
Okay. Do i use the refined flow or the formatted?
Let me know how to share the summary and i will
Review the section/prompt i added to the G-Doc. Analyze see if it will work and/or tweak it to give your direction, and then give it a whirl/run, is all.
Summary rt here for it, i just need to literally point out how bad the generation is so the refinements are on point for the next gen.
Same ole same ole.
Are you talking to me here? 😂
Alright brb.
Dang you deleted it super quick, drop it 1 more time. Mybad.
I don’t see an option to do that in the doc
alright were good. delete that sucka
Provide the last prompt you used really quick. Please & thank you.
[START SCENE] Jungle background. [0:00-0:02] Medium shot. Banana (Banana-head, athletic, floral shirt) leans 45-degrees toward Kiwi (Fuzzy-kiwi-head, safari shirt). [ACTION] Banana flashes a mischievous grin; Kiwi maintains a relaxed, neutral gaze.
[0:02-0:04] [SPEAKER: BANANA] [MUTE: KIWI] Camera focuses on Banana’s lips. [DIALOGUE] “Tonight during the games… I’m kissing Cherrita.” Banana speaks softly. Kiwi remains static, eyes beginning to flicker.
[0:04-0:06] [REACTION SHOT: KIWI] Sharp camera cut or pan to Kiwi’s face. [ACTION] Eyes snap wide, brow raised, jaw slightly dropped but [NO MOUTH MOVEMENT]. Silence on the audio track.
[0:06-0:08] [SPEAKER: BANANA] Banana pulls back to a confident upright posture. [DIALOGUE] “Hell yeah.” Banana smirks at Kiwi.
[0:08-0:12] [HOLD SHOT] Banana maintains steady eye contact with the shocked Kiwi. [0:10-0:12] [SPEAKER: BANANA] Banana adds quietly with a sly, growing smile: “Just wait for it.” [END SCENE]
Sequencing is off but i feel like the dialogue is a bit in the better direction.
Refine, refine, refine!
Oh no definitely better with both faces showing
Provide what you wish to accomplish for this prompt/scene again.
Or summarize it w/e..
We almost there.
🤏
I wish just for each character to say what they’re supposed to say. That’s all
Actually it looks alot better now that i review it, but lets get to best not better.
Alright
Bruh lmao, did you write up a script or we freefalling?
1 moment brb.
How do you mean
Disregard it, mybad.
Alright
Check the G-Doc, & then let me know how it looks after.
Should be at TT3 section.
Alright seen. Trying that out now
Its like the lip sync starts off connecting then falls off &out of sync at the end on this one. Pretty close.
The camera motioning is kind of fire though.
Preferably not bc it’s supposed to show both faces
Honestly this has been a nightmare
Idk how anyone else copes with this
Alright well, good luck.
The new update should allow users mark characters that should speak and in what order
Before hitting the generate
Asked a fellow creator and they got same issue. They literally have to generate tons of a scene to get a 8s scene and you may still end up not getting it
So far this week i haven’t created. It’s exhausting. Trying out other models too
Other than that i love the flow quality. Number 1 across board, no one else coming close
Cool, yeah, man, practice is still very much a thing.
Keep it up, & remember the refinements, summarizations, and sequencing helps alot.
I left you one last TT4 if you want to give it another refinement test in the G-Doc.
✌️
does flow not allow to select aspect ratio?
@steady bobcat
I think you're right about VEO's inability to recognize kiwis and bananas. I can keep the dialogue stable, but not swap their dialogues yet.
I've tried coordinates, left and right, pointers, height, colors it just doesn't know what these are.
EXACTLY. It doesn’t
Does that banana have abs?
The fact that it doesn’t even recognize the fruits
At this point I'd just swap their place at a new image, that could fix it
Its the abs probably. 😂
Mind you this is just two characters. Now imagine when they’re 3 or more
Let's try a different image generator just a bit just to be sure we're not nuts
Video generator you mean
I was thinking fruits distinction should work better than human distinction which is harder. But apparently the model is finding it difficult to differentiate a kiwi and banana
If anyone tries a different video generator and it works do share
I left a TT5 in the G-Doc, in case you get bored.
Thank you 🙏🏾
Any news about being possible transfer projects from whisk to flow?
I havent heard much as of late.
Any London meetup happening soon. Y’all let us know
London creatives i’d love to connect. You can pm
Well I got some more time to try and tackle this now, let'se see
Well again Nano Banana 2 itself has no problems to differentiate right from left
It takes it as viewer's right and left, and doesn't miss even once, this is how VEO should be thinking, it'd make videos so much easier
Well now that they're split I'll try it with them brb
Ok I almost got it this time, just had to deconstruct my prompt a lot.
