You're looking at a specific version of this model. Jump to the model overview.
irulkenzei /voice-over-app-cloud:f12eacee
Input schema
The fields you can use to run this model with an API. If you don’t give a value for a field its default value will be used.
| Field | Type | Default value | Description |
|---|---|---|---|
| text |
string
|
|
Text to synthesize into speech (single-speaker mode, leave empty if using 'script')
|
| speaker_wav |
string
|
Reference audio file for voice cloning (single-speaker mode, required if 'text' is used)
|
|
| script |
string
|
|
Multi-speaker dialogue script (dialogue mode). Format per line: '[Speaker Name]: text...'. A line without a tag is treated as a continuation of the previous speaker. If filled in, 'text'/'speaker_wav' are ignored -- use 'speaker_map' to assign a voice to each name instead.
|
| speaker_map |
string
|
{}
|
JSON string mapping speaker name -> speaker_wav URL, e.g. '{"Dad": "https://.../dad.wav", "Son": "https://.../son.wav"}'. Required if 'script' is used.
|
| speaker_pause_ms |
integer
|
500
Max: 3000 |
Automatic pause (ms) between each speaker turn in dialogue mode
|
| language |
string
|
en
|
Language code (e.g. 'en', 'es', 'fr'). Note: 'id' (Indonesian) automatically falls back to 'ms' (Malay) if the model doesn't support it directly -- see _LANGUAGE_FALLBACKS.
|
| exaggeration |
number
|
0.5
Max: 1 |
Emotion/expressiveness intensity (0.0 - 1.0). Higher values produce more expressive, exaggerated speech
|
| cfg_weight |
number
|
0.5
Max: 1 |
Classifier-free guidance weight (0.0 - 1.0+). Higher values increase adherence to the reference voice/conditioning; lower values (e.g. ~0.3) can improve pacing for fast-speaking reference voices
|
| speed |
number
|
1
Min: 0.5 Max: 2 |
Speech playback speed (0.5 - 2.0). Applied as pitch-preserving post-processing (ffmpeg atempo), since the model itself has no native speed control
|
| temperature |
number
|
0.7
Min: 0.1 Max: 1 |
Sampling temperature (0.1 - 1.0)
|
| output_format |
None
|
wav
|
Output audio file format
|
| chunk_by_punctuation |
boolean
|
True
|
Automatically split text at '.'/',' with manual silence (reduces drift/hallucination on long text)
|
| comma_pause_ms |
integer
|
300
Max: 5000 |
Pause duration (ms) after a comma
|
| period_pause_ms |
integer
|
600
Max: 5000 |
Pause duration (ms) after a period
|
| background_music |
string
|
Optional background music file. Automatically ducked (volume temporarily lowered) whenever the narration is speaking, and raised back smoothly during silence -- automatic audio ducking.
|
|
| music_volume_db |
number
|
-6
Min: -30 |
Base background music volume in dB relative to the original file (negative = quieter than the original). The music will automatically duck further while the narration is speaking, then return to this level during silence.
|
Output schema
The shape of the response you’ll get when you run this model with an API.
{'format': 'uri', 'title': 'Output', 'type': 'string'}