👥 Speaker Separation

Analyze multi-speaker audio files to automatically detect and separate individual speakers into separate audio streams.

Upload an audio file with multiple speakers, and this tool will:

• Detect all speakers automatically
• Separate each speaker's audio
• Export clean individual streams

📚 Usage Tips

How to Use:

1. Upload Audio: Select an M4A, WAV, or MP3 file with multiple speakers
2. Configure Detection:
   • Use min/max speakers for auto-detection (recommended)
   • Or set exact speaker count if you know it
3. Choose Output: Select format, sample rate, and bitrate
4. Separate: Click the button and wait for processing
5. Download: Get individual speaker files and a detailed report

Best Practices:

• Clear audio with distinct speakers works best
• If you know the exact speaker count, specify it for better results
• Processing time scales with file duration (expect ~2x realtime)
• M4A format provides best quality-to-size ratio
• For long files (>1 hour), expect several minutes of processing

Troubleshooting:

• If fewer speakers detected than expected, try increasing max_speakers
• If too many speakers detected, try increasing min_speakers
• For overlapping speech, the tool will assign to the dominant speaker

Extract a specific speaker from audio using a reference clip. Upload a short clip (3+ seconds) of the target speaker's voice, then upload the audio file to extract from.

Examples

Voice Denoising

Remove silence and background noise from audio files using voice activity detection (VAD).

How it works:

1. Upload an audio file
2. Adjust VAD and silence thresholds
3. Process the audio to remove unwanted segments
4. Download the cleaned result

Tips:

• VAD Threshold: Higher values are more aggressive (remove more segments)
• Silence Threshold: Larger values keep longer silent gaps
• Min Duration: Filters out very short voice segments (reduces false positives)

Examples

Extract specific voices from audio files using a reference clip. Upload a reference voice clip and one or more audio files to extract matching voice segments.

📚 Examples

Quick Start Guide:

1. Upload Reference Voice: A short, clear clip (5-30 seconds) of the voice you want to extract
2. Upload Audio Files: One or more files to process (can be long recordings)
3. Select Mode: Choose what to extract:
   • Speech: Only spoken words and sentences
   • Nonverbal: Sighs, laughs, moans, humming, etc.
   • Both: Everything from the matched voice
4. Start Extraction: Click the button and wait for results
5. Download: Get individual segments or download everything as a ZIP

Tips for Best Results:

• Use a high-quality reference clip with minimal background noise
• Reference should contain only the target voice (no other speakers)
• Enable VAD optimization for faster processing of sparse audio
• Adjust voice threshold if you're getting too many/few matches