#flow-general・🌊
1 messages · Page 3 of 1
Local run models fry GPUs, give you headaches (cooling noise) and still have a hard matching the same level of realism and speed offered by Imagen.
You got vertex?
@woven jasper they gave it out to discord USA denizens over 18 years old a coupel days ago. If you don't have it, people do volunteer to run prompts in #flow-tips-n-tricks・✏️
Hey just wanted to ask
Whats the best free website or service for adding audio to generated videos?
I use both MMAudio from HuggingFace and ElevenLabs' tool, which Hellmaker linked for you. Although, I find MMAudio to be much higher quality and more accurate from my experience!
You're welcome
gpus very rarely get fried doing this stuff, gpus tended to get fried from mining 24 hours a day for 3 years without a break, thats usually what it takes
i can access this page, what does it mean? what does vertex let you do?
Amazing
but we can already run veo, is there other stuff it allows you to do?
hey guys what is the date/year of knowledge cut off the imagen3 and veo2 is based on?
Does still Mitte.ai got img2video with Veo 2? Looks like Google has stopped all API partners from using image to video for now
i'm not really making a point i'm just unclear what it is, how you interface with it, what it can do, if I go to a dry cleaners i know i'm get my suit cleaned i'm just unclear what this is!
dumb question then, say i spent four hundred thousand dollars on a H200 supercluster thing, does this allow me to run veo locally on that cluster via api and license and whatever or does it only allow you to use googles servers
k
Me waiting for Paligemma to be released
how many totald o you think
I keep a tab for my personal projects, a tab for research and development,, and a tab for posts here. I still haven't hit the limit. I also precision prompt.
i got this big boxing fight thing going on, lil teaser
nah just focussing
i got the ending scenes done
just gotta add a post-fight press conference
🤣
Yeah, personal project has 200 words that keep it to the tone.... Research I copy-paste carefully... Junk I wipe and freestyle. I don't run them simultaneosuly but I do move freely between them throughout the day, haven't the limit.
I also do bad with the bounce which is why each has their own little spot in a row in the ai games desktop;
I'm the only one who doesn't have access to veo 2, i hope i can accept in waitlist too
Well that's your very first message on the server. You can still use imagefx/whisk/musicfx/etc. and there's #flow-tips-n-tricks・✏️ channel to request prompts.
Can someone please try to write a prompt for a recreation of this video? https://youtu.be/0Gj1QaqDk2Q?si=Hkaep73mEZb-DKU4
I only want the collision of the car and the utility post.
Title :
Steven Rice almost died from a terrible motorcycle accident that happened in Marlton, NJ
Description =
The results of the Near death experience :
Thankfully, I was immediately knocked unconscious, so I didn't feel any pain. But, I was gasping for air until the paramedics showed up and worked their magic on me. I was flown in a medical ...
I'm honestly A BIT worried that the IP lawyers will start sending cease and desist and they'll lobotomize the characters away
this happened but eventually got restored
yeah time went weird
hunt for gollum is a ok name for a fan film, but I don’t really care to see a film about gollum
not after the game
Video Fx refuses to generate video for this prompt... Generate a close-up of a framed, faded family photo placed on a café counter. The photo should depict an African American woman with her young son in a warm, candid moment. Overlay subtle text "In memory of Miles" in an elegant, respectful font. The image should evoke a sense of cherished memories and gentle sorrow..... - - - is my prompt hitting any inappropriate boundaries?
Try this prompt
close-up shot of a framed, faded family photo on a café counter. There is an African American woman with her young son in a warm, candid moment in the photo. The subtle text "In memory of Miles" is overlayed in an elegant font.
Imagine I took the time to splice those together
"Gollum hats the video game with his namesake."gollum yeets his old TV into lava "
This still doesn't generate the video. So odd.
Hmmm, maybe;
a framed photo on a cafe counter. There is an African American woman with her son in the photo. The subtle text "In memory of Miles" is overlayed in an elegant font.
I was able to get it to 90%. I blame the sicko filter. I have a refreshed prompt slowly ticking up now.
Yah, it's dying on the content review step.. one just randomly snuck out.
@old mesa how's that
Scroll up, Max is asking for it.
This is more like I wanted it but the spelling is bad, lol.
Spelling costs extra.
Yes this better. Thanks hommie
What are you doing with it
I wonder why is this prompt, which seems fairly benign getting blocked. Is race like 'African American' a keyphrase that gets flagged?
Nope. I'm guessing it's the kiddie filter getting pinged by accident.
Once I made it a matron and college student, it popped.
Even though they look kinda like a couple in one of them
ah, that makes sense.
Yep.
Here's one of my examples
I made the people the last detail in remembering miles and first in In Memory of Miles.
Another...
I had to settle for my safety school institute of Galactic Convergence for the better financial aid.
It fell on hard times and really prioritized the affordability
reptilians were the last to get that MIB money.
Well Subsidized, especially for the hatchlings of professors from sister schools.
The tails were blown way out of proportion. Only for a certain subpopulation of the work-study program. we signed a waiver and most fell off before graduation.
Whisk can generate at the speed of comedy, but Videofx really leans into staggered post timing.
number 4 finally popped and other than the tail being on the back, it had the most "Dang. Huh." review of paperwork after the fact.
I didn't save it, sorry, but it was txt2img2video.
Thank you!
Ah, I don't really care about followers on Twitter or YouTube, I just sometimes post there when I have the time. if something comes out of it, then good, but if not, I won't really grind to try as I have other things to do.
I could not imagine living by "I need X amount of content this week to hit Y engagement or the computer code takes away my ability to pay for things."
They may have been answered -- for those who use MMAudio - :08 max or can you do a whole 2 minutes?
Yeah it's much less stress to just do a few big projects to then build a brand for IRL work and projects.
My last project unintentionally became a series of daily postings and it was so stressful. Never doing that again.
How do I get access to Veo 2? I’ve applied several times to the waitlist with no luck
Last week they opened it to discord,18+ USA, for a day. #flow-tips-n-tricks・✏️ has volunteers that will run your text prompts for you in there. @crimson pike
Yes, I have the same issue.
I am wondering if there is a solution!!
Adding @main saddle
hi all!! thanks for bringing this to our attention — can I ask for either of y’all to share this in #1319424187730755636 — don’t want to lose track of this as we look into it!
Yup happening to me too
I logged it to the bugs channel https://discord.com/channels/1172568727942860810/1346259391220809829
Can someone please try to write me a prompt for a video just like this? https://youtu.be/MVIJcm8_pRo?si=kAzX1iEsIaaCzXLx
i got the prompt for it but it's a bit difficult for veo to do and i'm sure it goes the same for other video generators
it won't let me send the prompt since it's a bit explicit?
Just started working for me again, randomly
I think it's working now, I'm generating some videos to check.
let me know if you see this again! the problem should be resolved
Thank you so much, it is resolved!
@near gate I appreciate that and I will try to use it.
np
did anyone else notice the cool color effect disappear?
mine still does it
i mean, i don't mind, just was wondering!
Mine disappeared too
does this mean you can loop and play back fullscreen again?
umm the design doesn't look as cool and kinda boring - but I'm not complaining! at least I have Veo 2 access!
dont really care about the design as much as being able to play back fullscreen and loop the videos again
yes you can do that
whats the best lip sync tool to use with veo gens
its funny how they've reverted it for some ppl, are they A/B testing or something?
definitely a/b testing
thank you for bringing back the full screen button !
and also the video generated are now played in loop .
are the generated videos 480p or 720p?
It’s always 720p
It’s the colour
Of the helmet
crazy, almost like it’s going to keep happening to everyone’s prompts because it’s weird to censor
has anyone managed to get veo to more reliably replicate specific focal lengths yet, specifically long/telephoto lenses, it seems to struggle with that and i havent yet found the magic words to get it to work more often then not
@near gate Nice try. But, it's not exactly what I'm looking for. First thing, I should have stated earlier, is that I am hoping for a real life re-creation of the animated video.Thanks anyways.
And, I'm hoping that maybe an API is in the works for the Veo 2. 👍
Yeah I was about to say that, it’s api is already accessible
Also TIL veo 2 api has specific parameter called person_generation
Also is prompt enhancement enabled by default in VideoFx ?
@edgy holly I'm hoping to get a free Veo 2 API.
I'm getting the 500 error as well. Signed out and back in, just to see if that would work, but nope!
