IndexTTS2 - B station open source free TTS model, the first to support precise duration control

What is IndexTTS2?

IndexTTS2 is a new free text-to-speech (TTS) model open-sourced by the B station voice team, which realizes a major breakthrough in emotional expression and duration control, and is the first autoregressive TTS model that supports precise duration control. Support for zero-sample voice cloning , only one audio file can accurately copy the timbre , rhythm and style of speech , support for multi-language. IndexTTS2 support for emotional timbre separation control , the user can independently specify the source of the timbre and the source of the emotion. The model is equipped with multimodal emotion input, supporting emotion control via emotion reference audio, emotion description text, or emotion vectors.

Features of IndexTTS2

• zero-sample speech cloning: Only one reference audio is needed to accurately replicate vocal lines, intonation, and rhythm, with multi-language support for highly personalized voice synthesis.
• Emotions and Duration ControlIt supports zero-sample emotion cloning, and can control voice emotions based on reference audio or text descriptions. It has the world's first precise duration control function, which meets the needs of film and TV dubbing and timeline synchronization.
• high fidelity sound quality: Audio sampling rate up to 48kHz, support lossless audio output, combined with optimized vocoder, generates natural, smooth and emotional speech with less mechanical feeling.
• Multi-modal input support: It supports multiple input methods, such as text and audio, and allows users to control the style and emotion of the generated speech through text descriptions, reference audio, or emotion vectors.
• Localized Deployment and Open Source: It supports fully localized deployment and plans to open up model weights to provide developers with powerful tools to empower more application scenarios and promote the widespread use of TTS technology.

Core Benefits of IndexTTS2

• Precise duration control function: IndexTTS2 is the first autoregressive TTS model to support precise duration control, specifying the length of generated audio down to the millisecond level.
• Modeling of Emotional Tone Separation: IndexTTS2 enables separate modeling of emotion and timbre, allowing the user to control emotion and timbre independently.
• Multimodal Emotional Input Support: IndexTTS2 supports a variety of ways to control the sentiment of the generated speech through audio sentiment references, textual sentiment descriptions or sentiment vectors.
• Stronger ability to express emotions: IndexTTS2 has been optimized in terms of emotional expression to better simulate various emotional states.
• Better voice stability: IndexTTS2 enhances the stability of speech generation through techniques such as GPT latent representations and soft instruction mechanisms.

What is IndexTTS2's official website?

• Project website:: https://index-tts.github.io/index-tts2.github.io/
• Github repository:: https://github.com/index-tts/index-tts
• HuggingFace Model Library:: https://huggingface.co/IndexTeam/IndexTTS-2
• arXiv Technical Paper:: https://arxiv.org/pdf/2506.21619

Individuals for whom IndexTTS2 is indicated

• audiobook creator: Generate natural and fluent speech, provide high-quality speech synthesis for audiobook production, and enhance listeners' listening experience.
• Smart Assistant Developers: Provide natural and smooth voice interaction to enhance user experience in scenes such as intelligent assistants and voice broadcasting.
• advertising copywriter: Provides personalized speech synthesis for advertisement production, supporting multiple languages and emotional styles to enhance the appeal of advertisements.
• educator: Provide lively voice explanations in educational software and online courses to help students understand and learn better.
• content creator: Including self-publishers, podcasters, etc., who need high-quality voice content to enrich their works, IndexTTS2 can provide diverse voice styles and emotional expressions.
• Technology Developer: Interested in TTS technology, want to open source model for secondary development or integrated into their own projects, IndexTTS2 provides a strong technical basis and flexible deployment.