MODELS

Higgs Audio v3 TTS

by Boson AI

modelsource:boson.aitext-to-speechttsvoice-aivoice-agentsmultilingualvoice-cloningstreamingemotion-controlprosody-controlBoson AI

Visit Website Documentation GitHub

Overview

4B conversational text-to-speech model for low-latency voice agents with multilingual speech, voice cloning, and inline emotion/style control.

Details

Higgs Audio v3 TTS is Boson AI’s 4B conversational text-to-speech model for expressive voice AI. Official Boson sources describe it as built for voice-chat and low-latency voice agents, with expressive conversational speech across 100+ languages, zero-shot/reference-audio voice cloning, streaming speech endpoint examples, and inline text controls for emotion, style, prosody, pauses, pitch, speed, expressiveness, and sound effects. The documented API model name is `higgs-audio-v3-tts`.

When to Use

Use for voice-agent or voice-chat experiences that need low-latency expressive conversational speech rather than plain read-aloud TTS. Use when a TTS workflow needs multilingual output reference-audio voice cloning and inline control tokens for emotion style pauses pitch speed prosody or sound effects. Evaluate when you want an officially documented API model with a public demo docs Hugging Face model page and GitHub project links.

Getting Started

Read the official overview at https://docs.boson.ai/models/higgs-audio-tts/overview and note the API model name `higgs-audio-v3-tts`.
Try the official demo at https://www.boson.ai/demo/tts to assess speech quality and controllability.
Review Boson’s tags documentation at https://docs.boson.ai/models/higgs-audio-tts/tags for supported inline emotion
style
sound-effect
speed
pause
pitch
and expressiveness controls.
Inspect the GitHub repository at https://github.com/boson-ai/higgs-audio or the Hugging Face page at https://huggingface.co/bosonai/higgs-audio-v3-tts-4b before production evaluation.

Key Features

•4B conversational text-to-speech model for expressive real-speech style output.
•Supports expressive conversational speech across 100+ languages
•according to Boson’s announcement sources.
•Zero-shot/reference-audio voice cloning is described in official Boson docs and launch materials.
•Inline text tags can control emotion
•style
•vocalized sound effects
•speed
•pauses
•pitch
•prosody
•and expressiveness.
•Streaming speech endpoint examples are documented for the API model `higgs-audio-v3-tts`.

Capabilities

•text-to-speech
•streaming speech generation
•multilingual speech synthesis
•voice cloning
•emotion and style control
•prosody control
•voice-agent audio generation

Last updated Jun 14, 2026