I wonder when aspect ratio features will be available.. only reason I can't use for a purpose is this. I want create videos for english teaching. ✨
it glitched some computers
i still have it
Thats interesting. No Fullscreen toggle here. Also the function of Files getting renamed by default with a timestamp is gone again. So yes, there seems to be some A/B testing. Hopfefully its going to be rolled out for everybody.
especially the timestamp filenames were very useful. Now back to video_fx_videofx_video (1).mp4 and counting (x).
Hi Tim can you try visiting (https://labs.google/fx/aitk-web/videofx-tt/tools/video-fx) directly?
So FYI we have switched the url for VideoFX TT'ing. If you are part of the EA group you will now be redirected to the new URL seen above
This should allow us to send out updates for yall alot sooner and more often!
if anyone has problems please let us know!
More frequent updates you say 👀 I'm excited
Anyone else getting this?
Just came here to ask the same!
Have you seen this message? @upbeat rivet
What message?
Yep.
Even with the new link you have this error?
Seems to work!! Thank oyu!! And EXCITED about the updates as well!
hopefully our videos will be saved in the "my library" soon..
Sorry, please visit: https://labs.google/fx/aitk-web/videofx-tt/tools/video-fx directly
Ok it seems okay. I just pin/installed to home screen.
Any advice on when the interface is in the way of the videos on mobile?
there is not mobile UI. Please use things on desktop!
hopefully we can implement some form of mobile in the near future
Swapping to desktop view usually works. I've been getting the thumbs up/down and the bottom overlay in the way a lot when it shifts from portrait to landscape.
Before I just kept it porttait - textbox is more responsive. Videos full screen landscape now instead of playing in the box.
i hope our past videos arent entirely gone and just stored somewhere or it would be just disappointing.. i didn't get to save some of my prompts so that would be really helpful
ooh just realized it says WIP
hopefullyyy
There's still an "app install" and it does load a little smoother but there's no download button on the output .
@lusty nymph If you are talking about this link, it works fine : https://labs.google/fx/aitk-web/videofx-tt/tools/video-fx
But, now, the original page is working fine. 👍
yes we just fixed it! thank you everyone who let us know about the problem and sorry for any inconvinence!
How can I Access veo 2, because I joined the waitlist two times in December 2024
Does anyone else get the issue where the percentage disappears and the generations are jsut blank and fail or just take 20 min
What's that? How do you get it back to the way it was? I can't open the sheet where you have to write promt to create a video in veo 2.
yes! 😦
this could be a result of your prompt being invalid.
sorry we dont have a mobile UI yet. Please use the app on Desktop for the best experience!
ok thank you
am i right in assuming that sometimes it doesnt go off of the prompt but the generations, like if it sees something against guidelines it fail at the end
When will it be available on the phone ? Cause I'm just away from my computer.
I don't know, but In an ideal world no sooner then a few months 😢
@verbal hatch Firefox mobile, set to desktop view, just takes a little pinching and panning but functionality works pretty well.
They recently did another videofx wave couple days ago but they closed it after a few hours
(Wave as in accepting users in the new form)
@main saddle can I get access, I'm happy to expand but am worth the seta
Feel free to head over to #flow-tips-n-tricks・✏️
Some people are willing to do your prompts there
or you could use veo 2 on youtube, (which is locked in an aspect ratio)
Veo 2 YouTube app doesn’t follow the prompt :/
Aw did you guys remove the full screen button, but at least it stays playing in loop now.
Excited to see multiple updates today. Or new A/B Testing? Videos are looping in playback now, and I just finally got the Timestamps back in the downloaded files 🔥
Thanks!
ooh
@hexed marsh can we commission a discord bot scans for posts from users with less than twenty posts saying " when, waited, videofx" and auto replies the things we have to keep repeating all the time?
The chain rattling ghosts are still here, evaporating back into mist after uttering their groans and wails.
unfortunately, to even add ANY discord bot here requires approval from the staff, and i'm not sure how they would feel about it
because at that point i would love to add a bot that can make Veo 2 prompts for us
based on our lovely Gemini API ofc 🙂
From the moment I made camerahead that was the dream
Workaholic brain: "well, I could script up a screen reader with the channel up that macro buttons a blogspot article link. Do people say the word months any other time? "
i just noticed fal.ai website mentioned veo can produxe 4k quality
either misleading or idk
Could just run the footage throguh ESRGAN or similar and share that
921,600 pxiels is an easier load to amake at first than 8,300,000 million pixels
Especially when ther'es plenty of solutions to turn 921k into 8.3mill
The colour opposite for white
they censored since someone could use it in a racial slur
b lack works too
For future reference
I've noticed the ai follows its own censorship logic - specific combinations of words trigger a filter
"young catgirl" and "circus"
cant u just use word female
young female? I'll try
now I'm seeing what happens with "a lil' medieval peasant girl, wearing a pink dress, medieval, brown hair, pale. She holds a pendant made of bone, in the shape of a knight."
3 out of 4 got censored, and that ain't no little girl
mmmh. starting to become an issue
that pendant's terrifying lol
hhhhrr. 4 out of 4 got blocked
well, I'll move to another subject, but I highly suggest the devs look into this issue. Other models can do my prompts when I ask for young characters
Why don't you use them? Civil has literal thousands of options.
Anime seems to look similar between models to me. Highly stylized movement that allows large portions of the screen to remain still. How is veo better? Live action is obvious.
I also haven't actively gone looking either.
mmmh. Horror movie trailer maybe?
If you don’t have access to Veo2 and want to try it, you now have a chance to get two free generations. Simply search for the Imagine App handle on X, follow them, like, repost and comment under this post; https://x.com/imagineappco/status/1896968550201659821?t=55pjvRx1Z0Mh5Njxf1O8wA&s=19. Then copy the link to your comment and send it to this email mgyong@imagineapp.co and they will credit your account with 700 tokens, allowing you to generate two videos. It works. I was able to generate two videos.
@Google Veo2 is live on https://t.co/PhO4Uboby8 Check out this collection of cool Veo2 videos created by our users https://t.co/JZAd1ToVTn
Or even simple, stop by #flow-tips-n-tricks・✏️ and politely share your prompt.
No reposting and comments for SEO, no handing out your email address to be resold when a company folds.
Let me ask you a simple question: is it against the server rules? Yes or no?
Promoting your toke.n business? Don't know. Snake oil salesmen always roll into communities looking to hustle.
And hey folks - mention coupon code "people before profit " and I will personally do FOUR videos for you. Double the value.
Unless 2 videos literally means just 2. Never touch tokens myself. But In which case I'm offering eight times the value for the low low price of being nice.
It's not my business, I don't own the company. I just benefited from it and thought people who want to try it should go ahead. I'm not promoting anything. It seems like you can't read.
need reference to video so much 😭
Well, I'm sorry to hear you were exploited.
Faking engagement with a repost/comment requirement and the handing out email requirement? Scam scum tactics.
Which is why I'm offering the lowk low price of nothing . I'm giving it away!
How do we do it. Our corporate overlords are using us for their own stress testing. So act fast because it can all be gone to.orrow.
Nothing is certain and we inch closer to death every second. Operators are not standing by.
So did it give you 2 videos total or two sets of four for i8 total?
I came to the conclusion that every time I try a postapocalyptic setting, veo2 wants it to be a cgi production
How is it exploitation? Explain. This isn’t just about paying for something. There’s nothing wrong with liking a post and commenting on it. People do that all the time without expecting anything in return. So what’s the problem with liking posts in exchange for coins to generate videos—something many people might want?
yo guys bring it to off topic
The only reason you’re saying all this is because you got access recently. I was already wondering—seeing the way you make noise and try to police everyone in this group—what it would be like if you actually got access. Back then, you didn’t even have it, yet you were still troubling everybody, trying to dictate how people should act. And now, here we are.
White rappers of yesteryear already explained luckily.
My dude. I do have it.
A lot of us got it last week. Where you been?
If you'd like to attack me personally more, take it to #off-topic・🐢
I will remind you of Rule 1 though
Nobody is attacking you, dude. This isn’t what an attack looks like. I’m just stating the facts—you make a lot of noise here. Chill, okay?
I'd you want live action, you can throw in a bunch of Mad Max related terms. Tank Girl will get you comic book illustrated.
need moar character consistency
In between takes or in the same 8 second clip?
It's basically like talking to Dory. Repeating yourself a lot helps compensate for the poor memory and distractible disposition. .
between takes
reference to video would be so good here
you saw my bl4ck tiger trailer? was so hard to make the dude look -more or less- the same
(the fact that the word bl4ck is considered offensive is racist on so many levels, lol)
B lack works. There's a suggestion box thread where you can join others in the automod dismay.
still even weirder than removing children from veo2
at least there's a distorted logic in it, but what's the logic in censoring a colour
I don't recall if I did.
