ffio.py` with audio-loading functions based on `ffmpeg` and `librosa.resample` as an option for UVR5 #63

New issue

Open

opened 2026-10-20 09:53:35 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-20 09:53:35 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion/pull/63
Author: @hyperuser178
Created: 12/14/2025
Status: 🔄 Open

Base: develop ← Head: develop

📝 Commits (5)

f55ca2a Add ffmpeg-based function wavread to read audio files
772ae87 Added wavread_rosa based on librosa params & librosa based resampling
ea23680 Fix wavread_rosa call during handling single channel audio
e1192de Implement ffmpeg loading option for audio processing
72396aa Added an audio loading method based on ffmpeg as an option

📊 Changes

2 files changed (+138 additions, -8 deletions)

View changed files

➕ rvc/lib/uvr5_pack/ffio.py (+116 -0)
📝 rvc/modules/uvr5/vr.py (+22 -8)

📄 Description

Functions wavread & wavread_rosa are added, where the latter is called as ffread_rosa in the file rvc/modules/uvr5/vr.py as an alternative, in order to avoid bugs that librosa(.core).load may cause.

Due to that the param res_type is defined in librosa, the resampling is also performed by librosa.resample in the function(s).

However, due to that I'm not sure what kind of unit tests are required, it hasn't been performed / passed unit tests.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/RVC-Project/Retrieval-based-Voice-Conversion/pull/63 **Author:** [@hyperuser178](https://github.com/hyperuser178) **Created:** 12/14/2025 **Status:** 🔄 Open **Base:** `develop` ← **Head:** `develop` --- ### 📝 Commits (5) - [`f55ca2a`](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion/commit/f55ca2a3c17e6c0ee2af53d0ac6d12304a820e90) Add `ffmpeg`-based function `wavread` to read audio files - [`772ae87`](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion/commit/772ae871b26fcffd3524056b47949f015276a3c4) Added `wavread_rosa` based on `librosa` params & `librosa` based resampling - [`ea23680`](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion/commit/ea236804e7551d53c8d2c542dd7614d676504650) Fix `wavread_rosa` call during handling single channel audio - [`e1192de`](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion/commit/e1192deabaae0b6b705c9f34b32123a5721427ca) Implement ffmpeg loading option for audio processing - [`72396aa`](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion/commit/72396aa00fc448e2e402ed23e6e9522d67564359) Added an audio loading method based on `ffmpeg` as an option ### 📊 Changes **2 files changed** (+138 additions, -8 deletions) <details> <summary>View changed files</summary> ➕ `rvc/lib/uvr5_pack/ffio.py` (+116 -0) 📝 `rvc/modules/uvr5/vr.py` (+22 -8) </details> ### 📄 Description Functions `wavread` & `wavread_rosa` are added, where the latter is called as `ffread_rosa` in the file `rvc/modules/uvr5/vr.py` as an alternative, in order to avoid bugs that `librosa(.core).load` may cause. Due to that the param `res_type` is defined in `librosa`, the resampling is also performed by `librosa.resample` in the function(s). However, due to that I'm not sure what kind of unit tests are required, it hasn't been performed / passed unit tests. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-20 09:53:35 +00:00