High failure rate in T2AV generation
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.1k
- Forks
- 93
- Avg merge
- 1m
- Merged PRs (30d)
- 1
Description
Describe
When using the MOVA pipeline for pure T2V/T2AV generation, I am experiencing a highly unstable generation process. Approximately 40% of the generated videos result in corrupted, solid-color outputs (the entire video is just a flat color with no coherent structures or details).
Following the standard T2V approach for MOVA, I am passing a pure white PIL.Image as the image condition to the pipeline. I strongly suspect the issue lies in how the pipeline_mova.py loads, encodes, or concatenates this pure white frame in the prepare_latents stage, leading to a latent collapse during the diffusion process.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in pipeline_mova.py, focusing on the prepare_latents stage where the pure white PIL.Image is loaded, encoded, and concatenated for standard T2V generation. Reproduce the MOVA T2V/T2AV flow and compare successful videos with the corrupted solid-color outputs. Done means the generation process no longer produces the reported latent-collapse failures at the stated rate.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- audio-video-rtc, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100