Speech Synthesis API

Generate speech audio from text using ASAT v2, short for ASAT Speaks Any Text. It is optimized around Moroccan Arabic and Arabic speech generation.

Endpoints

Canonical endpoint:

POST https://api.sawalni.com/v1/audio/speech

Sawalni-style alias:

POST https://api.sawalni.com/v1/speech/synthesize

Request Body

Send an application/json request.

The current service does not expose NFE steps, speed, CFG, or chunk-size controls. Do not send those fields; the current server ignores extra request fields.

Streaming and Formats

The service keeps one speech session for each request. stream: true delivers output chunks as they become available. It does not select a separate quality path. Use response_format: "pcm16" for raw signed 16-bit little-endian PCM at 24 kHz, mono. For a streamed WAV, the initial RIFF header has placeholder lengths. Rewrite the RIFF and data lengths after the stream ends before saving a standalone WAV file. The playground does this for you.

Example — curl

curl -X POST https://api.sawalni.com/v1/audio/speech \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "asat",
    "voice": "male-1",
    "response_format": "wav",
    "input": "السلام، كيف داير؟ هادي تجربة ديال الصوت بالدارجة المغربية."
  }' \
  --output asat.wav

Example — Python

import requests

response = requests.post(
    "https://api.sawalni.com/v1/audio/speech",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "model": "asat",
        "voice": "male-1",
        "response_format": "wav",
        "input": "السلام، كيف داير؟",
    },
)
response.raise_for_status()
open("asat.wav", "wb").write(response.content)
print("Duration:", response.headers.get("x-audio-duration-s"))

Example — JavaScript

const response = await fetch('https://api.sawalni.com/v1/audio/speech', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'asat',
    voice: 'male-1',
    response_format: 'wav',
    input: 'السلام، كيف داير؟',
  }),
});

const audio = await response.blob();
new Audio(URL.createObjectURL(audio)).play();
console.log('Duration:', response.headers.get('x-audio-duration-s'));

Response

The response body is audio/wav or raw 24 kHz mono signed 16-bit PCM. Metadata is returned in response headers. Some fields, including duration and processing time, are only available for buffered responses. For streams, calculate duration from PCM byte count at 48,000 bytes per second.

Supported Voices

Retrieve the current machine-readable catalog with GET https://api.sawalni.com/v1/voices. The voice request value must be one of these IDs:

The service uses the selected voice for speech generation.

Model Details

Error Handling

Pricing

Speech synthesis is metered by generated audio duration.