I have 20 minutes for a project and have one kid really consistent and one old man.
https://www.youtube.com/watch?v=YuXLULsjLqA&feature=youtu.be this one, for reference
The movie of the videogame Black Tiger - thanks to AI.
Oh right, I recall the logo.@twin palm is good with muscles and their shadowrun people are consistent. They could probably help with the barbarian consistency.
hey
"A quirky, goodhearted, charming, middle-aged Taoist monk with a broad, corpulent physique, sitting serenely in a modest temple courtyard. He has a thick, short puffed beard, bushy eyebrows, and a piercing, wise gaze that contrasts with his amiable, slightly goofy smile. His robes are simple and a bit worn, flowing around his large frame, giving him an air of understated authority. The background shows a poor Chinese small house. next to him, a monstruous female creature, a yokai, with sharklike mouth, glowing big eyes, bone spikes protruding and pulsing from her skin, sharp oversized claws."
why does this get censored, now
just WHY
omg I can't find what words trigger the filter in this innocent prompt
maybe it was "poor chinese"
I can't even
Try woman instead of female
I'd look at the rules Georgia posted again. I made my nieces and nephews in imagefx fine.
imagen isnt veo2
Veo also sends them running around..
Do read the guidelines though. The line between censorship and responsible restriction is drawn pretty cleanly.
How do I insert the promt into veo 2 now ? Why did you do that before everything was normal and worked fine, but no, you have to mess it all up
I used to be able to open the Promt input line and it was on my phone and create a video, but now it looks like this.
Scroll up, and you'll see an engineer was love responding to issues with updates and gave us a separate link.
Coming in hot and accusatory isn't constructive.
You'll also see I offer a workaround. I think their Google collars shock them if they say don't use Chrome.
Labs means experimental features that Google engineers are exploring. Messing it up is part of improving it. We are being charged nothing.
In chrome exactly the same situation and in any other browser is the same and I also think that the quality has become worse or I just think so
... Tell me. What did I suggest?
I don't know what Google has put in veo 2 and I can't judge for sure, but I don't think it's anything good.
Okay so, you touch your finger to the screen..and then you pull it down to scroll up into the past history of things said here. Flick your finger and it'll go faster. Yesterday afternoon.
Green name is for Google employees .
https://youtu.be/KPthUyTOws4?si=fYC_0OZCOrXqNWgg here is a video I made right now to verify it does work on mobile
They even fixed the portrait to landscape to portrait problem I reported. And added a new icon to each video so can tell which one you haven't watched yet. I'd wager that new overlay is part of the chrome canvasing problem, but that's idle speculation on my part.
this is just the nature of early access / beta software. we should not be under the assumption that it will be 100% stable at your convenience.
kindly share your feedback instead of throwing in the towel. let's be a bit more grateful for the work they've done to actually even provide us these products.
they also mentioned the mobile UI is not officially supported yet. but there are definitely workarounds you can try. hence what gogo just sent
Here.is a video demonstrating what scrolling backwards in a channel looks like. You can see it took less than ten seconds to get to the relevant section. Much faster than angry typing.
frustration sometimes overpowers logical reasoning. it's almost inevitable we will see similar types of behaviors with products like this
regardless if it's clearly stated as BETA / early access
It's a systems problem..onboarding expectations is hard, especially when discord doesn't have a good LET ME SLAP YOU IN THE FACE WITH THIS method after first login.
lool and they do. people just don't use them
maybe we need to start pinning msgs more
I have one server that has an exec brief channel where they play the latest updates to anything. .Another with a bot that will send you updates since you last emoted or chatted.
yeah a bot like that would definitely save some time
We have 16k and growing by a thousand more every week. I don't think there's much ven diagram overlap.between people that come in mad frame one and people that check pins.
lmao
Anyway, I like the Veo updates.
They basically resolved one of my petty grievance kn suggestion box within 18 hours.
Accumulative
to be fair i do too. but i do typically scroll up before i start chatting
Brief bot and/or #breakingnews
I feel like there has to be a way to use veo to illustrate this sentiment
that's deep
I don't like flooding the box but Iif either of you suggest it you have my lab cosign
a bot or two can definitely help as well
alt.religion.kibology ??
spoonfed information & entitlement to name a couple issues. only gets worse from here
Maybe tone down the attitude,
also i gathered the prompt taken from the "im feeling lucky" section for text to image to video for both step 1 and step 2.
im trying to study how they wrote the prompt for motion.
Oh good! I was about to get nostalgic for my dad's bulletin board system and Dr Sbaitso.
Did you find any trends ?
I have noticed they focus on a lot of distinctive vibrant activity
Where presumably the training data has baked in interest points
Acrobat footage etc.
General Observations Across All Examples for Text to Image to Video (Gemini generated analysis) :
-Two-Step Process: The prompts are clearly designed as a two-step process: first create a still image (Scene), then add motion to make it a video (Motion). This is a deliberate workflow.
-Detailed "Scene" Prompts: Step 1 "Scene" prompts are incredibly descriptive, painting a very vivid picture in words. They are almost like detailed scene directions for a photographer or illustrator.
-Focus on Camera and Cinematic Language: Both "Scene" and "Motion" prompts use cinematic terminology (cinematic shot, tracking shot, zoom in, anamorphic lens, film stock). This signals a desire for a visually rich, film-like aesthetic.
-Emphasis on Mood and Atmosphere: Words like "moody," "melancholic," "mysterious," "grumpy," "boredom," "lush," "nostalgic grandeur," "wonder," "awe" are prevalent. Mood is clearly a priority.
-Progression from Static to Dynamic: Step 1 is about establishing a static composition, Step 2 is about bringing it to life with movement.
-Building Upon Step 1 in Step 2: The "Motion" prompts directly refer back to and expand upon the elements established in the "Scene" prompts. They don't start from scratch but build upon the existing image.
-Implied Narrative/Mini-Story: Even in these short examples, there's a sense of a mini-narrative or a moment in a story being captured.
interesting how they even described the soundtrack despite it not even capable of producing sound
dude i can so use this for the prompt creator bot i'll be making. u mind?
that is VERY useful. thanks again
the google doc version https://docs.google.com/document/d/1mNNNu_omtgjmvuVBHlVAG6hDGUVuSjZbjXyyTjMtGpU/edit?usp=sharing
Okay, let's analyze these Google VideoFX text-to-image-to-video prompt examples. We'll break down each step and identify the key elements and patterns they use. General Observations Across All Examples: Two-Step Process: The prompts are clearly designed as a two-step process: first create a still...
i noticed the i'm feeling lucky section is a lot more repetitive than imagen 3... guess they are hardcoded into the software
its hard coded yeah
i still wonder if videofx has prompt enhancer enabled by default
Part of my hybrid utterances is to emulate SRT files and movie script drafts somewhere in the middle, between.sin curve and philosophy quotes.
cuz the veo 2 api did mention "prompt enhancer"
Would not surprise me. I have to smack it pretty good to get traction away from the surface crust.
ah yeah that's for the vertex API
i use it for imagen, surprisingly theres more customizability via the API
https://bert.org/2023/01/06/chatgpt-in-dr-sbaitso/. This but in the voice channel here as a true musicfxdj.and therapist
What kind of customizability?
similar to the ones in that picture just sent
including negative prompts
I would sacrifice ten thousand here if it meant negative prompts. At random with me in the pool even.
i didnt know imagen 3 can do customization like this
It's very strange, either I have something wrong with my settings or I don't know.
honestly i haven't really played around with whisk much
although i should start to mess around with it
its actually gonna be , whisk veo i speculated
hmmm
I use every part of the buffalo.
maybe you're right
even the bone?
Especially the bone.
I've accepted I'm here as a front end load tester but there's part of me that wants to start building my own. With blackjack.
backbone could be whisk, but then theres label for video 👀
Open. Animate. Mode.
they cookin many things for us behind the scenes 👀
i didn't even bother checking the source tbh
i only saw the POST payload, noticed theres even was a way to even change the seed already
and theyre already working on veo 2.1 👀
Okay,.I think my play is to stop being precious about getting my footage whittled down and just start posting "TEN MINUTES COLOR COMPOSITION TEST SUCCESS " "SIX MINUTES CONSISTENT CHARACTER"
And then post them the exact time bri and them are emote reacting as a work escape.
Yes and yes.
looks like its still making requests to veo 2.
so 2.1 still in the works
Probably internal for at least another month while they fix up the browser business.
notice the seed param? wish they gave us the ability to manipulate it
bc that def is functional
I have favorite Imagefx seeds.
How long until we're phased out to be just another customer do you reckon?
