MisoLabsAI/MisoTTS

Miso TTS is an 8 billion, highly emotive text-to-speech model

What it solves

Miso TTS 8B is designed for high-quality, conversational English speech generation. It allows users to convert text into natural-sounding dialogue, with the ability to condition the output on prior audio for voice cloning.

How it works

The model uses an RVQ Transformer architecture inspired by the Sesame CSM architecture. It consists of two main components:

  • A large Llama 3.2-style backbone (8B parameters) that processes text and audio-frame embeddings.
  • A smaller autoregressive audio decoder (300M parameters) that predicts higher-order audio codebooks within each frame.

It utilizes the Mimi audio tokenizer and generates audio codes from text and optional audio context, allowing it to maintain conversation history for more natural speech.

Who it’s for

Developers and researchers who need a high-fidelity text-to-speech system capable of conversational dialogue and voice cloning, and who have access to high-VRAM GPUs (e.g., 24GB+ VRAM for bfloat16 precision).

Highlights

  • Conversational Focus: Specifically designed for dialogue generation rather than simple reading.
  • Voice Cloning: Supports prompted generation by conditioning on existing audio samples.
  • Watermarking: Includes default watermarking for safety and authenticity.
  • Llama-based Architecture: Leverages a large-scale transformer backbone for high-quality output.

Related

  • Project
    OpenBMB/VoxCPM

    VoxCPM is a tokenizer-free, diffusion-based Text-to-Speech system that provides studio-quality multilingual speech generation, creative voice design, and high-fidelity voice cloning.

  • Project
    OpenMOSS/MOSS-TTS-Nano

    A multilingual, tiny speech generation model with 0.1B parameters designed for real-time, CPU-based inference and high-fidelity voice cloning.

  • Project
    resemble-ai/chatterbox

    A family of open-source text-to-speech models providing low-latency voice cloning and multilingual speech generation across various model sizes for GPU and CPU deployment.

  • Project
    fishaudio/fish-speech

    A state-of-the-art multilingual text-to-speech system that uses a Dual-AR architecture to provide realistic voice cloning and fine-grained emotional control via natural language tags.