Three Voice Modes
DeepVideo provides three voice source modes for translated videos
Mode 1
Voice Clone
Extracted vocals from the original video. The TTS engine replicates the original speaker’s voice characteristics in the target language.
- Same-language dubbing
- Preserve original speaker’s voice
- Automatic extraction from video
Mode 0
Built-in Role
System voice library with 358+ professional voices per language. Ideal for cross-language translation with consistent quality.
- 358+ voices per language
- 10 supported languages
- Gender-aware auto-selection
Mode 0
Custom Role
User-uploaded reference audio for branded voice. Perfect for specific tone/style requirements and brand consistency.
- Upload your own reference audio
- MP3 / WAV / M4A supported
- Branded voice identity
Cross-language restriction: If
voice_clone_cross_language = false in config, clone mode is automatically disabled for cross-language projects (forced back to mode 0).
Voice Roles Management
Manage your custom voice roles and browse the system voice library
▸ Custom Voice Roles
▾ System Voice Roles
zh-001
Female
zh-002
Male
en-042
Female
ja-007
Male
▾ ByteDance Online Voices (Pro TTS only)
Operations per Custom Role
| Button | Action | API |
|---|---|---|
| ▶ Preview | Play reference audio | GET /custom-voice/audio |
| Enable / Disable | Toggle role availability | PATCH /custom-voice/{lang}/{name} |
| 🗑 Delete | Remove role permanently | DELETE /custom-voice/{lang}/{name} |
Creating a Custom Voice Role
Upload your reference audio to create a branded voice
📁 Click to select audio file…
Reference text: “The quick brown fox jumps over the lazy dog. This is a sample text for voice recording. Please read naturally at a comfortable pace.”
System Voice Roles (Built-in)
358+ professional voices