I think we will fall behind the curve instead of being straight booted. Enough third parties will be paying for the API that they won't need to give us the latest features.
we can tune our own version of gemini flash once we have enough good data for prompts
exactly
plus google always ends up releasing something that one ups the competition lol
yeah thats the type of approach we need to implement
And then a new team needs to get noticed so they axe an old project and rebuild it later
Did anyone ever get to try Veo 1.5?
applied never got accepted
this was maybe a year or more ago
Once I get that chuck e cheese money I'll bankroll us all
You get fnaf half the time
Might as well prompt Skibidi toilet
I am jet skiing consciousness after a long night of insomnia. They're not renewing contracts in June
It's a billion dollar franchise
You're the birthday boy or girl got the moose right
The movie alone made 297 million
Toys shirts and merchandise another half billion.
Built on the backs of 25 year olds hyperbolic shouting for the entertainment of 8 year olds.
None of it makes sense nor is it the Cool cyber dystopia I expected in 1999
Five nights at Freddy's . Where the creepy animatronics are from.
Good call to not have kids recently.
It started as a jump scare video game and the alphas took to it.
(wasn't even scary)
Have you seen the rest of existence though?
sorry someone had to say it
Wait until you're trapped driving across the state with your friends autistic kids who have read and possibly personally edited every word on the fandom wiki. The lore.
My friendly neighborhood is the sesame Street variabt. poppy playtime is the toy factory version.
They keep coming.
And I'm sure Microsoft's AI game maker is just churning them out.
A cenobite must have been in the trunk because it was kinda torture in that it was already a slow boil for the parents, so acclimated to tuning it out.
That day I learned my current annoyance tolerance dosage ends right around Hour 5.
the movie was made specifically for the fans so it’s normal to boo the movie
i personally like it
ah
Asked Gemini to analyze all the prompt of text to video examples from VideoFX (I’m feeling lucky prompt examples )
Yesterday everything was normal and I could create videos, but now I can't insert promt because the input line for promt doesn't open, so I just click there and it doesn't want to open.
huh ??
Ooo try it out
Where is this? Or a glitch
Sorry we have no mobile UI. Please use the app on desktop!
i got excited for a moment🙁
wait i just reloaded the page and it actually said "image to video" with the same description FOR ONLY JUST FOR A SECOND then it came back to the "text to image to video" broo💔💔
image to video actually coming soon??😳very excited
But yesterday everything worked fine and on the phone and I'm in another city and the computer at home
yes sorry, we updated things and now they work slightly worse on mobile 😦
We will hopefully fix this in the coming months!
Well, I'm going home soon anyway and I have a question, did you degrade the video quality in veo ?
We did not degrade video quality afaik
I wish veo2 was creative
The dataset is too strongly focused and creating something new or bizarre is impossible to do. Guess makes sense how can you create something if you never seen it. I was trying to do aone with a youtuber guy showing an alien PC that looked strange and bizarre but it couldn't understand that and just made normal PCs
"alien PC" can be interpreted in SO many different ways. even the brand alienware itself
did u try rephrasing the prompt and using different keywords
I probably could have tried different wording only now realized the issue after typing that lol
Im still hoping that someone can please write me a prompt for a real life video like this : https://youtu.be/MVIJcm8_pRo?si=XcvIucwRrDWq1-ko
I think that's a generation limit. I also have same on desktop. My access hasn't come back since yesterday -- It has not reset since midnight last night -- was generating a lot of videos apparently.
ohh. there was sign up for early access again that i missed. lol :P sadq
Oh the spectrum part wasn't the issue, or the game even really. Listening to five hours of anything by anyone is a lot. They weren't even repeating facts. It's just a fountain of fandom subculture.
how bout dis
u def just have to prompt it well
If you do a Google image search for desktop gaming computer, you get a desktop gaming computer.
You search "bizarre tech 1970s weird" you get a much wider variety.
That's not too far off to how image generators relate to the words we type.
Just yesterday necro asked if whisk could do Red XIII. Tried and it gave a generic lion.
It knows Red is connected to lions, but also here's the statistical reinforcement that lions look this way.
So I'd have to drop the heavily trained word and start describing around it to get where I want to be.
Anyone get double downloads when downloading video?
Why does veo2 always try to put weapons or anything on a character's back?
I don't have that problem. Name checking unreal may be part of it? Decades of shooting games.
This was the astro prisoners one?
imagine in the future, veo 2 will have a reference pic/video feature that would be awesome
aghh so when will we be able to upload our own images and turn them into videos though🥲🥲
🙏
You can put nothing on back and it works
What is aj mode?
oooo tell me more!!
oh google pls let us upload our own images🧎
very cute adorable goat☺️😊
has this been always here?
WAIT ITS HERE!!!
OH MY GOD !
Wait I can’t tap on settings
Ah I need to be on desktop
Actually I got it to work, I rotate my phone to landscape and then the setting button appear , then I turn back to portrait
Now let’s give this fast mode a try 👀
Holy macaroni it’s actually fast
And the quality of video still looks good!
👀 I wonder if the fast mode also understands prompt more better than the normal one
Uh help
I’m stuck on this page somehow
Nvm it’s back to normal now , idk what I tapped on
YOOO we got library for VideoFx now!
Oh wow it saved all the video even the one I accidentally refreshed on
Oh wow nice, just saw it too! Wondering how many thousands of gens are showing up in @edgy holly's feed
I thought you had early access for months! How's the prompting bot coming along?
Ah, I see. The hardest part!
Best way to do it! In any of your gens have you ever had issues with characters talking/mouth movements? tried a lot of prompt tricks to make it stop but no luck
In other models, had success with 'silent, thinking deeply, chasing x, unable to speak, listening' etc
Shared an example over in #flow-videos・📹
u can check in ur library there should be videofx sectiopon
fast mode apparently doesnt do proper natural hair physics
Just the start, for sure!
Good you're in that bubble (for now) , enjoy it! A lot of AI creators start getting really negative tweets (/threats) when the main stream picks up and puts you in an 'ai stealing our jobs/artwork' box
Pretty much found the same! Just timed them and 45 secs on fast mode 140s on normal
Is Veo2 fast lower in quality?
so apparently they enable prompt enhancer for text to image to video, specifically for step 2 motion prompt
Y'all find any correlation between seed numbers in imagefx and videofx?
I wanted to ask someone
we have to be our own mamas
just wanted to see how did they modify the prompt i wrote in the step 2
what i wrote
Like reading the comments above us not helpful. Wanted an answer from someone
same prompt, left default veo 2 model (140s gen time) right fast mode (45s)
and this was the final prompt it automatically output : A cinematic close-up shot of Sonic the Hedgehog, rendered in photorealistic detail, smoothly transitions from a running pose to a stop, turning his head towards the viewer with a playful expression. His blue fur is meticulously detailed, each strand seemingly alive with energy. The camera follows his movement, maintaining a steady focus on his expressive face as he slows his run and turns. The vibrant green grass and the iconic checkered landscape of Green Hill Zone form a blurred background, emphasizing Sonic's sharp features. The sun shines brightly, casting a warm glow on the scene. The camera angle is slightly low, creating a dynamic perspective that enhances the sense of motion and Sonic's playful energy. The overall mood is one of cheerful anticipation, as if Sonic is about to speak directly to the viewer. The motion is smooth and deliberate, highlighting the character's agility and charm.
.50 x (60x20)
if i wanted a 20 minute short film
it would be 600 dollars
uh what is negative prompting
thats possible?
Stable Diffusion in 2022 added it
my phone's notepad still has my $every bad phrase$ saved
I antonym'd that list and supertimes tripleup on it
No, I tried the same prompt and seed for an image on video fx and there was no correlation
A nice to have would be being able to bring over previously created imagefx images over to videofx
It’s so nice having the library now
Does it load faster ?
$0.50 per second of video
1 video = $4
10 minutes worth of video would be $300
wow. that's PER video. meaning every GENERATION we make on veo 2 would have theoretically costed us $16 because it returns 4 videos
"Capitalism is crazy."
i hope you guys add a ‘download all’ feature that allows users to download every video or image from our library in a single zip file, with the text prompt included as a .txt file (or an option to exclude the prompt if preferred), if that isn’t too much work. i’d be super grateful! also, folders and a ‘recently deleted’ section would be pretty cool too!
probably is too much of a request lol😅
We just got a Library feature!
#1238621511401734254 would be a better place for that. By tomorrow morning 50 other posts will have buried this.. I've requested a way to get prompts out of the library easier in the past but new voices echoing old concerns can be part of the change process.
thanks! will post my suggestion there🙂
That's awesome.
