Text to speech

Product

POST /v1/tts

Voices that pronounce Indian names correctly, switch scripts mid-sentence, and sound the same next month as they do today. One published version, so there is no tier to choose wrong.

TRY IT IN THE BROWSER READ THE DOCS

Models

Version	Aliases	Rate / 1M chars
orphix-multilingual-1.0.1 The only published version — every public voice runs on it.	echo-1	$0.0189 / min

Aliases move when a new version ships. Exact versions never do — pin one to a key and that key is frozen until you change it.

Languages

78 voices, unevenly — and that is the useful part.

A flag row would imply ten equal libraries. English and Hindi carry most of the catalogue, which is worth knowing before you build a Portuguese IVR on two voices.

Language	Voices	Accents
English en	44	American · Indian · British · Australian
Hindi hi	18	Indian · Gujarati · Bengali · Marathi
Spanish es	4	Latin American · Peruvian
Tamil ta	3	Indian
Korean ko	2	Seoul
Arabic ar	2	Saudi · Jordanian
Portuguese pt	2	Brazilian
Italian it	1	Tuscan
French fr	1	African
Chinese zh	1	Beijing Mandarin

The model itself understands thirteen — German, Dutch, Japanese and Russian have no public voice yet, so reaching them today means cloning one.

Controls

text	1 – 5 000 characters, whitespace collapsed
voice_id	78 public voices, plus your org's own
speed	0.25 – 4.0× · out of range is rejected, not clamped
seed	any integer · same seed, same bytes
output_format	wav_<rate> · pcm_<rate> · mulaw_<rate>

✓ mulaw_8000 for telephony✓ pcm for streaming into your own mixer✓ mp3 and opus are not available yet

Example

One request, one wav file.

import requests

r = requests.post(
    "https://api.vocalix.ai/v1/tts",
    headers={"Authorization": f"Bearer {KEY}"},
    json={
        "text": "आपका parcel आज शाम तक deliver हो जाएगा।",
        "voice_id": "anika-hi",
        "model": "echo-1",
        "output_format": "wav_22050",
        "speed": 1.0,
        "seed": 42,
    },
)
r.raise_for_status()
open("reply.wav", "wb").write(r.content)
print(r.headers["X-Vocalix-Model-Version"])   # echo-1.0.0