THE BLACKCAT Voice Extractor Studio

# THE BLACKCAT Voice Extractor Studio

## Purpose

THE BLACKCAT Voice Extractor Studio separates human voices from music beds, soundtracks, and mixed interview audio.

## Current Location

```text
/Users/pabloscandoglierro/Documents/General Stuff
```

## Current Engine

Demucs installed in:

```text
/Users/pabloscandoglierro/Documents/General Stuff/.venv-demucs
```

## Local Dashboard

Start the browser app:

```bash
scripts/open-voice-extractor-studio.command
```

Or run the server directly:

```bash
python3 tools/voice_extractor_studio/server.py \
  --port 8770 \
  --audio-root "/Users/pabloscandoglierro/Documents/General Stuff"
```

Then open:

```text
http://127.0.0.1:8770
```

## Dashboard Flow

1. Put the source MP3 in `/Users/pabloscandoglierro/Documents/General Stuff`.
2. Open `THE BLACKCAT Voice Extractor Studio`.
3. Choose the audio in `Fuente local`.
4. Keep `WAV master` and `MP3 review` selected unless a smaller delivery is needed.
5. Press `Separar voces`.
6. Open the output folder when the job completes.

## Current Run Pattern

Manual fallback:

```bash
cd "/Users/pabloscandoglierro/Documents/General Stuff"
. .venv-demucs/bin/activate
demucs --two-stems=vocals -o separated "INPUT_FILE.mp3"
```

Then post-process the extracted voice stem:

```bash
ffmpeg -y -i "separated/htdemucs/INPUT_FILE/vocals.wav" \
  -af "highpass=f=80,lowpass=f=12000,loudnorm=I=-16:TP=-1.5:LRA=11" \
  -ar 48000 \
  "voice-only-output/INPUT_FILE/INPUT_FILE_voices_clean.wav"
```

## Output Pattern

- Raw voice stem: `separated/htdemucs/<file>/vocals.wav`
- Clean edit WAV: `voice-only-output/<file>/<file>_voices_clean.wav`
- Review MP3: `voice-only-output/<file>/<file>_voices_clean.mp3`