Models

Async offers three text-to-speech models, each tuned for different use cases. All models are available via the Streaming HTTP and WebSocket endpoints.

Model Overview

	Async Pro v1.0	Async Flash v1.5	Async Flash v1.0
Model ID	`async_pro_v1.0`	`async_flash_v1.5`	`async_flash_v1.0`
Best for	Highest-quality, low-latency English speech	Low-latency streaming with built-in text normalization	Real-time multilingual applications
Languages	English	English, Spanish, French, German, Italian, Portuguese	English, French, Spanish, German, Italian, Portuguese, Arabic, Russian, Romanian, Japanese, Hebrew, Armenian, Turkish, Hindi, Chinese
Supported endpoints	Streaming, WebSocket	Streaming, WebSocket	All (Streaming, WebSocket, HTTP sync, Timestamps)
Text normalization	Built-in	Built-in	Not available
`speed_control`	Not available	Not available	0.7 – 2.0
`stability`	Not available	Not available	0 – 100

Async Pro v1.0

Our highest-quality English TTS model, designed for natural speech generation with strong non-standard text handling and fast inference speed. Optimized for low-latency streaming, making it equally suitable for real-time conversational applications.

Model ID: async_pro_v1.0

Languages: English

Endpoints: Streaming (POST /text_to_speech/streaming), WebSocket (WSS /text_to_speech/websocket/ws)

Text normalization: Automatically handles dates, currencies, numbers, and abbreviations — no preprocessing needed

Best for: Voice assistants, conversational AI, audiobooks, narration, podcasts, and any use case where English voice quality is the top priority

Async Flash v1.5

A latency-optimized streaming TTS model supporting 6 languages, with strong built-in handling of non-standard text such as dates, currencies, numbers, and abbreviations.

Model ID: async_flash_v1.5

Languages: English, Spanish, French, German, Italian, Portuguese

Endpoints: Streaming (POST /text_to_speech/streaming), WebSocket (WSS /text_to_speech/websocket/ws)

Text normalization: Automatically handles dates, currencies, numbers, and abbreviations — no preprocessing needed

Best for: Real-time voice assistants, chatbots, and conversational AI across European languages

Async Flash v1.0

Our legacy low-latency multilingual TTS model, supporting 15 languages and optimized for fast real-time speech generation.

Model ID: async_flash_v1.0

Languages: English, French, Spanish, German, Italian, Portuguese, Arabic, Russian, Romanian, Japanese, Hebrew, Armenian, Turkish, Hindi, Chinese

Endpoints: All — Streaming, WebSocket, HTTP sync (POST /text_to_speech), Timestamps (POST /text_to_speech/with_timestamps)

Unique parameters:

speed_control (0.7 – 2.0) — Adjusts the speaking speed of the synthesized voice

stability (0 – 100) — Adjusts how stable or expressive the synthesized voice sounds

Best for: Applications requiring broad language coverage or synchronous audio generation

Choosing the Right Model

Need the best English voice quality?
Use async_pro_v1.0. It produces the most natural-sounding speech, handles non-standard text out of the box, and is fast enough for real-time streaming use cases.

Need low latency across European languages?
Use async_flash_v1.5. It combines fast streaming with built-in text normalization for 6 languages.

Need broad multilingual support?
Use async_flash_v1.0. It covers 15 languages and is the only model available on the synchronous HTTP and timestamps endpoints.

Important Notes

async_pro_v1.0 and async_flash_v1.5 are streaming and WebSocket only. They are not available on the synchronous POST /text_to_speech or POST /text_to_speech/with_timestamps endpoints.

The speed_control and stability parameters are only supported by async_flash_v1.0. They have no effect on the other models.

All models support the same voice library — you can use any voice with any model.

All models support custom phonemes, digit pronunciation, and silent pauses.

Model Overview#

Async Pro v1.0#

Async Flash v1.5#

Async Flash v1.0#