Models

Revolab offers five models — three TTS (Nada) and two STT (Aisyah) — at different quality and latency tiers. List them programmatically with GET /v1/models (OpenAI list shape — client.models.list() works).

TTS models (Nada)

Nada models synthesize speech from text. Pass the model ID as the model field in POST /v1/tts.

nada-1.0-flash-lite

Ultra-fast synthesis for low-latency applications.

Best for interactive applications where speed matters more than voice quality.

nada-1.0-flash

Balanced speed and quality for most applications.

Default TTS model. Recommended starting point for most integrations.

nada-1.0-pro

Studio-quality voice synthesis for premium applications.

Best for voiceovers, audiobooks, and branded voice experiences.

STT models (Aisyah)

Aisyah models transcribe audio to text. Pass the model ID as the model form field in POST /v1/stt.

aisyah-1.0-flash

Fast, accurate transcription for most use cases.

Default STT model. Best for real-time applications.

aisyah-1.0-pro

Highest accuracy transcription for critical workflows.

Best for medical, legal, or noisy-environment transcription.

Choosing a model

• Interactive / real-time apps (chatbots, IVR): use nada-1.0-flash-lite or aisyah-1.0-flash — low latency.
• General production apps: use nada-1.0-flash or aisyah-1.0-flash — good balance of speed and quality.
• Premium content (audiobooks, voiceovers, legal transcription): use nada-1.0-pro or aisyah-1.0-pro — highest quality.
• Rate Limits & Quotas — per-key RPM limits and how to handle 429 responses.