Haven't added the same character structure yet but I do know how to keep them talking at the right side now.
Where'd his shirt go?
Like said I haven't added the structure yet, gotta be really careful so the dialogue works correctly
I can work on that now, I think
We rocking chucks and just throwing the boots outs. 😂
The flow is there, nice.
Well I could absolutely use this too, sadly the other easier vid platform I use suddenly wants to charge $30 a month so I couldn't practice faster there.
So it goes like.
[First action:
Only the character from the left of the image repeats this quoted sentence exactly without any creativity freedom: "Tonight during the games (pause) I'm kissing Cheria"
-break sentence no dialogue here-
Second action:
Only the character from the right of the image repeats this quoted sentence exactly without any creativity freedom: "Mate, are you serious?"
- break sentence here space with no dialogue-
Third action:
Only the character from the left of the image repeats this quoted sentence exactly without any creativity freedom: "Hell yeah!"
[Video Quality: Keep the characters accurate to reference image, no creativity freedom. All professional. Remastered 4K HD ultra high resolution, sharpest definition, denoised, unblurry, clean lines, High contrast, fix any jpg artifacts.]
Ends up VEO can difference left from right, it just for some reason doesn't know the right of what I am talking about, so gotta specify "The right of the image"
You're good, do whatever is in your budget frame atm
Either way, the dialogue is in sync which you tackled.
Looks like a book written in json, interesting pattern lock with this prompt.
Might think that tomato or whatever is leftist
those sentence breaks, fascinating.
But yeah that consistently works to keep the dialogue on one character or the other
Its like giving the model a que to breath and switch over to the right and/or left with this prompt.
Interesting formatting.
Also idk why it said "Boor" probably need to remove more creativity freedom there
I think it was the "Pause" thing, it didn't listen to me on that anyway
Yeah that 0:02-0:03 wt was that, tried to pick up rawr, oddly it sounds like compression at that marker.
-break sentence no dialogue here- <- def the reason why the boor came out.
Maybe a sequence like this could be added later.
Sequence:
[Left Character]: "Tonight during the games... I'm kissing Cheria."
[Technical Cue]: 2-second pause. Total silence. No background dialogue.
[Right Character]: "Mate, are you serious?"
[Technical Cue]: 1-second pause. Total silence. No background dialogue.
[Left Character]: "Hell yeah!"
Has it ever listened to that for you
It seems to step onto itself when one gets technical
Audio wise(boooor) its happened to me but I can't remember which gen it exactly was.
That balance between the creative/freedom to technical ratio is important.
I dont blame it, lol
Sounds like a back drop to a heavy metal chorus tbh, "Boooooor". 😂
Also I was using brackets, but seems it's using them to connect the sentences when there's two characters in the image
Brackets were working fine for me when it was just one character
Have you tried json prompting before?
{
"video_sequence": [
{
"action_number": 1,
"character_source": "left_of_image",
"dialogue": "Tonight during the games (pause) I'm kissing Cheria",
"constraints": "repeat exactly, no creativity freedom"
},
{
"action_number": 2,
"technical_cue": "break_sentence",
"audio_state": "no_dialogue"
},
{
"action_number": 3,
"character_source": "right_of_image",
"dialogue": "Mate, are you serious?",
"constraints": "repeat exactly, no creativity freedom"
},
{
"action_number": 4,
"technical_cue": "break_sentence",
"audio_state": "no_dialogue"
},
{
"action_number": 5,
"character_source": "left_of_image",
"dialogue": "Hell yeah!",
"constraints": "repeat exactly, no creativity freedom"
}
],
"technical_specifications": {
"visual_fidelity": "Keep characters accurate to reference image, no creativity freedom",
"quality_standards": [
"Professional",
"Remastered 4K HD",
"Ultra high resolution",
"Sharpest definition",
"Denoised",
"Unblurry",
"Clean lines",
"High contrast"
],
"post_processing": "Fix any jpg artifacts"
}
} <- this is what that prompt would look like in a json. Might be a completely different gen/output though.
Have YOU tried it
Back in like 2023-2024. lol
Been awhile for me. Now you making me want to try it.
idk now that I try VEO more it seems to be based a lot on some Chinese models, it has the same stumps and works differently from the Nano image models.
