MisoTTS

MisoTTS is an 8-billion parameter text-to-speech model that generates emotive conversational speech.

Miso TTS 8B is an 8 billion parameter text-to-speech model designed for high-quality conversational speech generation. It features an RVQ Transformer architecture inspired by the Sesame CSM architecture, using a large Llama 3.2-style backbone and a smaller autoregressive audio decoder. The model generates Mimi audio codes from text and optional audio context for voice cloning. It currently supports English only and includes audio watermarking via SilentCipher. The repository contains inference code, model definitions, and setup instructions for running the model locally.

MisoTTS at a glance

Best for: Social media creator, Writer
Runs on: Web, self-hosted, open source
Pricing: Free
Released: May 2026
Category: AI audio tool
Website: github.com/MisoLabsAI/MisoTTS

Key features

Generates conversational speech from text
Conditions on prior audio for voice cloning
Uses an RVQ Transformer architecture
Applies automatic audio watermarking
Processes interleaved text and audio tokens

Use it for

Generating conversational speech from text
Cloning voices using prior audio context
Running local speech generation inference

Alternatives to MisoTTS

VoxCPM — Provides tokenizer-free multilingual speech with controllable voice creativity.
Fish Audio — Provides studio-grade voice cloning and text-to-speech with emotion control.
MOSS-TTS — Delivers zero-shot voice cloning and fine-grained pronunciation control.
Eleven v4 and Eleven v4 Turbo — Generates low latency expressive speech with real time streaming.
GPT-SoVITS — It allows high-fidelity voice cloning with as little as one minute of data.
IndexTTS — Supports zero-shot voice cloning with multi-language emotion and speed control.
AuK — It handles zero-shot TTS, editing, and enhancement via natural language.
Superwhisper — Transcribes speech offline and online with custom AI modes and templates.
Voicebox — Perform voice cloning and generation entirely on local hardware.
Applio — Runs locally with open-source code and no telemetry.
DiffSinger — It uses a shallow diffusion mechanism for singing voice synthesis and text-to-speech.
Respeecher — Engineered for professional production workflows with ethical standards.

All MisoTTS alternatives