RVC-Project / RVC-Project/Retrieval-based-Voice-Conversion-WebUI
`preprocess.py` could silently drop training data
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 38.4k
- Forks
- 5.3k
- PR merge metrics
- No merged PRs in 30d
Description
During the parameter rename (idx0 → output_key), the final norm_write call was dedented out of the slicer loop:
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/81eed5e8f68b6bed1789f682fe78cdd324495afc/train/preprocess.py#L52
The tail chunk is a full-length training sample. And since the slicer's min_length is 1500 ms, any sliced utterance shorter than 4 s consists entirely of that tail chunk, so it's dropped whole.
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
Open train/preprocess.py around lines 52 and 96-109 and compare the norm_write call with the surrounding slicer loop. Confirm that the tail chunk is processed for each sliced utterance, including utterances shorter than 4 seconds, then run the preprocessing path on representative short and longer audio to verify no training data is silently dropped.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- data, machine-learning
- Issue type
- Bug
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Quiet
- Clarity
- Clearly specified
- Newbie friendliness
- 78/100