And it's missing a lot of reasoning, stuff like Grok is miles ahead
Yeah, but the ui and the tool comprehension take a massive hit on the UI/UE imo, but I see what you mean.
Did u try it yet or not
Let me finish with this piece first really quick. Working on the sizing/formats 1:1 3:4 with some comic panels first.
But Has the mega busting person tried it yet
Yeah @pulsar coral have you?
Yeah but for chatbot making, I've never put it in an image model
For what it could matter
We've seen amazing Flow videos created using JSON prompts! It's not the only or even "best" way, just one way to help structure hyper-specific visual directions. No JSON? No problem! Regardless of how you prompt, it helps to be specific with your setting, subject, action, lighting, shot type, camera angle, and audio for strong results.
︀︀
︀︀See the prompt used to generate this video below!
︀︀
︀︀Static 35 mm full-frame 16:9 shot reveals a vast pale-grey warehouse: plain back wall, concrete floor, grey rafters above. A cardboard box sits center frame. In one seamless motion the box bursts open; furniture erupts upward, sweeping out on fast, elastic, physics-true arcs. Graphite sofas, patterned rugs, tall shelves, and bronze floor lamps land neatly behind and around it, while a solid table thumps at the exact center of the layout. A cushion rebounds onto the sofa, a framed picture snaps to the wall, and a bronze pendant fixture swings from the rafters as dus…
What I see in that prompt is it seems to follow the back to front order.
I guess that works better for multiple elements than a single moving character.
That’s not the json prompting format
Yeah something told me when things weren't pink
It’s been a while since I used the JSON prompt format. Maybe I should give it another go
Every time I try to keep the characters as they are with this, it breaks the dialogue.
He'll yeah.
I'd have to dive deep into my project work, projects, and etc. just to even find the last time i used it lol
Do better.
Yeah I don’t know
I think json format gave me uhh choices rather than followed my core idea I guess?
I got to get the used to all these options now.
... c'mon brain you just gotta think in retro prompt here
Don't post too many screenshots of that or a green guy will come and take them away again
It’s the HEêA🤖 LL Yeah! For me
Are you serious
It could be "reference image" chinese generators don't use that
Yeah one of those
We should have nametags for this stuff by now 🙄
Nope, but it made the thing more kiwi-like. But yeah this kinda proves this keeps the characters talking on their side
Wait I like the kiwi design
It’s the robotic hell yeah for me
Gonna try some other thing than "characters"
Are you serious
Yeah I can try adding the tones next
Also I was pretty sure VEO didn't know what a Kiwi is
If it can give it a Kiwi head then why can't it recognize it to make it talk
I feel like I'm with sticks and stones here
He'll yeah, the kiwi is rocking that 🥝
Yeah, that he'll yeah was weak.
Lol is that elvis on this mans shirt. 😂
I wouldn't think that vid is fake without the watermark tbh
I think it's the image detection of the antropomorphic characters that's failing
Or fruitmorphic whatever they are
Wait the banana is more an avocado than a fruit
make in flow : prompt : Final stand, operator back to wall, low ammo, multiple hostiles approaching through smoke, last stand defiance, shot on RED V-Raptor, 50mm, f/2.0, Call of Duty Zombies desperation, heroic sacrifice, impossible odds, cinematic hero moment
Upload 3 images of your face and use the prompt.
for NFS FANS prompt: Cinematic medium shot,wearing hoodie under leather jacket, young man leaning against BMW M3 GTR silver with blue stripes, arms crossed, confident stance on abandoned industrial dock at blue hour, carbon fiber details visible, city lights behind, shot on ARRI Alexa 35, 40mm, f/2.8, wet asphalt reflection, Most Wanted hero and his machine, driver-car bond, underground racing legend
His feet kicked off his chucks
Yeah bottom line here is the prompts I try work just fine when the characters are alone or without reference image, so they're probably fine. But when there's two, VEO's detection goes wild.
Hi, flagging a case study for the safety/compliance devs. While working on a stylized, gritty 'SpongeMan' project, Flow bypassed standard IP guardrails and generated high-fidelity versions of SpongeBob and Mr. Krabs.
Observation: The model seemed to over-index on the character names, actually overriding my 'hand-inked/gritty' style prompts to deliver the branded look instead. Attached are the reference style I was aiming for vs. the actual branded outputs I received.
Happy to provide the full prompt strings if needed for the training team. I also, have already messaged @real wyvern as well.
Thx!
I mean you know your flow the best dawg. Just refine, refine, refine!
