Fano Speech API

Fano Speech API is an API-first Speech-to-Text service for teams building reliable voice workflows on AWS. It supports both Streaming (real-time) STT and Async (batch) STT, so you can transcribe live audio streams and process recordings at scale using one consistent platform.

Streaming Speech-to-Text (Released Dec 2025)

Transcribe audio in real time with low latency, ideal for

• Live captioning and real-time transcription;
• Voice agents and conversational interfaces;
• Immediate command recognition and in-the-moment workflows

Key capabilities include

1. Seamless Multilingual Code-switching that acurately transcribes Cantonese, Mandarin, English within a single audio stream, without manually switching language codes;
2. Keyword biasing provides a list of keywords (brand names, product terms, industry jargon) to boost recognition accuracy for your domain;
3. Automatic punctuation: Output includes intelligent punctuation so transcripts are immediately readable and ready for downstream use.

Async Speech-to-Text (Released Nov 2025)

Process recordings efficiently with asynchronous jobs, optimized for scale and operational stability.

Key capabilities include:

1. High-accuracy monolingual models: Dedicated models optimized for English, Mandarin, and Cantonese for single-language audio;
2. Multilingual ASR model (yue-x-auto): Designed for Hong Kong and Singapore language mixing, supports Cantonese/Mandarin/English code-switching without manual language tags;
3. Default speaker diarization: Automatically labels up to 8 speakers, optimized for call logs, meetings, interviews, and multi-speaker recordings;
4. Keyword biasing: Improve accuracy on domain terminology by passing a keyword list in the request;
5. Webhook callback for completed jobs (Enhanced Dec 2025): Receive a callback notification when transcription is finished, no need to poll job status repeatedly.

Common use cases include:

1. Contact center transcription, QA, and analytics;
2. Real-time agent assist and voicebot experiences;
3. Meeting and interview transcription with speaker separation;
4. Compliance and audit workflows that need readable, structured transcripts;
5. Encourage correct recognition of product names and jargon via keyword biasing