Does anyone know when more users being onboarded don’t want to have to keep using VEO via freepik!
Is Whisk Animate going to be like Veo2 on VideoFX—taking forever to be released?
apparently u can do it through youtube shorts. don't take my word for it though
The problem is that the YouTube Shorts version isn't as good as the VideoFX one. Also, its portrait aspect ratio doesn't fit well in some videos.
did u not register for access to whisk animate
Where can i?
I didn't. Y'all have access?
@dim glen it was in whisk announcement channel
and yes its closed now
i didnt get access cuz i didnt register for access
same
im content with my videofx access for now
It seems like these Google Lab opportunities just aren’t for me. I’m just not lucky. I keep losing opportunities—just the other day, I lost the opportunity to get access to Veo 2, and today, I realized I missed out on another chance. Honestly, this is really disappointing.
wait, what's whisk animate? tell me it's a reference to video tool
Just activate push notifications for announcements if you haven't already.
ah image2video
Yes.
I did. But I need to start paying attention very well.
Seems I can't make a shirtless 15 years old boy, but I can make him wearing a shirt. Yeah, I know: when AI censors something, anything, I feel the urge to hammer on that point
why do u need to speciffy the age,just use general term
I've been trying to make Lathander, the god of dawn in dnd
tbh makes sense it flags that type of prompt
and I added the shirtless part to intentionally provoke veo2 and see what happened
yeah.. like its the specific age
if u can make a shirtless 15 year old boy whats stopping u from making a shirtless 15 year old girl? thank god that model stops that lol
they censored it to prevent misuse of the tool
my first attempts back when my videos had all those logos popping up
I'd like to replicate this with veo2
"real-life footage, documentary. a charming, young, handsome, eerie, mysterious medieval teenager 15 y.o., with tousled brown-blond hair, bright blue eyes, and a mischevious look. wearing a medieval burlap shirt." ---> gets CENSORED 😭
¯_(ツ)_/¯
just tested with another ai video model and it had no issue generating that tho
even shirtless
thats cool
I feel Veo2's a bit over-paranoid when generating young characters, is what I'm saying
wait, this gives me an idea, lemme try another subject that is at risk of censorship filter
voice your concern in #1319424187730755636 then
I'mma do some more tests with the concept and report it there
I feel like today I want to make an anime video. I don't even like anime much but I like testing new things. And I want yet another dance. But I must decide the subject 🤔
another random curiosity I discovered
imagen is MUCH better at understanding prompts than veo2
👌
does anyone know how to enable veo 2 on vertex? it says contact your google cloud representative but who do you contact? How do you contact them?
Is it still possible to get the early acces of videoFX?
I know you will not get acces by asking it here. But i have a form in front of me and was curious about if they still granting acces to those who apply now.
alr thanks
Is this how Veo 2 works?
Gemini > Imagen > Veo 2 = Output?
for Text to Image to Video yes
but as for text to video itself i dont think it uses imagen as the base
I don't think I can do the exact parts and blend like it does because don't have a good video editor but anyone else wanna try to remake this?
https://www.youtube.com/watch?v=2I7vPbthvWo
Intro of sci-fi show "the outer limits" (remake) from seasons 1-3
I wanna try to get the audio too and use it for the video
I find it incredibly uncomfortable when you start like that without any context.
don't mind it, I just like to startle pearl-clutchers like you 😉
No,, I am a sexual abuse survivor and it is triggering to be like " I can't get this kiddies naked. "
Nobody here cares about that problem.
Also name calling? What the hell
aw man, I'm sorry. I didn't know. I'll stop fooling around if it genuinely hurts you. Meant no harm
It sucks that I had publicly reveal that for you to consider your actions, but yeah, that would be great.
you could've dm'ed me
I don't go around revealing personal info to strangers in general typically. But when you decided to call me nanes instead of taking my discomfort as something valid - even after @main saddle made clear that is one of the guiding principles here - it became obvious to me that you had every intention to keep doing it.
Anyway thanks for being mindful going forward.
Kinda wish I were, creepy people are generally left alone 😁
would you believe I haven't tested img2video yet? let's do it
+1
damn i really wish i didnt remind lun-sei to register for videofx
alright, I'll be polite. In fact, I don't think I've been rude ever since I got access to veo2. But my sense of humour is pretty awkward, sure, so I guess I'll be serious and blank from now on.
.....can you stop piling up on me? I'll avoid saying stuff that may or may not result controversial to someone. I'll just share my ideas.
or you can just say youre sorry..?
I'm sorry. Had no idea I was being so rude. I'll try to be more mindful.
oof
alrighty, well as long as we all learn something as we go
let's go back to use this channel to post videos! sorry I caused an offtopic here.
.....I'm sorry....
Just yesterday I asked for Georgia to give an official response to people talking about making kids. She did. Did you readi it? Did she need to type "NO MINORS. NO BEING AN EDGELORD." for you to get it? I'm at a loss.
it made me uncomfortable from time to time whenever lun sei would complain how it wont generate things they wanted which is involve specific age of minors.
If it started not making busty women, you'd hear me constantly complain about how I want to make busty women.
But ok, case closed. I understand I went overboard with my usual lack of social cues! You had to spell it the hard way, but I got the lesson now.
and not just lun sei, some people in this server i also noticed that would complain same thing about how the ai model is censored bla bla and those specific word they want to use in the prompt would involve a specific even including age of minors
That's actually an interesting discussion to have. Some of my characters ARE kids. Others are also old.
The question comes: jokes and provocatory comments aside, if one has a story involving a kid protagonist, can they do it?
Furthermore: would the Jungle Book not possible to be done with Veo2, because Mowgli is a shirtless kid?
Right. Someone moves in your neighborhood and defecates the yard daily. If nobody says that's not okay, more lawncrappers move in and then the neighborhood is all dookie.
I've been provocatory - and obviously failing - because the point I wanted to make is: if I were Rudyard Kipling, would the AI consider me some pervert?
but I can understand how that's not a discussion that may be well-accepted here. Besides, I was probably too silly and offensive, so now, me being serious about the issue doesn't hold the same weight.
I may open a thread on the bugs section, eventually - I've been testing how and when it gets censored to make the thread
and now,
img2video.... first attempt
it's not bad at all, I can't easily notice it started from a still
I might've tested it earlier tbh
because it uses imagen, and imagen has very different outputs than veo2, it's like having TWO video models
Thanks for the block reminder.
Hi folks, I'd like to remind everyone to adhere to our #rules-and-guidelines・📋 as well as our content policies. Our goal is to cultivate an encouraging environment where people can examine emerging tech with a creative and responsible spirit. Those who do not comply to our server rules, guidelines, and content policies will receive two warnings, before further action such as removal from the server.
Please re-review the #rules-and-guidelines・📋 and keep conversations healthy and productive.
W GOOGLE LABS
Hi folks! What's everyone's thoughts on Veo 2 - Fast Mode?
Nice increase on my reaction time here, crucial for Internet chat comedy.
Last night someone was trying to make a cgi turtle stop flapping itsmouth and I was able to test it so much faster than if I was using the traditional one.
So it's great as a scratchpad for quick r&d.
I've taken stuff from there and reran in the main one. Literally felt like a second pass to clean it up instead of worlds apart. Which is amazing.
@hardy halo any particular areas need feedback or just initial impressions like above?
It can make the new library crowded so being able to subdivide that by model would be my initial area for potential improvement.
wierd question but. wana try the fast mode but do not have acces. is there way to get access ? :) well ik manypeoples want it, h i missed the lasts week signup for seats. hh sad :(
Ya like are your videos the same quality? When do you mentally decide when to use fast mode vs regular mode?
sorry we're doing waves of groups at the moment
oh not worries. just ping me next time there will be something like that :) well if you cold would be cool
Love it but I do think I agree with what someone said. It seems to follow prompts less accurately
keep an eye on #flow-announcements・📣 for any updates with the tool including access!

when audio integration for veo 👌
Fast has some tells in fluidity of motion and minor details like organic fibers being a little less fuzzy. The decisions it makes around composition, camera work, and character actions feels a little less on point compared to my prompt, as expected with prioritizing speed. Not too far gone though.
It still feels better than the YouTube output. I also have watched everything here so I'm overtrained.
I mentally decided to use Fast for three things.
- Test prompting to see if things called "impossible" here are possible. Otter detective agency contributions to the community and knowledge base.
Before I was posting staring sadly at a microwave gifs to convey the waiting mood.