Hello. When I was creating video in Whisk, there was written, how many generation is available, but in Flow I can´t find it. Can I find out somewhere, how many videos I can create in Flow?
why is this keep happening to me?, ive already wait multiple days but it still dont work
Will Flow have the options to make auto aspect ratio images based on the aspect ratio fo the input reference images?
Indeed that's also called "Auto" aspect ratio, they haven't said anything about it so I don't think we users know that, but yes it should absolutely have it. Google AI studio and Gemini have it too so it exists.
Basically if you could pick Auto aspect ratio, you upload any reference image of any reasonable size to the image generator, and Nano Banana should create it in the exact same aspect ratio. But I've only seen Nano 1 doing it, not Pro or two.
Because Nano 2 and pro can also select aspect ratio via prompts but can't be done at Flow only Gemini and I think at Google Ai studio too, for example you end your prompt like
Aspect ratio is 6:2
And they do it.
Only two mods alive, new record. Good night too tho.
Also might wanna ask for the feature at the "labs-bugs-feedback", could happen.
thank you for you reply!
You could be generating more than 10 images at once.
My issue is different, but no idea what's the cause, mine is generating images, there's a random chance the generation doesn't start and result on "Failed Something went wrong."
I haven't seen the "You ran out of daily image generations for Nano Banana 2 🍌 for today" message anymore in 3 days, so more likely Flow has a bug about it.
And instead of that message when you ran out of image generations you get the "Failed" or "You're generating too fast" messages
So yeah if after 5 minutes you keep getting one of those you might as well just come back in 24 hours to it.
I doubt I've generated more than 100 images today, I just started xD
Also besides that one, seems some free accounts are getting limited even if you haven't used the account for days. So that could be a bug, it happens during peak hours.
But yeah, I know the limitations of Nano 2, but not sure about Pro
Or at some point you tried to generate more than 12 images at once and it put rate limit on that account, specially if it's free.
Peak hours could be the issue too, seeing that NanoBanana has such realistism in their generative
Eitherway, the 3 things are happening.
Cause yesterday Flow did go offline for 3 hours or so seems.
I got the Premium plan as their video gen are also fairly good
I'm aware of that too.. since they are closing Whisk soon
Anyway whichever of those happens, the account is fine the next day at worse so yeah we'll live, I think.
would anyone like to view my app and give me ur honest critisim
And sometimes refreshing the page helps
works 👍
incognito mode
#4:3
make in flow : prompt : Final stand, operator back to wall, low ammo, multiple hostiles approaching through smoke, last stand defiance, shot on RED V-Raptor, 50mm, f/2.0, Call of Duty Zombies desperation, heroic sacrifice, impossible odds, cinematic hero moment
I used a panel set up instead, of 3 whole images. 🤙
#4:3
Is this some F joke from Google ??? I am trying to generate a video and I am getting a constant policy violation error. I kept trying and for a moment I thought it was the dialogue but no matter what I was input as a dialogue, I was getting an error. Then I swapped the character from female to male, and it generated. Is this a joke??? I am asking seriously here. Where can I send feedback about this??? I am so tired of this.
PROMPT: Camera Movement: The video starts with a highly dynamic "crash zoom" and "speed ramp" effect—the camera aggressively and rapidly pushes in towards her face, then suddenly smoothly stabilizes. Action & Expressions: As the camera stabilizes, she looks up from her smartphone, smiles confidently, and gives a playful wink. She taps the phone screen with her finger. Audio/Dialogue: "Great matches coming up this weekend! We have football action from all around the globe. Are you ready???!!" Style: High-energy, cinematic sports promo, 4k.
PS: There are no breasts or cleavage exposed. Is just the company's logo that I can't disclose.
just a quick question if i sub to Pro and generate videos in Flow , can i download them without watermark ?
no
thank you
Thanks
was this image to vid?
im gonna try wit this
workflow : ingredients to vid
managed to generate all four of them
same with this
now imma try using image to vid
A tight, cinematic interior cockpit shot from a low-angle POV, looking up at [Insert Character/Animal], callsign [Enter Flow], piloting a high-performance skimming craft at Mach 1. The craft is banking hard over the mirror-like glass of the equatorial ocean at sunrise. The focus is razor-sharp on [NameTag]’s specialized flight helmet, featuring integrated haptic sensory-conduits tailored to their anatomy that vibrate in sync with the craft's external sensors. Outside the canopy, the horizon is a seamless blur where the deep indigo ocean meets a burning orange and gold sky, the two elements bleeding together in a shimmering heat haze. The cockpit glass is detailed with micro-vibrations and beads of tropical condensation. Internal cockpit lighting is a soft, dim amber, contrasting with the blinding, cinematic golden-hour light flooding in from the side. Shot on RED V-Raptor, 35mm, f/1.4, emphasizing the "motion blur" of the passing waves and the serene, intense focus of the pilot.
