/redub

description: Redub video with a different voice

Redub Video

Replace the voice in an existing video using ElevenLabs transcription + TTS.

This is a **utility command** - it works on any video file without requiring a project structure.

Pipeline

Input Video → Extract Audio → Transcribe → TTS (new voice) → Replace Audio → Output Video

Entry Point

Step 1: Get Input Video

Ask the user for the input video:

/redub - Replace voice in a video

Enter the path to the video you want to redub:

If user provides a path, verify it exists. If not found, list recent MP4 files:

find . -name "*.mp4" -type f -mtime -7 | head -10

Step 2: Choose Target Voice

Help the user choose a voice. Options:

**Option A - Use default voice:** Check `ELEVENLABS_VOICE_ID` env var or `toolkit-registry.json` config.

**Option B - List available voices:**

# User can find voices at: https://elevenlabs.io/app/voice-library
# Or use the API to list voices

**Option C - User provides voice ID:** Ask for the ElevenLabs voice ID directly.

Present as a question:

Which voice should narrate the redubbed video?

1. Use default voice (from .env or registry)
2. Enter a voice ID
3. Help me find a voice (opens ElevenLabs voice library)

Step 3: Transcript Handling

Ask how to handle the transcript:

How should we handle the transcript?

1. Auto-transcribe and generate (no review)
2. Auto-transcribe, let me review before TTS (recommended for accuracy)
3. I have my own transcript file

If option 2: set `--save-transcript transcript.txt` and pause for review. If option 3: ask for transcript file path.

Step 4: Output Location

Suggest default output path:

Output: {input_stem}_redubbed.mp4

Accept default or enter custom path:

Step 5: Execute

Run the redub tool:

source .venv/bin/activate && uv run tools/redub.py \
  --input "INPUT_PATH" \
  --voice-id "VOICE_ID" \
  --output "OUTPUT_PATH" \
  --json

For transcript review workflow:

# Step 1: Transcribe only
source .venv/bin/activate && uv run tools/redub.py \
  --input "INPUT_PATH" \
  --voice-id "VOICE_ID" \
  --output "OUTPUT_PATH" \
  --save-transcript transcript.txt \
  --dry-run

# Show transcript to user, let them edit
# Step 2: After approval, run with edited transcript
source .venv/bin/activate && uv run tools/redub.py \
  --input "INPUT_PATH" \
  --voice-id "VOICE_ID" \
  --output "OUTPUT_PATH" \
  --transcript transcript.txt \
  --json

Step 6: Report Results

Parse the JSON output and report:

Redub complete!

Input:  original.mp4 (45.2s)
Output: original_redubbed.mp4 (47.1s)
Voice:  {voice_id}

Duration change: +1.9s
(The new voice speaks slightly slower than the original)

Keep intermediate files? [y/N]

If duration differs significantly (>5s), suggest:

• Adjusting `--speed` parameter
• Editing the transcript to shorten/lengthen

---

Quick Mode

For experienced users, support direct invocation:

/redub my_video.mp4 --voice-id ABC123

Parse arguments and skip interactive prompts.

---

Tool Reference

| Option | Description | |--------|-------------| | `--input, -i` | Input video file | | `--output, -o` | Output video file | | `--voice-id, -v` | Target voice ID | | `--transcript, -t` | Use existing transcript (skip STT) | | `--save-transcript` | Save transcription for review | | `--language, -l` | Force language (auto-detects by default) | | `--speed` | Adjust TTS speed (default: 1.0) | | `--keep-temp` | Keep extracted/generated audio files | | `--dry-run` | Preview what would happen |

---

Use Cases

• **Voice replacement**: Replace your voice with a professional narrator
• **Anonymization**: Replace identifiable voices in demos
• **Localization prep**: Get transcript for translation (then TTS in target language)
• **Voice consistency**: Re-record old videos with your current cloned voice
• **Accessibility**: Generate clearer narration for videos with poor audio

---

Limitations

• **Timing**: New TTS won't match original timing exactly. Use `--speed` to adjust.
• **Non-speech audio**: Background music, sound effects are lost. Extract those separately if needed.
• **Multiple speakers**: Currently treats all speech as single speaker. Diarization possible but not implemented.

---

Error Handling

| Error | Solution | |-------|----------| | `ELEVENLABS_API_KEY not found` | Add to `.env` file | | `No voice ID provided` | Use `--voice-id` or set `ELEVENLABS_VOICE_ID` in `.env` | | `Failed to extract audio` | Check FFmpeg is installed | | `Failed to transcribe` | Check API key, file size (<3GB) |

---

Evolution

This command evolves through use. If something's awkward or missing:

**Local improvements:** 1. Say "improve this" → Claude captures in `_internal/BACKLOG.md` 2. Edit `.claude/commands/redub.md` → Update `_internal/CHANGELOG.md`

**Remote contributions:**

• Issues: `github.com/digitalsamba/claude-code-video-toolkit/issues`
• PRs welcome for multi-speaker support, timing sync, etc.