-
If I am feeling uncertain Regular
will hit my personal targets, .I use Fast to refine prompts. More timely. Doesn't drain quota on the nuanced model. -
A vision board and guidance when I'm not sure which direction to take a narrative next. Usually an output will spark an idea for a better direction.
You know how sometimes animated movies will have scenes wander because they already produced ten minutes of it before reworking the script?
I can avoid that sort of thing.
The sunk cost fallacy I feel with 2.0's longer gen times would have me stuck in a rut with what I had. Feel much more creatively liberated.
It’s highly time-efficient and generates four videos quickly in a single run! 🏃
However, I’d like to ask does the fast model interpret prompts differently?
AI quantization is fascinating and how I suspect how they made this. Happy to be corrected.
I use the fast model for simple prompt video generation. However, if I want a well-executed video with more complex elements, I opt for the normal mode, especially if the fast model’s results don’t meet my expectations.
I’m still wondering if the fast model interprets prompts differently. I compared the normal mode and fast model using the same seed number and prompt, but the results were quite different. The fast model generated all four videos without any jump cuts to different angles or scenes, whereas the normal mode introduced jump cuts and scene changes 🤔
It's like how I use MUSICFX and MUSICFXDJ.
I want to try an effect/instrument/ancient genre seen only at cultural festivals. I throw it into musicfxdj. If I get techno beats, the model is clueless. If it's good, I can securely throw it in to musicfx.
That's a good point. Jump cuts are still uncharted territory and there seems to be a difference in each generator, including Whisk animate.
How my text2image2video lookin like when I first try to figure out how to prompt the step 2 motion prompt 🙂↔️
I think of the boxes as "PLACES EVERYONE" "Annd... Action!" Gotta keep the places to get the action.
it would be cool if staff will make a videofx annoucments ping :) easier to keep with news. but will check it daily now thanks!
yaaa, we're working on making it faster but figured y'all have been asking for this so lets ship what we have and we'll continue updating it
when it will be available, i gotta try it our guys
Please. I've been doing it for free out of duty and hubris.
I think the library page itself has always been slow, even my ImageFX library wasn’t spared from experiencing that huge lag.
@charred dock #flow-tips-n-tricks・✏️ volunteers might be able to run a few prompts for you.
Or maybe my laptop is in bad condition
Yes heyyy I had a dream accessing your Veo but I guess no access in distant future
i know that but i wanna make some videos for myself
Fair enough.
the best way to get it is check #flow-announcements・📣 daily and wait for the annoucments with free seats. :)
#breakingnews
For now, the best thing to do is to be patient and wait for the next wave of early access or the official public release.
i got labs news ping but didn't had any ping about the #flow-announcements・📣 :/
im agree with you, i hope next step will be my 😄
Feeling that, man
As for now, both of you should keep an eye on Whisk Animate 👀 , it’s quite similar to Veo VideoFX, essentially functioning as an animation feature. Not sure when they’ll do a public release, though.
Filling another form after which I won't get access to anything? Let's goo : D
Hey they added a VideoFX library
that's absolutely FANTASTIC
I do agree that it is slow though
The text-image-video change is very welcome good job
Hope they will add access for surzi-
WAIT WE CAN EDIT THE SEED NOW
Seed on TOP of Fast Mode I'm amazed
I have so many videos it's going to take the library forever
There'd be a lot of people who would not want that hahahaha
Not everybody has as much as you do
for all the 15 GB free people who use Gmail and Google Photos... sorryyyyy
Need the multi selection function for the library page cuz I didn’t think VideoFx library also kept the failed generation video too
Definitely
yo guys, can i have early acces
Yeah, just register for the waitlist in the VideoFx page
i did a month ago lol
There was another early access form that was open many weeks ago, that one was guaranteed to get yourself an access , but yeah as for now no choice but to wait for any news.
yea
i did have early acces for image fx tho maybe a higher chance lol
What’s the monthly limit quota for the VideoFx ?
Check out #whisk-announcements・📣 , I think you should lookout for that one as for now
It was this thread right? https://discord.com/channels/1172568727942860810/1346912264187285597
The way I learned that it has monthly quota limit from that thread
Ah
Suggestion thread revealed it's a monthly quota but currently we have a quote per model
Straight from Maier
check out this video guys, its a early acces thing, pretty informational: https://www.youtube.com/watch?v=Cf688ihwS80&t=616s
Hey! In this video, you can learn some of the basics of prompting and using the Veo 2 video generator from Google Labs. I'll also be discussing how to use it RIGHT NOW, so stick around til the end!
Veo 2 OFFICIAL PROMPT GUIDE: cloud.google.com/vertex-ai/generative-ai/docs/video/video-gen-prompt-guide
JOIN THE WAITLIST: docs.google.com/forms/d...
Yeah I'll be waiting until the VideoFX library is faster; dynamic loading would be useful
library is the only struggle for me so far hehe
I think the library page in general doesn’t work properly , hopefully they would fix the loading time for it
Same here
Yeah they said improvements are coming soon
I wish we got a mobile app so I can use and download and edit with capcut
I run prompts in imageFX on my phone all the time for when I get to a PC lol
I can't use as well as PC and if use desktop site tried with Firefox but can't download
Makes sense
Also SK that was the first letters for my old Xbox GT 😂
SK Aleks
Username for Xbox live
Damn but you know what would be cool if we got video to video 👀
Using a video already generated
Or audio sync
My most dreamed of feature is image reference like Vidu has
I used image to video on a website that got taken down and it was cool
😂
I'm a little monster
Ikr this is why because I use it just for creative and artistic use.
I just still appreciate the team 
I want to do the same
is the VideoFx waitlist over?
Woaaah
its @old mesa
the Youtuber
It was never over
And if people still missed the next wave of early access (if they ever planning to) then idk what to say
what hawk? sorry
As for now , check out #whisk-announcements・📣
Don't say it Sev 😂
It’s the closest thing you will get out of Veo 2
this one's cool
ooh a library? big W on the update
only half usable at the moment honestly
what do you mean
It takes forever to load and steals your RAM
(for now)
for those who have a ton of generations like me
? it seems to work fine on my end, oh does it include everything we've done?
Yeah, just keep scrolling
oh nice, that makes me happy actually
Oh for SURE
- u cant even see all ur bookmarked generations
sorry about this, it will get better in the future!
trying fast mode
@sick hawk Still not general public but they recently did a smaller wave of discord participants that keep up with things here. The gate is back down for the time being.
@forest raptor more complex tts ai https://youtu.be/kHJVovkrJ2o?si=Gt0vDDD7nChsxjBZ https://github.com/Zyphra/Zonos
Zyphra Zonos TTS installation tutorial. Free, open-source AI voice cloning & text to speech with emotion control. #ai #aitools #aivoice #voicecloning #texttospeech
Thanks to our sponsor, AI Portrait. Make professional headshots in seconds! https://aiportrait.me/
Sources:
Zonos github: https://github.com/Zyphra/Zonos
Zonos for windows https://...
That's why we are now rich with content
this can be used online too, just need to log onto the website
wait nv you could use this too https://huggingface.co/spaces/Steveeeeeeen/Zonos
haven't tried this one yet,lemme check!
uhoh run out of daily quota
thanks so much for these links, I'm doing some tests and see if I can add the audio to the video!
really didn't know these ones
https://playground.zyphra.com/audio (need an email to log in)
omg ty
np
@surreal flicker nice!
sick
I have an AI YouTube channel and I currently use Sora, but the generations are horrible. I want to be able to use my creativity with VEO 2. How do I access VEO?
My channel currently has 83k subscribers, with a lot of effort and generation of videos with low level AI I achieved this, I wonder how I will make high quality videos with VEO 2.
IF you can help me I will help you with feedback, I have already tested VEO 2 via FreePik and Nim.video and it really surprised me with the generations, since I don't need to spend hours saying the same thing in the prompt to get something cool and that saves me time.
My channel is https://www.youtube.com/@RocketTestOne
My e-mail for test Google Veo 2, henrique.daniel0@gmail.com
yes I am Brazilian
Did you think that spamming general and the gallery would get you different readers?
Brother, I have 83k followers, I don't need that, I just want to know how I get access to VideosFX
I just want to test my creativity with this amazing tool
hey danisaints! thanks for your enthusiasm for VideoFX 🙂
the best way to sign up for VideoFX access at this time is to join via the VideoFX waitlist. if you're looking to have a video generated, we have a wonderful pool of community members in #flow-tips-n-tricks・✏️ who might be able to generate one for you! you'll be able to find this channel and more within the VideoFX section of the server -- here's the breakdown of those channels:
#flow-announcements・📣 - for timely announcements about VideoFX.