Choose your [character otter, dolphin] or whatever comes to mind. Dont forget your [name tag].
image to vid no issues as well
idk why you encountered these errors
veo 3.1 fast
Allllllllll arrrround the globe, stutterman.
Sup @crisp basalt
I tried generate like 10 times now same error
🔥
Lol you got to fill in the callsign but 🔥 .
Here fixed it for you homie.
That first video zoom out was so 🔥
I am a paid subscriber of Google’s AI services. I am currently using the VideoFX 'Flow' tool to create videos, but I am experiencing a persistent error.
After generating a video using the Flow feature, the 'Add to Scene' button does not work and keeps showing an error message. I have already paid for the subscription, but I cannot proceed with my project due to this technical glitch.
Could you please investigate this issue and let me know how to resolve it? I would appreciate a prompt response.
can u screenshot the error
elias is online rn so it would be more helpful if u include a screenshot of the erro
so the team can try reproduce the error u facing
Hey, could you open a bug report in #1238621511401734254 and include details like your browser and OS as well? Labs team looks through the reports often.
How scnee builder works
i even dont find this
how extend scene
old UI was working new i dont find ?
I'll translate it and send it to you since I'm in Korea
The site itself is not translated separately, so I will translate it into text.
I have successfully generated a video using the Flow feature, but the 'Add to Scene' (장면에 추가) button consistently fails. It keeps triggering a popup: 'Error: Something went wrong.' Please help resolve this as I've already paid for the subscription.
I am a PRO user and the 'Add to Scene' button is not working. It says 'Something went wrong' every time. I have already tried switching accounts and testing on different devices (PC and Laptop), but nothing works.
Reptilian shapeshifter. I knew it google.
I think that stuff can be somewhat dealt with with Fixes: No strabismus, no uncanny valley effect.
Then again VEO works a little different than Nano, but as long as it's just one person I'd think it'd get it
Or two people without reference image, it sure showed me a lot of probs to follow two characters at once yesterday
I’ve been WAITING for Aspect Ratios! Thank you so much, Google Flow. What’s funny is I’ve been having to use OpenArt simultaneously because they have Aspect Ratios…I literally had to resub to them last night because I needed 1:1 for Etsy. I wish I had checked Flow first!
... yeah putting requested features up is something that causes people to sub things.
Whodathunk it
Prompt: A tight, low-angle cinematic portrait shot from behind the right shoulder of [Insert Character/Animal], wearing the sophisticated, hydro-morphic segmented combat armor with its highly detailed, scuffed texture. The focus is razor-sharp on the [Insert Character/Animal]'s right shoulder pauldron, where the custom callsign, [Callsign], is deeply etched into the scuffed matte-black metal. The focus extends to the subject's specialized helmet sensory assembly, tailored specifically for [Insert Character/Animal] anatomy, featuring integrated haptic sensory conduits and intricate, pressure-sealed membranes. The environment has opened up dramatically: the background is dominated by the broken, skeletal wreckage of a massive commercial freighter that has crash-landed onto a lunar outpost. Its colossal fuel tanks, engines, and cargo holds are twisted metal, spilling a vast debris field of broken structural beams, cargo containers, frozen propellant lines, and shattered electronics across the jagged surface. A specialized multi-tool tactical axe is held defensively. The air is thick with a green-tinged fog that drifts through the entire wrecked site. Harsh, powerful rim-lighting from emergency flares and the freighter's dying reactor core illuminates the scene, casting dramatic shadows and creating a massive, piercing blue lens flare that washes over the foreground. A horde of infected cosmonauts in torn suits stumbles through the green fog, navigating the treacherous debris. Shot on RED V-Raptor, 50mm, f/2.0, hyper-realistic detail and scale.
FYI: If you want 100% of the prompt make sure to insert your chosen animal twice -> [Insert Character/Animal]
&
Your callsign/name once -> [Callsign].
#4:3
can we get the scene to download at 1080 stil or have they stopped that now?
People, would you rather use the 4k version of the veo 3 video or upscale it yourself with other Ai tools? do you prefer 1080p over 4k?
Honestly, the 4K from VEO is better quality than the tool I use to upscale
Anyone else consistently getting you're requesting images too quickly, even though it's been a while?
