Text to speech and speech to text API
Built for how Southeast Asia actually speaks.
Aisyah understands speech, Nada generates it. Both trained in-house on Bahasa Melayu, Manglish, regional accents and the everyday switching between Malay and English that generic models handle badly.
50 free minutes of speech to text and 50 of text to speech. No card required.
Word error rate, Aisyah 1.0 Pro. Most accurate of the fourteen models tested.
Invented words. The lowest of all fourteen, including models it beats on accuracy.
To first audio from Nada 1.0 Pro. A blink takes about 400.
Real Malaysian utterances. Every model transcribed the same set.
Two families. One understands speech, one produces it.
The accuracy tier, for messy, real-world, code-switched Malaysian calls. Most accurate of the fourteen models tested, and the least likely to invent words that were never said.
- Word error rate
- 5.79%
- Invented words
- 0.62%
- Rank
- 1 of 14
The fast tier, for live transcription where latency matters more than the last decimal point. Still ahead of Gemini 2.5 Flash and Whisper large-v3.
- Word error rate
- 10.03%
- Invented words
- 0.72%
- Rank
- 6 of 14
Malay-first voices that speak a sentence switching between Malay and English without stumbling. Audio starts before the sentence is finished, so a live agent answers without dead air.
- Time to first audio
- 300 ms
- Streaming
- Yes
- Voice cloning
- Yes
The number that breaks automation is not accuracy.
Word error rate is how much of what was said came back wrong. Invented words are text the model produced that nobody spoke, and that is the number that does the damage: it flows straight into call summaries, routing decisions and compliance logs. Aisyah 1.0 Pro has the lowest invented-word rate of all fourteen models, including ones it beats on accuracy. Lower is better in both columns.
| # | Model | Word error rate | Invented words |
|---|---|---|---|
| 1 | Aisyah 1.0 Pro | 5.79% | 0.62% |
| 2 | Scribe v2 | 6.67% | 0.70% |
| 3 | Gemini 2.5 Pro | 7.00% | 1.00% |
| 4 | Qwen-Audio 3.0 ASR Flash | 9.17% | 1.11% |
| 5 | Ilmu ASR v4.2 | 9.48% | 0.87% |
| 6 | Aisyah 1.0 Flash | 10.03% | 0.72% |
| 7 | Gemini 2.5 Flash | 10.76% | 2.48% |
| 8 | Whisper large-v3 | 17.30% | 2.73% |
| 9 | Gemini 3.6 Flash | 24.55% | 14.35% |
| 10 | Nova 3 | 31.99% | 0.94% |
Ten of fourteen models shown. Benchmarked on 1,899 real Malaysian utterances, 820 published and 1,079 held back. Aisyah 1.0 Pro scores 4.89% on the published half and 6.58% on the held-back half, so its accuracy holds up on audio it has never seen. Compare Aisyah with each model, category by category.
Built around Malaysian speech from the start.
Set diarize and every segment comes back tagged with who spoke it. An agent and a customer, or three people on a conference bridge, arrive as separate speakers instead of one run-on block of text.
Code-switching between Malay and English is a first-class setting, not a workaround. Turn on language tags and each segment is labelled with the language it was actually spoken in.
Real call audio is compressed, noisy and narrowband. Denoising is a request flag on the same endpoint, ahead of a voice-activity pipeline that splits long recordings into segments.
Set stream and Nada returns raw PCM as it synthesizes, so a live agent starts speaking before the sentence is finished. Billing meters only the audio actually delivered.
Clone a voice from a short reference sample through a guided wizard, then use it like any premade voice. A pronunciation dictionary lets you fix how specific words are said.
WAV, MP3, M4A, FLAC, OGG, Opus and WebM, up to 50 MB or 30 minutes a request. Give it a language hint or let it detect one.
Keep the SDK you already use.
The interface follows the OpenAI convention, so existing SDKs work by pointing at a different address. The gateway also speaks the ElevenLabs and Cartesia APIs, and LiveKit agents connect through the stock Cartesia plugin.
curl -X POST https://api.revolab.ai/v1/stt \
-H "Authorization: Bearer $REVOLAB_API_KEY" \
-F "model=aisyah-1.0-pro" \
-F "file=@call-recording.wav"Published pricing is coming.
Per-minute rates are being finalised and will be published here. In the meantime every new account starts with free minutes on the flash models, so you can build and test the whole integration before anything is priced. If you need a number against a volume commitment now, we will quote it.
Or run it inside your own walls.
The models already run on our own GPU servers rather than someone else's API, so the same stack can be deployed in your environment when audio is not allowed to leave it. Banks, telcos and government work usually start here.
Talk to the team- Two-factor authentication, required on every account
- API keys scoped to speech to text, text to speech, or both
- Optional per-key expiry, and revocation that takes effect immediately
- Every record in the platform belongs to an organisation
- Your audio is never used for training without an explicit opt-in
- Deployment inside your own environment
- Single sign-on
- Role-based access across your organisation
- Pricing scoped to your volume
Scoped with our team rather than switched on at signup.
Text to speech and speech to text, answered.
Does Revolab have a Malay speech to text API?
Yes. Aisyah is a speech to text API for Malay (Bahasa Melayu, also called Bahasa Malaysia), trained on Malaysian speech including Manglish and the everyday switching between Malay and English. It covers Malay, not Malayalam.
Is there a Malay text to speech API?
Yes. Nada generates Malay-first voices that handle sentences switching between Malay and English, and streams audio back with about 300 ms to first audio. You can clone a voice and fix how specific words are pronounced.
Can it transcribe Manglish and code-switched audio?
Yes. Code-switching between Malay and English is a first-class setting. Turn on language tags and each segment is labelled with the language it was actually spoken in.
How accurate is the Malay speech recognition?
Aisyah 1.0 Pro scored a 5.79% word error rate on 1,899 real Malaysian utterances, the most accurate of the fourteen models tested on that set.
Is there a free tier?
Every new account gets 50 free minutes of speech to text and 50 of text to speech. No card required. Per-minute pricing will be published soon.
Does the speech to text API support speaker diarization?
Yes. Set diarize and every segment comes back tagged with who spoke it, so an agent and a customer arrive as separate speakers instead of one block of text.
Which audio formats can I send?
WAV, MP3, M4A, FLAC, OGG, Opus and WebM, up to 50 MB or 30 minutes per request. Give it a language hint or let it detect one.
Can I keep my existing OpenAI or ElevenLabs SDK?
Yes. The API follows the OpenAI convention and also speaks the ElevenLabs and Cartesia APIs, so existing SDKs work by pointing at a different address. LiveKit agents connect through the stock Cartesia plugin.
Can the models run on-premise?
Yes, for enterprise customers. The models can be deployed inside your own environment under an enterprise agreement.
Run your own audio through it.
Every new account starts with 50 free minutes of speech to text and 50 of text to speech. No card required. If your recordings are Malaysian calls, that is the test worth running.