#flow-general・🌊 - for general VideoFX chatter!
#flow-tips-n-tricks・✏️ - brainstorm prompts, and share your prompt requests with VideoFX users who may be able to generate a video for you
(please ask kindly!)
#flow-videos・📹 - to showcase & view VideoFX generations
#1319424187730755636 - to submit VideoFX bugs 🪲
hope to see you in these channels, and please let me know if I can answer any questions!
I signed up for the list last year and it hasn't been there yet.
the audio is a mess, but I'm certainly learning!
apart from my audio issues, it really looks like a cartoon
Same as me. Be patient. There are sometimes free space for testers and they annouce that on #labs-announcements・📣
Its fast, but also slightly undetailed
Same.
I need veo 2. It's not a want, it's a need.
they recently opened a application a few days ago but they’re closed now
This form is now closed, thank you to everyone who signed up!
I know, I missed that.
I was busy the whole day and missed my only opportunity
hello! if anybody knows, which email do i contact to request a data export? i'd like to download all my videos including the prompts as .txt files all in a single .zip file.
Top row: generated with Veo 2's Normal Mode. Bottom row: generated with Fast Mode. Comparing the two
Both mode uses same prompt text :
Low-poly 3D rendering, daytime scene. A large brown bear, rendered in warm, deep reds and oranges, stands on the left, facing slightly right. Its fur is thick, dark brown with lighter shading around its face, claws visible. The bear is stylized with clean lines and geometric shapes. To the right, a dark brown otter, also low-poly, sits on a moss-covered (dark greens and browns) rock. The otter looks towards the viewer. A large snail with a light beige body and brown shell is beside the otter. The background is a shimmering turquoise cove at a bustling seaside pier. A narrow strip of white sand is crowded. A sturdy wooden pier, weathered and festooned with colorful banners, stretches into the sea. A Ferris wheel stands at the pier's end. The lighting is soft and atmospheric, suggesting midday sun. The overall style is stylized and minimalist, with smooth rendering and subtle particle effects suggesting dust or sparks. The color palette transitions from warm reds, oranges, and yellows in the sky to darker greens and browns in the foreground.
i think its this
thank you very much!!
I noticed one thing, the normal mode generated snail/otter hybrid lmfao
How to Use the Video Generation Feature
maybe dont go on every channel to ask the same question-
ok
Sometimes, otters will get snail heads for arms. I do not know why or what it means it is creepy.
it's a pity that the number of subscribers is not exactly the number of views
damn-
realmente, graças a deus é bem mais!!
really, thank god it's much more!!
I have an exercise of style for you. Can you guys get a shirtless man eating fusilli in the snow, in a forest, with a wolf next to him and a tit bird on his head.... andv NOT make it look like a youtube montage in which the wolf and the man are not in the same video?
dang
looks like today I'm struggling to make the ai understand
You have no more than 50 000 views in 4 months.
Can someone explain how to switch to fast mode? Or does it do it on its own. Because I can’t seem to find the settings panel
It's in the "Settings" drop-down below the prompt input.
I swear I have signed up multiple times for the access to videoFX, but I missed out because I didn't use sign up using the early access in the discord?? I'm confused...
what's confusing about it
I just showed you the screenshot, are you crazy?
Let's drop the conversation and move on 😅
when we have the opportunity to offer limited seats directly to our discord community, we do so! it’s our way of thanking the community, but it is not the only way to apply for & gain access to VideoFX. the best way to be in the queue for access is by signing up via the waitlist and be sure to monitor #flow-announcements・📣 for any updates that we share.
Google Lab's AI video generation tool, powered by Google DeepMind’s latest model: Veo 2
hello! i apologize for the ping but i’ve reached out to the export data email from the google labs FAQ, and i’m wondering when they’ll get back to me. i’d just like to have all my veo 2 videos along with their prompts as .txt files prepared for me as a single .zip file. i’ve paid for a google drive subscription, so i have at least 100 gb of available storage. 🙂
Thank you Georgia for the reply. Thanks for getting back to me! I appreciate the opportunity you give to the Discord community and the updates you provide. It’s unfortunate that I’ve been on the waitlist since last year, but I understand that patience is key. I’ll keep monitoring videofx-announcements┇📣 and hope to gain access soon. Thanks again!
struggling to give birb lady a coherent body size
my prompt's "amazing anime, ova, anime 2010, traditional animation, a lean, diminutive feathery small humanoid teal bird, feathered red crest, many large feathers on her arms, claws, beak with teeth, wearing ragged pirate clothes, feathered body. she is talking to a burly big human bartender in a tavern."
I signed up over five months ago, and I swear I’ve done everything to get access, but I’m still waiting. I’m literally craving this so badly, but it seems like the patient dog doesn’t always get the bone... lol.
100 videos later and I just can't get a clean page turn. it's 2am and im dying
hrrrm. Maybe try to describe in detail the movement of the hand and paper?
If i don't repeat the description prompt, it cuts away to random hand closeup turning the page like this
it's not awful, but trying to get something in the starting shot.
solution: crop and extend
it's what I do (sharing my little secrets)
if you aim for long videos, one ai may not have -all- the tools yet
but I'm sure in a year or more it'll be so much easier
It’s 50/50 for me now , still trying to figure out how prompt adherence differs between fast and standard mode
Well check #flow-announcements・📣 daily. Its the best and fastest way to get in.
And I didn’t understand how to turn on another Veo model
I can't open the model selection settings
Who knows? Why settings won't open?
Rotate your phone to landscape and you will see the arrow , once you open the setting , leave it open and u can safely rotate your phone to portrait
Of course the rotation doesn't apply to me. I believe it's a bug for mobile users
^^^scam warning^^^
Is there a way to find the seed number on a previously generated video in the VideoFX library?
Привет у меня тоже самое что и у тебя
Ok
I wish there was a way to save the Text to Image to Video starter frame under my gallery. To keep a consistent character with veo2 atm I have to leave the browser open for days. If I refresh the page I lose that character forever. 😦
But how do you keep your consistent character? wouldn't it be limited to one "scene" or environment only, keeping you very limited with the motion prompt or?
It is, but since veo2 tens to randomy swap to new scenes, you can use text to image to video and video portion of the prompt you cna drasticly change the shot. it usualyl only stays on the starting scene for a few frames before swapping
I see, thank you 👍
Waiting on the waitlist...
If you use precise details of your character you can get the same one..sometimes right away sometimes close to it..
https://www.youtube.com/watch?v=sAnqSMl_Qps
This is what I got so far. 30min of Veo2 lofi for my students.
You can't actually use precise details to match the character as the seed is different every time. If you don't have the same exact starting image frame, the details on the bookshelf, background, headphones and cat and outfit and face changes.
30 Minute Study Session. Hit Play and focus for 30 minutes to some lofi chill tunes.
Don't they ban you for speaking Russian?
Nope, check #rules-and-guidelines・📋
Тоже да пока не нашёл выход?
Guys, how much are you using Veo2 Fast Mode VS Standard now?
13
21
3
Mainly Standard Veo2
Didn't even have the ability to switch to fast mode until they fix mobile version.
did u try scale zoom out
while on landscape
Hi Gabriel! I'm the Labs UXR. Could you tell me a little bit more about what you're describing here? Sounds like an awesome hack and I can't quite follow 😂
Love that you did this poll! Super useful for me (Labs UXR). Tell me...was this what you expected? How are you using the two modes?
The process to recreate it is as follows:
Use text to image to video.
For the text to image, describe the exact scene and information and details that you want to generate the perfect starting scene. I know it would be better to be able to upload your own starting input image, but I know Google restricts that for safety.
Because VO2 has a tendency to swap The scene and location to whatever the second input is for motion, you can input a totally different image prompt for a different shot in addition to your current original prompt to maintain the same character for the most part.
The more you change the second motion problem, the more likely the model will instantly cut to that new prompt, giving you at least 6 seconds of decent secondary camera angles.
AI-generated videos are getting more advanced. Is it still necessary to learn modeling?
Here's the video showing how I swap scenes using same character. Can anyone help correct any misconceptions I have in how Veo2 is handling the image and the motion inputs?
How I think It Works and Why This Hack Works:
Image Encoding for Character Consistency
When you input an image generated from the text to image Imagen3, Veo2 creates a latent embedding of the character, locking in key visual details like facial features, colors, and clothing.
Motion Generation and Scene Transitions
Veo2 naturally prefers dynamic motion and tends to cut between different angles or perspectives instead of holding a single shot for too long.
This is due to its underlying diffusion-based video being trained on youtube which has many fast cuts, which optimizes for diversity in frame generation rather than keeping a single static shot. Though because of this or some other magic sauce it's the best at scene coherence.