I know I've put the request in before, and I was told "the request has been sent" but am I the only one that spell check with the red squiggly underline does NOT work in my prompt box? "Spellcheck" works in every single program I own...ALL of them, even the latest Chrome and all the chrome settings tick boxes for English language.
It just does not work in the Flow Image generator prompt box. It even works in all the other Google labs products and in Discord. 1 misspelled word and I have to regenerate till I can figure out why it couldn't understand my prompt. IDK.
Please give me your feedback. Is it that difficult to program Flow for this? Clearly it's an issue if the Lab admin told me they have been sent the request. Does spellcheck work for you? Any suggestions?
Does the Ultra plan have NBP limits?
anyone else having issues logging in to use Flow?
is flow nano banana more censured than the gemini apps?
Youtube channel views
That more often means your account either ran out of Nano Banana2/Pro images for today or the account is being limited due to requesting too many images at once (more than 10 or 12), or at worse if you haven't used the account for a couple days and it got limited at random maybe cause Flow is at peak hours at the time. It'll usually be fine in 24 hour or less, if after 5 minutes the bug isn't gone yeah more likely get ready to wait some hours on it.
Flow used to display a more related message alike "You ran out of Nano Banana 2 images for the day come back tomorrow 🍌 " for this, but I haven't seen it in like 4 days anymore, so yeah that's a bug.
Yeah I've also tried various upscalers and Flow's keeps more small details for some reason, happens to the Nano images too. I got a big enough monitor have zoomed and done comparisons and yeah, it's worth the wait.
Even with Ultra? It’s been like that for past couple hours
I've seen this too, but the built in for chrome seems to work. If not you could paste the prompt at a social media account where the spell check works, or at PCs notepad, it'd work there for sure.
If you got Ultra and happened to you yeah more likely. But if you've only made a few images and got it I guess you gotta mention it at the labs bugs feedback. My $5 account was good and fine for the morning.
I'd say it's much less at Flow than at Gemini, WAY less. I've used the Gemini one a lot too. However you can get luckier at Flow with the images since you can generate like 8-10 at once and something might pass. However gotta be careful not to go faster than that, accounts seem to be getting limited for it atm.
I've thought about prompting in another program but that's a lot of copy/paste. I'm working on an illustrated movie with over 10,000 (usable) images and I'd rather have Flow correct me than use another goto. Thanks, MG.
Maybe you can find a browser extension or something for it.
Damn, I have created a few, but didn't realise Ultra would have a limit, the amount is paid for it!
tbh with Ultra plan you should have been there all day, but Flow seems buggy about it atm, you're gonna have to report it. Then again the rate limit does seem to be affecting everyone if they overdo it.
Thanks google! I fulfilled my first dream, I still need to buy a guitar. Maybe I'll be able to in 10 years.
Nice guitart.
Thanks, I generated a few, but nowhere near as many as I have in the past. I'll give it a bit
i dont understan why ai studio charges for nano banana 2 since flow and gemini use freely
When something is free, you are the product. Just wait for it.
Also that place still has possibly the most viable way to use original Nano Banana, which is what Whisk uses for precise mode... so yeah. Tho I don't think they might have noticed too much yet what they have there atm.
Also it could be that sometimes putting something up for free or very cheap even at your own cost is a way to dent on the competition. War times strategy. Or it's both things.
Oh yeah, also the "You're requesting generations too quickly. Please wait a moment and try again." message bug is surely affecting Nano Banana Pro more than Nano Banana 2, try swapping models and at least that one might work.
Good idea, I'll try the NB2
NB2 appears to be working
...dam that just shows more people are using NB pro since so many are having the prob.
Yeah I do prefer NBP
Hi I have issue when create sence from flow, is there anyone have same error {
"error": {
"json": {
"message": "Internal Server Error",
"code": -32603,
"data": {
"code": "INTERNAL_SERVER_ERROR",
"httpStatus": 500,
"path": "project.createScene",
"zodError": null
},
"status": 500
}
}
}
flow server error plz....
Is anyone else having issues with upscaling and downloading today? Sometimes it downloads, sometimes it doesnt.
I get the same error when trying to add videos to scenes. Most videos also aren't generating; they reach 99% and then hang for several minutes and fail with a generic error. Seems there's some issue with Google's systems, or they're just overloaded today.
