API Reference

Endpoint VoiceStudio

Overview

Endpoint	Method	Fungsi
/health	GET	Cek status backend
/engines	GET	Daftar engine TTS
/v1/audio/voices	GET	Daftar voice (bawaan + clone)
/profiles	GET/POST	Daftar / tambah profile clone
/profiles/{id}	GET/DELETE	Detail / hapus profile clone
/v1/audio/speech	POST	Sintesis TTS
/engines/select	POST	Pilih engine aktif

1. Cek status

# Response
{"status":"ok","device":"cpu","version":"0.5.7"}

2. Daftar engine

# Response (ringkas) — hanya 2 engine yang tersedia
{"tts":{"backends":[
  {"id":"omnivoice","available":true,"supports_cloning":true,"reason":null},
  {"id":"omnivoice-subprocess","available":true,"supports_cloning":true,"reason":null}
]}}

• id → nama engine, dipakai di kolom model TTS.
• available → true = siap pakai.
• supports_cloning → true = bisa clone suara.

3. Daftar voice

# Response (ringkas)
{"voices":[
  {"voice_id":"alloy","name":"alloy","type":"openai_alias","language":"id","sample_rate":24000},
  {"voice_id":"60e831c4","name":"suara_kak","type":"profile","language":"id","sample_rate":24000}
]}

• type:"openai_alias" → voice bawaan (alloy, ash, …).
• type:"profile" → voice clone buatan sendiri.

4. TTS / Sintesis

# Request (JSON)
curl -X POST https://celsyotk.tailba6639.ts.net/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
    "model": "omnivoice",
    "input": "Halo, ini contoh suara.",
    "voice": "alloy",
    "response_format": "mp3",
    "speed": 1.0
  }' --output hasil.mp3

Field	Wajib	Keterangan
model	ya	omnivoice atau omnivoice-subprocess
input	ya	Teks yang diucapkan
voice	ya	voice_id dari /v1/audio/voices
response_format	tidak	mp3 (default) | wav | opus | flac | aac
speed	tidak	0.8 – 1.2 (default 1.0)

5. Tambah voice clone

# Request (multipart)
curl -X POST https://celsyotk.tailba6639.ts.net/profiles \
  -F "name=suara_baru" \
  -F "kind=clone" \
  -F "language=id" \
  -F "ref_text=HALO ANAK-ANAK KU! ALAKAHUNA SYUDUDU!" \
  -F "ref_audio=@/path/ke/ref.wav;type=audio/wav"

Field	Wajib	Keterangan
name	ya	Nama voice clone
kind	ya	Wajib clone
ref_audio	ya	File audio referensi (wav/mp3/flac, min ±3 detik)
ref_text	tidak	Transkrip audio (meningkatkan akurasi clone)
language	tidak	Kode bahasa, mis. id

# Response (ringkas)
{"profile_id":"65ab12cd","name":"suara_baru","kind":"clone","language":"id","created_at":...}

6. Hapus voice clone

# Request
curl -X DELETE https://celsyotk.tailba6639.ts.net/profiles/65ab12cd

# Response
{"deleted":"65ab12cd"}

7. Daftar / detail profile

8. Pilih engine (opsional)

# Request
curl -X POST https://celsyotk.tailba6639.ts.net/engines/select \
  -H "Content-Type: application/json" \
  -d '{"backend_id":"omnivoice"}'