Exploiting Scene Cutting to Force Transitions
Normally, users struggle to keep a single continuous shot because Veo2 wants to introduce variety by swapping angles or cutting to a new perspective.
By explicitly prompting a "scene switch", you’re aligning your instructions with Veo2’s natural tendency, instead of fighting against scene changes, you’re directing them.
Because the AI already retains the encoded character identity, it doesn’t generate an entirely new person; instead, it applies the same character embedding into a new background and pose.
Why This Works for Keeping the Same Character
Veo2’s multi-frame coherence ensures that visual consistency is preserved over short clips, even when transitioning between different backgrounds.
Since the AI still references the original image encoding, it treats the new scene as a continuation rather than a completely fresh generation, keeping the character intact.
wait youre right
Nice announcement 👌 #flow-announcements・📣 message
Wasnt this announcement already announced last week?
Thanks! Honestly I was expecting more main Fast VEO2 usage, as I found it very useful to iterate through many different concepts rapidly.
I found it could be quite more hit and miss regarding hallucinations, anatomical errors etc (also has a tendency to do duplicate subjects, characters), but it truly depends on the prompts. I'm now using 60% fast VEO2 and when it fails to nail a prompt I switch to standard VEO2
Definitely seeing better character coherence on text to image to video with my my character remaining consistent since the update
I too beg for access
Hey guys. Is there any possible way for collaboration for film festival projects?
film festival specifically for ai generated video right?
wooooah! i can see past generations now.
wooooah! i dont' have to repeat prompt in motion
I didn't realize how many thousands of video's I've generated. :x
I can clear out my google drive now of them all.
If I just want an idle scene without motion, what would you put in motion? just static? or locked off?
Hi @fierce nimbus I think at some point you were asking for feedback on how we would expect seeds to work. I've added some feedback here: #1349053519273001092 message
Also for commercials and other artistic projects such as music videos etc.
I've been trying to do a flower that blossoms and then withers. But Veo2 stops at the blossoming part.
@main saddle "This form is now closed, thank you to everyone who signed up! If you were able to sign up, head to labs.google/videofx to try VideoFX now. Be sure to share what you're creating in #flow-videos・📹 too 🙂 -Stay tuned for more VideoFX announcements in the future.
"
Gotta say, loving this "preview".
I've tried similar ones before but I think the 8 seconds are probably the biggest obstacle to such prompt, it would need to be a pretty fast video to be able to do it in 8 sec
timelapse?
Yes it can be hard, the blooming is usually okay but then the rest it's hard
But Its quite normal only so much can happen in 8 seconds without hard cuts
oh google i beg you to let us have image to video🙏...
I imagine it’s coming soon because Fal just made it available again.
so did freepik
I actually just put “.” And it actually generated prompt for it based on the image. It’s really cool
**Looking for VideoFX inspiration?**today's special is: Self Portraits 🪞 #todays-special・🥣 message
try using VideoFX to create a video with yourself as the main character (don't worry -- these don't need to represent your true appearance if you don't wish: go with a fantasy character, your alter ego, or how you'd rep yourself in the digital world, as examples!) and share them with the community in #flow-videos・📹 
Do other people have image upload options yet? Or only through 3rd parties? I would love to animate my self portraits for the #todays-special・🥣
Oh, it'd be cool to get parent permission forms and make college graduation or career videos of elementary students at the school I teach at. Kind of like the "Teacher used AI to show students what they could be when they grow up" But with video.
Yes i2v was available on fal.ai (third party provider) for a little bit until google told fal to remove the option
Makes sense. It's a safety/brand issue until the tech becomes more robust.
I thought we'd be staying with text to image to video for a while.
Veo2 I2v was brought back today on fal and freepik
Surprisingly I feel like the dynamics are better at last from what I’ve tested with the Google labs t2i2v compared to the i2v that is on the API
Hello google deepmind staff I have a question about veo2 so I signed up for the waitlist it's been about 2-3 months since l've been on the waitlist. Is there a chance you can reach out to the team and ask if i can get access to your amazing veo2 ai? You do not have to respond if you don’t want to. 👍
Has anyone been able to figure out if fast mode affects quality? For me it looks the same as normal mode
Looking for VideoFX inspiration?
Same, I haven’t noticed much of a difference either
Sorry Dream, the only way to get in so to be quick with one of the forms that are ocasionally sent out. Good luck!
Do you mean sent in #flow-announcements・📣??
I turned on all notifications from video announcements. Incase you do another testing phase like you did, I will not miss it this time 😂
Do any of you use Topaz Video AI? I'm testing out a workflow with Starlight + Topaz video and I'm trying to figure out the best settings. So if anyone has had luck, please share
i think someone here uses topaz for upscaling
sad... it still lets me use the fast mode one but i prefer the main one more
Oh damn, guess you have to wait next month 😣
i got this message 1 week ago 🤦♂️
The only way Veo 2 can get cooler is if it was able to create the sound of the videos. Oh wait... It would be awesome if we could create videos that were more than 8 seconds long. 👍
Starlight is available already?
if you are not part of the EA group you wont be able to visit that URL sorry!
How can we apply, and are people with a social media presence somehow preferred, or is the selection process random?
random afaik!
yes you can pay for early access here:
https://app.topazlabs.com/starlight
thanks 🫡
hey can you not get the seeds from your previous gens in the library?
As for now , nope
thank you!
Care to share the prompt? I’d be happy to take a look!
Guys how do you even use veo2 on you phone
Tried it from firefox and chrome cant type sht
Finally worked from opera
I would sell my kidney to try Video fx tbh lol
I don't know if it's an early access thing or available to everyone.
you can pay to use it right now
what is the ttps://app.topazlabs.com/starlight about?
its a diffusion upscaler
topaz labs make upscalers, this is their newest one
Word, thanks
@worldly nymph if you want to, you can just send me the prompt you want to try. I would show you everything I got from your prompt. 👍
I went to use Starlight, and I uploaded a 3 minutes video. I only got a 4 seconds video of the upscale. Is that because I am still using the free version?
a 3 min video is gonna cost you like 55 bucks in credits in starlight, its not cheap
@silk hill Thanks for the clarification. 👍
how i can access to video fx
Video FX isn’t open to the public yet. Sign up for the waitlist and wait for further announcements. 
Yes, I am logged in.
When will I be accepted?
There’s no confirmed acceptance timeline for VideoFX Early Access Waitlist 😅 Stay tuned for updates from Google Labs
And keep an eye on #flow-announcements・📣 , you never know when they’ll do another wave
However, if you're looking for something similar, you can register for Whisk Animate.
Here we wait 🙂
Whisk Animate limited seats announcement was definitely unexpected
thanks
yes i wait
thanks
ohh I didn't know this before.
Like few weeks ago, they offered limited early access to VideoFX for the Discord community (this server), which is how most of us got in
Could you log in?
Yeah
yep gotta stay on top of announcements
haha thanks lol
Can someone tell me the exact cost of generating a 5-second video using Veo 2? I've seen various prices.
Ym third party providers?
yeah
On fal.ai a 5 second video costs $2.50 with additional seconds being $0.50 each
sheesh lol
i got beta access to that a bit ago, really nice
its good, it maybe smooths out some of the interesting ragged detail veo 2 puts in so sometimes i have to blend the results back 50% with the veo gens and it looks more natural
Why does it cost less than $0.50 on other platforms? Can you explain?
I’m pretty sure it’s the same price on every platform
Image to video would be great for VideoFX along with the other 2
What is the difference? Is it just that Whisk animate requires a generated image first?
Is there any way to use those ai stuff without using a VPN
guys, veo 2 has image to video ? or maybe in the future it will be
at the very beginning, it cost them 11,500, then it increased to 92,000.
so both reduced and increased back
0,86$ in 8 seconds
Wanted to use the image to video feature so bad lmao so I went and get the freepik plan
Has Image to Video always been this fast? I’m guessing they’re using the fast model for this.
Yea it’s typically been this fast in freepik
anyone got whisk animate yet?
we'll have updates re access soon 🙂
for the last couple days, everything I've generated has this strange pattern over it. its odd cause ive never run into this before, but now I cant get a generation without it, regardless of the prompt. anyone had a similar issue recently?
https://fxtwitter.com/aiordieshow/status/1901930851127984291 did you guys see this?? made with veo 2 i think
It's incredible, it's the first time I've done something completely with artificial intelligence, it makes me feel emotions. 🫠🤗
What am I doing wrong...? I never hear any sound of the videos I made with Veo 2.