PUBLIC_ERROR_VIDEO_GENERATION_TIMED_OUT ????
daffuuqq cant generate anything
hello. you asked me for an example of where flow is worse than the Gemini app in image generation for the same models.
back then I only had private images so i couldn't post this to you, but now I'm working on a game now and preparing assets for it and i have an example that i can post
as you can see nanobanana2pro in flow added a line in the middle that didn't exist
and is less accurate
generally also when generating images of people of various positions in anime style, flow is more likely to miss a leg or a limb and to have less spatial understandin
and more likely to reverse the order of toes
I’m having the same issue—videos get stuck at 99% then fail with PUBLIC_ERROR_VIDEO_GENERATION_TIMED_OUT. Anyone else seeing this or know a fix?
having the same
are more people having it?
maybe a temporary issue on their end
Same here.
yep, seems like a temporary issue Flow side. EU servers here.
I just think their servers are being very slow for a reason, hits a timeout error internally
are u guys using low priority mode
There is currently a flow server error, right?
How do I fix the issue with Google Flow Labs not generating images?
Gemini just has free aspect ratio, or it can be prompted to be the exact same ratio from reference image. Even if in both cases that image seems to be around 16:9, Gemini will struggle less to fill it since it has total freedom.
And since I don't think you prompted it to use ratio there yeah it just went by perspectives and the other prompts
Flow on the other hand has locked ratios so it will try to force elements to fill them
Also about Flow's current errors, It's just not giving the precise error messages, in some cases it could dbe you just be you ran out of image generations for the day and it's not giving you the correct "You ran out of Nano Banana image generations for the day come back tomorrow 🍌 ", instead displays it as if you're generating too fast or an "oops" message.
But in some other cases the account is indeed being limited for not much reason, specially if you were doing nothing with it during the day, that could be some error/bug during peak hours. More likely in both cases the account will be fine again in like 24 hours, but you might still wanna slow down.
gemini seems to be good at following aspect ratios aswell when mentioned
24 hours passes way slower than one would think, tho sometimes seems to be just 12 or so too for it to work again
Yeah whenever you can end your prompt like
and the problem here is much more than just 'perspective' or 'ratio' problem. it added a split in the middle
But yeah I think for what you're doing it's fine as it is
i really don't care about exact aspect ratio. iw ant the image base to be correct and then i can use any cheap image model to get the image in the aspect i need
Yeah ratio also counts for the camera composition and how it will be filled. The thing either has to guess how to do it or try to fill the empty elements moving them where it can't calculate
I've also struggled with that a lot with fullbody images of characters with weapons, sometimes they go out of the image's borders
Then noticed it was ratio problems
Nano Bannana pro on Flow not working
Did you even manage to generate some images today? is your account free or pro?
pro
Or ultra whatever
Yeah then it's possible you just ran out of image generations for today and it's displaying the wrong error message, I think it should be fine again around the same hour when you started making images.
hmm thats weird
That is indeed a bug tho, Flow's UI's messages should be explaining what's really going on for each case
It's been going on for 3 days now, but I've seen accounts do mostly reset themselves back to normal in 12-24 hours. Anyway Google staff at labs bugs and feedback did say they're fixing it, but we don't know when it'll be.
can anyone generate videos with VEO? getting failed generations with every model and type
Using Flow, I've had a boatload of misfires over the last few days. Just wanted to say hi and let you know what's also happening to those of us with the Ultra account. I've tried fast and fast lower priority and a few of them I've had to do 3 or 4 times to get it to stick. It mostly happens on extend. The initial ones I'm generating 2 videos and in most cases this morning 1 of them has failed. So, scrolling through, thank you to who posted about them being aware of it, so hopefully it'll work itself out today or tomorrow. And I love everything about this lab.
Wait a few hours and try again.
ironically I don't think Flow is being able to take the same amount of punishment Whisk takes
Which maybe could have to do with how Whisk's original Nano Banana would use less resources yet mostly give people what they want, but what do I know about how that works.
I think I've only seen Whisk go on maintenance once during all the time I used it.
There's a reason the Nano models are called "Nano". Doesn't seem Nano 2 is so "Nano" anymore.
Not to mention people seem to have to struggle way more with Nano 2 to get what they want, which is obviously gonna cause a lot of wasted resources.
how can i generate the ai videos?
When will this feature error be fixed?
The "Add to Scene" function is an error
It's not just me. I think the server itself is the problem
What's up with this google... no generation and it still remove the credits??? Since the latest update I have nothing but issues with Flow!!!
so who’s gonna tell him..
