deezer / deezer/spleeter

[Bug] The separation results from python library perform not well.

Open
#840 0 comments 0 reactions 0 assignees View on GitHub
bug invalid
Dominant language
Python
Stars
28.4k
Forks
3.1k
PR merge metrics
No merged PRs in 30d

Description

- [☑️] I didn't find a similar issue already open.
- [☑️] I read the documentation (README AND Wiki)
- [☑️] I have installed FFMpeg
- [☑️] My problem is related to Spleeter only, not a derivative product (such as Webapplication, or GUI provided by others)

## Description

I try spleeter using command line(https://github.com/deezer/spleeter/blob/master/spleeter.ipynb), it works really well.
When I try it from python library (https://github.com/deezer/spleeter/wiki/4.-API-Reference#separator), the output from the separator of the audio "audio_example.mp3" is not separated.
Is there anything wrong when I got the separation results?

## Step to reproduce

The below python script shows how I got the separation result.
```python
import numpy as np
import soundfile as sf
from spleeter.audio.adapter import AudioAdapter
from spleeter.separator import Separator
# Load wav using adapter
input_file = "audio_example.mp3"
audio_loader = AudioAdapter.default()
sr = 44100
signal, _ = audio_loader.load(input_file, sample_rate=sr)
# separate
separator = Separator('spleeter/resources/2stems.json')
pred = separator.separate(signal)
sf.write("output/vocals.wav", pred["vocals"], sr)
sf.write("output/accomplaniment.wav", pred["accompaniment"], sr)
```

## Output
The output vocals and accompaniment seem like compressing the volume of the input "audio_example.mp3", not the correct separation result.

## Environment

| | |
| ----------------- | ------------------------------- |
| OS | Linux |
| Installation type | pip|
| RAM available | XGo |
| Hardware spec | CPU|

## Additional context

Contributor guide

Open the contributing guide

Research direction

Start with the Python API example using AudioAdapter.default(), Separator, and spleeter/resources/2stems.json, then compare it with the spleeter.ipynb command-line flow. Reproduce the issue with audio_example.mp3 and inspect whether the generated vocals.wav and accomplaniment.wav contain separated stems rather than a quieter copy of the input.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
audio-video-rtc, machine-learning
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.