Video generation using Veo 2 stopped working for me. It worked some days ago and generated one or two videos and then came up with “quota exceeded”.
After a few days it has stopped generating videos. And then throws up the quota error.
What is the quota on Veo 2 for the testers? Does quota get exhausted even in case the video prompt is rejected?
3 Likes
I just used it for the first time today and generated 5 in a row before it gave me that message. Annual paid pro subscriber.
Thanks for your inputs. I am not sure if the tester needs to be an Annual paid pro subscriber. Google may please clarify. Today it generated once (cows grazing in a green pasture) and then stopped saying “quota exceeded”. So non pro subscribers get one video? If yes, then what’s the frequency/time limit before another video can be tried?
Have you tried with images having people? Did it work?
I guess it depends on the queue.
I could generate 2 videos in a row (no subscription; it threw the error several times, just be naggy and try again), then the third 2 hours later.
Thank you for your response. Yes I had been able to generate videos without subscription. However for some reason it throws up the error “Your prompt may have been blocked due to safety reasons, please update it and try again.” when there is any human in the image. For example, I had an AI generated image with three people looking up at the clouds in the sky. It errored out. On the other hand, it generated a video of a white elephant walking in the clouds.
Can someone from Google explain what’s the logic?
1 Like
I tested Veo via api and it’s amazing! In addition to the very low quota for free use, the cost is too high to invest in creating videos using Veo, so far it is cheaper to pay an influential person to speak or present in a video instead of making the video with this model.
Learn from my mistake – I switched over to Colab and ran everything via Vertex AI to experiment with videos. In just 446 seconds I racked up a £175 bill (around $230), and I suspect some of that cost comes from failed image-generation attempts. I did not realise until it was too late that the charge is $0.5 per second
@Ganesh_R , for safety purposes, the models will not generate humans by default but you can enable while using the api by changing the “personGeneration” parameter to “allow_adult”
also the model will not generate videos of Kids or babies.
please refer the following model parameters for documentation [https://ai.google.dev/gemini-api/docs/video#veo-model-parameters]
Hi @Akhilesh_Kambhampati
Thanks for your reply. I am referring to Veo2 which is in the experimental phase, not Veo.
I see that the documentation link you provided states that person models aren’t not allowed in “Image to Video” irrespective of the personGeneration parameter.
I was actually able to generate a couple of videos in Veo2 under Image to Video feature. However it seems to have been turned off after the last update. Why so?
It is so strict that it doesn’t even accept as input an image with a marble idol of Shirdi Sai Baba which was also generated using AI.
Please consider removing this control and allow generation of videos in Veo2 at least during the experimental stage using image having people or statues as inputs. It would help you refine your product. It would be up to Google to decide whether to block it once the product is in production. By the way there are other competitors who do not have such rigid controls.
Thanks.