Anyways ! Good morafterevening the netizen of labs animals
you during your micro analyzing about whisk and hallucinated stuff about it
So anyway,
idk just trying to find an explanation on why Flow has been derping during the past days, and Nano 2 does indeed take more time to generate each image... while Whisk's Nano never had those problems pretty much spammed with users
It could also be aliens or anything else tho, so yeah you're right about that one
Then again it seems it's not just the Nanos and the website, it's also veo failing
Yeah i had a weird output last night testing veo specifically as well. Such a odd night.
I've seen that kind of stuff happening at other AI web sites and it usually points to overload, peak hours and server issues
Sounds like common talk, but yea best tend to it early.
So, is this a bit cumbersome for anyone else besides me when making images that the dismiss box covers up the menu for when you want to say download it while you're still creating content? Like, I get it that it's nice because it shows you the work is being created, but maybe put it in the middle or put it somewhere where it's not going to interfere with the 3 dots to enable what you want to do next with the image or video.
I see what you mean here. Sometimes theres a whole string of those upscaling messages going down in that area. If i can remember if you dismiss it to early it'll cancel the download as well.
A container you can minimize a bit is a decent start, nice eye here.
Thanks and actually I've discovered that if you dismiss it early it will not cancel the download as long as you stay on that page but let's say you do a video and you click over to the extend page then it's possible that that one won't save but I've had them show up after that when I go back to it because I forgot that it was running in the background because I needed to close all those boxes to download the one that it was covered. But the last few days I've just stayed on the page because all the let's just call it stuff happening. Thanks again. I appreciate your comment
If you're talking about using their login information, I don't know that I would recommend them doing that because there's no guarantee that the person you're allowing as a member wouldn't violate terms and then get all of the owner's Google taken away. It's just too much risk. I think a better ask would be if you can work out with someone that has an ultra account that perhaps has extra credits towards the end of their cycle and then you just send the prompts over to them and they generate them and upload them to a Google Drive or something. I guess you could do that. Good luck, though.
Are you on chrome, IOS, or ??
I couldve sworn there was a discrepancy on my end, but I usually switch from Safari to Chrome, when it acts up or there's a lot of traffic.
Ahh i see, yeah if there's "traffic spikes", as i refer to them as, just try it on a different browser on your browser list.
Either way nice mentions here.
I cannot get anything to generate, regardless of prompt. Instant error message. Cleared cookies, logged in and out. Any tips? Been going on for at least two days
Same issue in AI studio now
halo-cinating huh.
Just started getting that message myself. Haven’t the last two days.
A lighter website with a more cost efficient image model wouldn't have had this prob, whodathunk it.
Why do all the videos I create now have a watermark?
i wonder why
It's just going through some peak use atm. idk how it really works but seems Flow's been biting a little more than it can chew in the past week.
there's some possibility that they downgraded the server capacity due to the high-cost computing power.
ai bubble incoming
Yeah that's what I was saying, but you said it better
Also if it's somehow linked to Nano 2's usage, there goes another reason to just re-enable Nano 1 as introductory model as Google AI studio does.
Feedback only seems to be making Nano 2 worse anyway, or so it seems from what new people that join the Discord keep saying about it.
Whisk would rarely go through something like this.
/imagen prompt: Cinematic timelapse, 9:16 vertical, minimalist kitchen renovation begins with an unfinished subfloor and empty modern space, morning light entering through large windows, subtle construction setup, clean architectural framing, ultra realistic, smooth motion
Just got it now too.
@tough yacht Could you please check it for me?
more than 24 hours and still cant generate any video. shame on you google
Network issue
I also face that
Handheld, slightly shaky footage on a research vessel above the North Atlantic. Cold gray lighting, overcast sky, and rough ocean waves. A marine researcher kneels at the edge of the boat, carefully holding a live stingray partially submerged in a container of seawater. Close-up macro details: wet skin texture, water droplets on the lens, gloved hands tightening a compact GoPro-style harness onto the stingray’s upper body. Audible clicks as the mount locks into place. Wind howls softly, ropes creak, and waves slap the hull. In the background, equipment monitors flicker. The researcher lowers the stingray into the ocean—camera jolts slightly as the animal slips free and disappears beneath the surface.
first i put the the scipt into the flow app , but it only gave a picture script
what do i do now to make the script into video
can someone help me out
Screenshot the issue pls then we will help
How to stop the people from singing when i tell them to SPEAK
how do I fix this?
its a dog
?
Network issue
Go to video section click on start fram and select your generated pic then same with end frame then booom
I changed and still not working