Classai Router · Voice

Indonesian voices that sound human.

Turn text into natural Indonesian speech, clone your own voice with consent, and plug it into any app with an OpenAI-compatible API — paid with the same credits as your AI models.

Selina · Female · Narration

“Malam itu hujan turun pelan. Di ujung gang, sebuah warung kopi masih menyala, dan Pak Darto menunggu pelanggan terakhirnya.”

Press play and watch the orb — it listens to the real audio. Drag it, tap it, scroll.

Drag to spin · tap to hear it

Listen

One text box, five ways to speak.

Real, unedited output from our calibrated voices — numbers, Rupiah and English loanwords included. Tap a card to play.

Try it — no sign-up

Type anything. See exactly how it will be read.

Rupiah, dates, phone numbers, “AI”, “ChatGPT”, “yg”, “anak2” — rewritten the way Indonesians actually say them, before any voice speaks. Live, in your browser.

You type

The voice reads

Promo seratus empat puluh sembilan ribu rupiah cuma sampai tujuh belas Oktober dua ribu dua puluh enam jam sembilan belas lewat tiga puluh we i be!

Why Classai Voice

Made for Indonesian, from the first syllable.

Indonesian that sounds right

Numbers, Rupiah, dates, times, abbreviations and terms like AI or ChatGPT are spelled out the Indonesian way before the voice speaks. Add your own words to a pronunciation dictionary.

Calibrated voices

Every voice is tested on hard Indonesian text and scored on articulation and pace. The best ones come pre-tuned, so the first take already sounds clean.

Speaking styles

Narration, ad, casual, news or education in one tap — pace, stability and expression adjust for every engine. Expressive tags like [whispers] or [laughs] on Eleven v4.

Clone your own voice

Record 15–30 seconds and a consent sentence, then use your voice in the studio or through the API. Consent recordings are kept as proof.

Several engines, one place

Compare engines side by side — even blind — and pick the one that fits your content and budget.

OpenAI-compatible API

POST /v1/audio/speech with your existing Classai Router key. The OpenAI SDK works as is; mp3 or wav back in seconds.

Pricing

Pay per minute of audio, from 5.3M credits.

No subscription. Billed per character from your credits (±1,000 characters ≈ 1 minute); failed requests are free.

EngineCredits / minute
VoxCPM2OpenBMB5.3M credits+6.7M credits if the GPU is asleep
dots.ttsrednote5.3M credits+6.7M credits if the GPU is asleep
Eleven v4ElevenLabs28M credits
Eleven v4 TurboElevenLabs14M credits
Eleven Flash v2.5ElevenLabs14M credits
Eleven Multilingual v2ElevenLabs28M credits
OpenAI TTSOpenAI9.4M credits
OpenAI TTS HDOpenAI18.9M credits

Credit packs — the same credits work for AI models

Pro

250M credits

$0.02 / 1M credits

To try every model, or for light use.

  • All active models, all API keys
  • Valid 30 days from payment

One-time payment

$4.99

Buy Pro

Max

Save 25%

500M credits

$0.015 / 1M credits

Regular daily use, cheaper per credit.

  • All active models, all API keys
  • Valid 30 days from payment

One-time payment

$7.49

Buy Max

Ultra

Best value

1 billion credits

+ Bonus 100M credits

$0.012 / 1M credits · save 40% vs Pro

The best value per credit for heavy use.

  • All active models, all API keys
  • Valid 30 days from payment

One-time payment

$11.99

Buy Ultra

Voice cloning

Your voice, with your permission.

  1. 1Record 15–30 secondsIn a quiet room, or upload a clean recording you already have.
  2. 2Record the consent sentenceThe voice owner reads one sentence. We keep it as proof — no consent, no clone.
  3. 3Use it anywhereIn the studio, or through the API with "voice": "clone:<id>". The clone fee is charged once per engine.

Imitating public figures or anyone without permission is prohibited.

For AI agents & developers

Built for AI agents & developers.

Let your agent make speech through MCP, call the OpenAI-compatible API from your code, or hand your coding agent one prompt that wires it into your app.

  • OpenAI-compatible
  • One key for chat, voice, image, video
  • Pay only for what comes out
  • llms.txt + OpenAPI
  • MCP server

Connect once and your agent can make speech itself — in Claude Code, Claude Desktop, Cursor, opencode or any MCP client. Streamable HTTP, authenticated with your sk_live_ key.

Tools

speaklist_voicesgenerate_imagestart_videoget_videolist_media_models

Server URL

https://router.classai.id/mcp
claude mcp add --transport http classai https://router.classai.id/mcp \
  --header "Authorization: Bearer sk_live_…"

FAQ

Questions, answered.

Something else? Contact our support.

Can I use the audio commercially?+

Yes — ads, videos, courses, podcasts and apps. Tell listeners the voice is AI-generated, and never use it to imitate someone without their permission or to deceive.

How is it billed?+

Per character, from your Classai Router credits (about 1,000 characters ≈ 1 minute of audio). Failed requests aren't charged. Credits are valid for 30 days per purchase. Token packages for coding models don't cover voice.

Is it only Indonesian?+

It's built for Indonesian: the pronunciation engine, the calibration and the curated voices all focus on it. Many voices are multilingual and read English fine too.

How does voice cloning work?+

Upload or record a clean 15–30 second sample, then the voice owner records a short consent sentence. The clone fee is charged once per engine, the first time you use the voice. Cloning other people's voices without permission is prohibited.

Do I need a separate API key?+

No. The same sk_live_ key and base URL you use for AI models work for voice. List engines and voices with GET /v1/audio/voices.

Can I try before paying?+

Every demo on this page is free to play, and so are the voice previews in the studio. Generating your own audio uses credits — packs start small.

Give your content a voice today.

Type, pick a voice, press play. Your first audio is a minute away.

Create a free account