Text to Speech
Synthesize natural Arabic speech — voices, quality tiers, blocking, streaming, and realtime.
توليد الكلام
حوّل نصاً عربياً إلى صوت طبيعي، إمّا دفعة واحدة أو ببثّ متدفق يبدأ قبل اكتمال التوليد. تُحتسب الفوترة بعدد أحرف النص المُدخل.
ASAS TTS turns Arabic text into natural speech. You can get the whole clip in one response, stream it as chunked audio for low-latency playback, or drive it over a WebSocket for realtime apps.
Endpoints
| Endpoint | Shape | Use it for |
|---|---|---|
POST /v1/tts | Blocking — returns the full audio clip | Short, non-latency-sensitive synthesis |
POST /v1/tts/stream | Chunked HTTP audio | Low-latency playback (p90 first-audio < 150 ms) |
WS /v1/tts/realtime | WebSocket protocol | Realtime / interactive synthesis |
All three take the same body.
Request
{
"voice": "layan",
"text": "أهلاً بك في أساس",
"speed": 1.0,
"quality": "standard"
}| Field | Type | Notes |
|---|---|---|
text | string | Required. 1–500 characters. Longer input returns 422. |
voice | string | Required. A voice slug from your org's voice registry. |
speed | number | 0.5–2.0. Default 1.0. |
quality | fast | standard | high | Default standard. |
500-character limit. v1 accepts up to 500 characters per request. For longer text, split it into sentences and call in sequence. An Arabic-aware chunking pipeline that stitches the audio for you is a planned fast-follow (v1.1).
Quality tiers
quality trades latency for fidelity. It also sets the credit rate:
| Quality | Relative latency | Typical credit rate |
|---|---|---|
fast | Lowest | ~0.5 credit / char |
standard | Balanced | ~1 credit / char |
high | Highest fidelity | ~1 credit / char |
Exact rates are visible in your dashboard. Billing is always by input characters, counted before synthesis — a client abort mid-stream still bills the dispatched characters (the audio was already generated), and cached responses bill normally.
Blocking example
curl https://api.asas.flitc.tech/v1/tts \
-H "Authorization: Bearer $ASAS_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "voice": "layan", "text": "أهلاً بك في أساس" }' \
--output speech.wavStreaming example
import { AsasClient } from "@asas/sdk";
const asas = new AsasClient({ apiKey: process.env.ASAS_API_KEY! });
const stream = await asas.tts.stream({
voice: "layan",
text: "مرحباً بك في منصة أساس للصوت العربي",
quality: "fast",
});
// `stream` is a ReadableStream of audio chunks — pipe it to a player or a file.
for await (const chunk of stream) {
process.stdout.write(chunk);
}Voices
List the voices available to your organization from your registry (platform voices plus any of your own):
curl https://api.asas.flitc.tech/v1/voices \
-H "Authorization: Bearer $ASAS_API_KEY"Voice cloning is not enabled for customer keys at launch. When it ships it will be an entitlement with recorded consent and audit logging.