← All models
A unified speech-language model that synchronizes speech and text into a single, cohesive stream via 1:1 alignment.
Details
- Accepts
- audio
Pricing
- Price
- 0.010 cr / character
Prices in credits (1 credit = $0.01).
Data schema
Input
| Field | Type | Description |
|---|---|---|
| top_p | number | Top-p (nucleus) sampling parameter for text generation. |
| promptrequired | string | The text to synthesize into speech using the reference speaker's voice. |
| language | stringenum: en, ar, ch, de, es, fr… | Language for text alignment. Use the appropriate code for non-English synthesis. |
| audio_urlrequired | string | |
| transcript | string | Transcript of the reference audio. For non-English audio, providing a transcript is required since the built-in ASR is English-only. |
| temperature | number | Sampling temperature for text token generation. Must be greater than 0. Higher values produce more varied output. |
| output_format | stringenum: wav, mp3 | The format of the output audio file. |
| num_extra_steps | integer | Number of extra autoregressive steps for speech continuation beyond the input text. Useful for generating trailing prosody or silence. |
| speed_up_factor | number | Factor to speed up or slow down the generated speech. Values > 1.0 speed up, < 1.0 slow down. |
| noise_temperature | number | Temperature for noise in the flow matching diffusion process. |
| acoustic_cfg_scale | number | Classifier-free guidance scale for acoustic feature generation. |
| repetition_penalty | number | Penalty applied to repeated tokens during generation. |
| num_inference_steps | integer | Number of ODE solver steps for flow matching acoustic generation. More steps improve quality at the cost of speed. |
Output
| Field | Type | Description |
|---|---|---|
| audio | — | The generated speech audio file. |