Skip to main content

Voice API (TTS)

The Dhamma.Gift voice is a text-to-speech service: the same one that reads in the site's player and the dictionary. It can be called from a browser or a server, no key, no login.

Terms of use
  • pratham, priyamvada, alan, kathleen, irina, puriso may be used freely, within the limits below and each voice's own licence.
  • dg, dgru, dgen (Cūḷa DG, Yojanāni DG) are made from recordings of real people's voices. All rights reserved. They may be used only on the Dhamma.Gift website and in the Dhamma.Gift apps. Any other use through this API — personal and non-commercial included — requires prior written permission from Dhamma.Gift.

Address: https://api.dhamma.gift · CORS is open · a limit of 60 requests per minute per IP.

Read a phrase​

POST /api/tts/pali with JSON:

FieldMeaning
textthe text, up to 2000 characters. Pāḷi in IAST (Evaṁ me sutaṁ), a translation in English or Russian
voicethe voice (below); pratham by default
ratespeed, 0.25–3; 1 is the voice's own pace (the site reads Pāḷi at 0.875)

The answer is {"audioContent": "<base64 MP3>"} — the same shape as Google's Text-to-Speech, so a player written for Google plays it as is. MP3, mono, 22 kHz, ~40 kbit/s; repeated requests come from a cache.

curl -s https://api.dhamma.gift/api/tts/pali -H 'content-type: application/json' \
-d '{"text":"Evaṁ me sutaṁ.","voice":"pratham","rate":0.875}' \
| python3 -c 'import json,sys,base64; open("evam.mp3","wb").write(base64.b64decode(json.load(sys.stdin)["audioContent"]))'
const r = await fetch('https://api.dhamma.gift/api/tts/pali', {
method: 'POST', headers: { 'content-type': 'application/json' },
body: JSON.stringify({ text: 'Evaṁ me sutaṁ.', voice: 'pratham', rate: 0.875 })
});
new Audio('data:audio/mp3;base64,' + (await r.json()).audioContent).play();

Voices​

voiceLanguageWhat it is
prathamPāḷi"Pratham": Piper neural voice + our Pāḷi pronunciation rules (IAST → sounds, stress, long vowels). Default
dgPāḷi"Cūḷa DG": a voice trained on real Pāḷi readings. Beta
priyamvadaPāḷi"Priyamvada": a female Piper voice with the same rules
dgen, alan, kathleenEnglishPiper voices. dgen is "Yojanāni DG" (the default): trained on a real person's readings. Pāḷi words and abbreviations (Ven., SN 56.11) are read properly
puriso, dgru, irinaRussianPiper voices. dgru is "Cūḷa DG" (the default); puriso is Supertonic's male voice M1 taught to Piper. Pāḷi in Latin script is read the Pāḷi way, Russian words with the right stress

GET /api/tts/dg-voices gives the list as {"voices": [{"id", "lang", "label"}]} in menu order (a language's first voice is its default; Yojanāni DG is listed there by its internal id miles, which works the same as dgen); GET /api/tts/health shows that the service is up and which version of the pronunciation rules it uses. GET /api/tts/offline/pali-ipa.json gives the rules, the engine and the voice models that DG downloads to read on the device, offline.

How Pāḷi is read: every letter is sounded, punctuation becomes pauses, …pe… is read as "peyyāla".

A whole text as one MP3​

POST /api/tts/memo — the lines read one after another with pauses of any length between them (Google stops at 10 seconds); this is what the Memo page saves.

FieldMeaning
segmentsthe lines, up to 200 and 20 000 characters in all
voice, rateas above
delaythe pause between lines, seconds, up to 3600
end_delaythe pause at the end, seconds, up to 3600
soundgong.mp3 or tick.mp3 at the end, or empty

The answer is the MP3 itself (audio/mpeg): the speech itself at most 30 minutes, the whole file with its pauses at most 180. One such request counts as ten against the per-minute limit.

Licences​

  • The service code and the Pāḷi rules: github.com/dhammagift/pali-tts.
  • pratham: weights MIT (rhasspy/piper-voices), recordings AI4Bharat IndicVoices (CC BY 4.0).
  • dg, dgru, dgen: all rights reserved; only on the Dhamma.Gift website and apps, anything else with prior written permission (see Terms of use above).
  • priyamvada, alan, kathleen, irina: rhasspy/piper-voices, each under its own licence.

Also at this address​

The same host serves the MCP server (/mcp) and the site's API (/api/…), with the same texts and database as dhamma.gift.