Veo / Flow Generation Issues, Lost Credits, Consistency Problems, and Excessive Failed Generations

Google AI Studio
,
1

I’m submitting this because I’m honestly frustrated and I’ve reached a point where using Veo and Flow feels more stressful than creative.

I create story-based content and spend a significant amount of time planning scenes, continuity, prompts, references, and character consistency. My projects are not random one-off generations. They are connected sequences with recurring characters, environments, and ongoing narratives.

The biggest issue is that I’m repeatedly losing generations while trying to create content that should reasonably work.

Some of the problems I keep running into:

1. Constant failed generations

I repeatedly receive:

“I can’t generate that video. Try describing another idea. You can also get tips for how to write prompts and review our video policy guidelines.”

The issue is that many of these prompts are NOT violating policy.

These are often simple lifestyle scenes:

  • characters walking

  • entering rooms

  • sitting at tables

  • adjusting bags

  • rooftop scenes

  • museum scenes

  • friends talking

  • everyday interactions

Nothing explicit.

Nothing unsafe.

Nothing extreme.

Yet the system refuses them anyway.

After spending time planning prompts and references, this becomes extremely frustrating.

2. Lost generations feel expensive

Every failed attempt matters.

Creators do not have unlimited attempts.

When a generation fails:

  • time is lost

  • workflow breaks

  • creative momentum disappears

  • generation limits get burned

  • upload schedules get delayed

This is especially frustrating when working with daily limits.

I feel like I’m spending more time fighting the system than creating.

3. Character consistency is extremely unstable

This is probably my biggest complaint.

I use recurring characters.

I carefully establish:

  • outfits

  • shoes

  • hairstyles

  • accessories

  • body proportions

  • environments

  • visual identity

Then Veo changes things anyway.

Examples:

Character shoes randomly changing

Hair changing unexpectedly

Facial details appearing or disappearing

Objects moving

Characters morphing

Unexpected body changes

New features appearing that were never requested

Wardrobe swaps

Continuity breaks between clips

This becomes a huge issue for episodic storytelling.

4. Movement quality feels inconsistent

Sometimes motion looks great.

Other times:

Characters move in slow motion

Walking becomes unnatural

Simple actions barely happen

Small gestures get ignored

Characters freeze

Movements become stiff

Actions requested in prompts do not occur

A scene asking for “walk into room and sit” can become:

walk halfway

pause

stand still

slow-motion movement

incorrect body positioning

This makes multi-scene storytelling difficult.

5. Flow and Veo struggle with continuation projects

I build connected stories.

Clip A leads into Clip B.

Clip B leads into Clip C.

Current issues:

identity drift

environment drift

camera mismatch

character replacements

object switching

continuity loss

Characters stop looking like themselves.

Scenes stop matching.

Previous references appear ignored.

It feels like every clip starts from zero again.

6. Prompt sensitivity feels unpredictable

Sometimes small wording changes completely break generations.

I’ve had prompts rejected that were almost identical to prompts that worked earlier.

The inconsistency makes it difficult to understand:

what is allowed

what triggered rejection

whether policy actually blocked it

or if generation simply failed

More transparency would help creators.

7. Error feedback is too vague

The current message:

“I can’t generate that video…”

does not explain:

WHY it failed

Which part triggered it

If it was policy

If it was motion

If it was character issues

If it was system instability

If it was temporary

Creators are left guessing.

Please give actual feedback.

Even:

“camera instruction issue”

“motion conflict”

“policy review”

“continuity instability”

would help.

8. Request to Google

Please improve:

  • continuity retention

  • recurring character memory

  • outfit locking

  • shoe locking

  • object consistency

  • movement realism

  • continuation workflows

  • clearer rejection reasons

  • protection against wasted generations

  • recovery when generations fail

Please also consider refunding failed generations when videos never render.

Creators are investing real time into these projects.

I genuinely want to keep using Veo and Flow because the potential is incredible.

But right now it feels like I’m battling the tools instead of creating with them.

I’m frustrated because I care about the platform and want it to work.

2

I have the same frustration. I don’t know what to do.

Today it’s more frustrating cause when i provide a ref image, ut changes everything characters and environment.

3

Keep changing my Character in the prompt and cause inconsistency using the Agent

4

Yup, the quality that Veo generate is below cheap / free ai generator. It didn’t give what you want. Unknown character, location, wrong ratio, don’t know how to follow a simple prompt.

It just take/rob your credit. Veo can’t even make a proper video, there are more toward robbing your money.

5

Our team all feel the same. Google appear to be launching updates and celebrating while the reality for heavy ultra users is very different. Characters created as part of Omni update and then day or two later the character has disappeared from inventory.

Constant battle with prompts that go against policy - but nothing specific is stated.

Try lodge a ticket and you will get 10 hours of homework designed to wear users down. There is still unresolved tickets from late Feb when Gemini lost all of our history in research chats.

Ultra users should have been taken better care of.

6

I can relate. I do experience same. It’s just so frustrating. Even when you want to edit a video generated already with a prompt which Omni flash or so is to handle, the video is generated almost thesame way, may be with little different which does not even honour the prompt. Even when you use extend, the last word of the dialogue still flows into the the beginning of the second clip. Google needs to improve this. AI for coding is doing well when it comes to writing code to build software but this text/image to video generation is at infancy to me. More work needs to be done to enable people create cinematic ai films.
I don’t know if anyone has tried Grok Imagine for comparison

7

I have exactly the same issues. I have limited time to work on projects and it drains away in constant regeneration, spurious red flags and highly variable response to prompts. Every now and then it spits out something amazing and you think you have cracked the system but the next second it goes into a loop, hallucinates something weird or raises yet another red flag. It’s soul destroying because it clearly could be amazing. The failure to address consistency in generation, guard rails and constant failures makes it pretty much unusable for anything other than ai slop.

8

I am here because of having similar issues. Lost all my credits. Had total of 20 creation left and I cant even use any single one of them. Tried to create simple video from 2 images. 1 first frame and 1 last. It always adds unnecessary stuff or changes things midway. There is no consistency and all my credits are gone. I’m at a point where I can’t stop working, because Its been almost a day for trying to find right prompt and describe what i want clearly. But even with the same prompt or same pictures, outcome is always different and with a lot of mistakes and issues. Lastly, I tried to use my wife’s account because I dont have any limit in my cards right now to purchase any more credits or upgrade my plan. Even tho, when I FINALLY managed to get it right, THIS TIME the video is corrupted. Some fractures and glitches happening all around in the upscaled export.

So, I am asking: If any one sane person requested a job or project from someone.. Would they still pay full price for incomplete job and pay the same amount for every revision, yet still not getting what they wanted.. What would you do? I’d personally punch the shit out of them till they give my money back. SO?

9

It is definitely very frustrating, it’s getting worse and worse, I can’t generate anything in Flow anymore. That blessed message constantly pops up: “we detected unusual activity”, and honestly, I don’t see any such unusual activity when the only thing I do is create architectural images. I’m going to run away from there.

10

same issue last2 days i got bad result ..worst lost all credits

11

when I used a free account I had no issues, I paid for a account and I get unusual activity and it does not make things, I log on to another google account and sure it will make images again…

12

I have the same issue. I am running a trial with google flow and so far its hit and miss. I have created characters and upload them as reference. Nothing complicated, and it completely re-renders the characters into something completely different to the reference I upload. I have very detailed prompts and specifically state that the characters must remain the same as in the reference images but it still changes them. So 75 credits later and I still dont have something useable. I dont think I can rely on this product. BTW Its an 8 second clip.

13

Hello All,
This community dedicatedly supports the Gemini API and Google AI Studio.If you have issues with Google Flow, please visit the Google Flow App, click on the three dots in the menu, and select Feedback to report your issue.

14

Does anyone know if this terrible service and frustration that we get with google also generated badly or possibly doesnt generate at all when you use VEO via other ai platform like Higgsfiels etc ?