Breeze TTS 2 - A New-Generation Speech Synthesis Model from BreezeBlue

What is Breeze TTS 2

Breeze TTS 2 is a new-generationspeech synthesismodel from BreezeBlue. It supports zero-shot character voice design through natural language and precise control of performance details such as emotion, speaking rate, and accent with natural language instructions. The model-side latency is as low as 40ms, and it ranks first globally in all three authoritative evaluations: Voice Design, Voice Direction, and Latency. It supports 50+ languages.

Main Features of Breeze TTS 2

• Voice Design: Create unique character voices from scratch by describing characteristics such as age, temperament, and personality in natural language, without relying on a preset voice library.
• Voice Direction: Precisely control emotion, speaking rate, accent, physiological state, and dynamic performance changes with natural language instructions while maintaining the character's voice identity consistency.
• Ultra-Low Latency Real-Time Generation: The model-side first-byte latency is as low as 40ms and supports WebSocket streaming, meeting the needs of real-time conversation and interaction.
• Multilingual Coverage: Supports more than 50 languages, maintaining voice naturalness, character consistency, and performance control in cross-lingual scenarios.

Technical Principles of Breeze TTS 2

• Large Language Model Paradigm Shift: Breeze TTS 2 follows the LLM technical approach, transferring the language model's generalization ability, instruction-following ability, and Scaling Law to the field of voice generation, enabling the speech model to understand complex character settings and context.
• Long Dialogue-Oriented Data Strategy: The team collected real long dialogue data from films and television, short dramas, interviews, and podcasts, with emphasis on annotating character relationships, emotional tension, turn-taking rhythm, and pause intent, so that the model learns how humans express themselves in complex interactions rather than standardized reading of text.
• Scenario-Driven Evaluation System Reconstruction: After discovering that academic Benchmarks were disconnected from creators' actual needs, the team rebuilt evaluation standards around the two core stages of voice design and voice direction, and open-sourced the corresponding Benchmark, using cross-scenario character consistency and continuous change capability as the basis for model iteration.
• End-to-End Unified Architecture: Integrates voice design, voice direction, and low-latency streaming generation into a single model, achieving character creation, performance control, and real-time response through a unified inference pipeline, avoiding the latency and consistency loss caused by stitching multiple modules together.

How to Use Breeze TTS 2

• BreezeCreator Web Platform: Visit the BreezeCreator platform official website https://breezeblue.ai/app，用自然语言描述角色特征设计专属音色，再通过自然语言指令调整情绪与表演，直接生成音频内容。
• HTTP API Calls: After developers obtain an API Key, they submit text and voice/performance parameters through standard HTTP interfaces to obtain synthesized speech files, suitable for batch content production.
• WebSocket Real-Time API: Establish a persistent WebSocket connection to append text in real time and receive raw PCM audio in a streaming manner, suitable for low-latency interaction scenarios such as virtual streamers, game NPCs, and AI Agents.
• SDK Integration: Use the official JavaScript/TypeScript SDK for quick integration. With just a few lines of code, you can create real-time voice sessions, manage conversation turns, and play audio.

Core Advantages of Breeze TTS 2

• Zero-Shot Voice Design: Create unique character voices from scratch by describing characteristics such as age, temperament, and personality in natural language, without relying on a preset voice library.
• Natural Language Voice Direction: Use natural language instructions to precisely control emotion, speaking rate, accent, physiological state, and dynamic performance changes while maintaining a consistent character voice identity.
• Ultra-Low-Latency Real-Time Generation: Model-side first-byte latency is as low as 40ms, ranking first globally in the TTS Latency Benchmark, meeting real-time interaction needs.
• Multilingual Character Consistency: Supports 50+ languages, maintaining voice naturalness, character consistency, and performance control across languages.
• Comprehensive Benchmark Leadership: Ranked first globally in all three authoritative benchmarks: Voice Design, Voice Direction, and Latency.

Breeze TTS 2 Project Address

• Project Website：https://breezeblue.ai/breeze-tts-2

Comparison of Breeze TTS 2 with Similar Competitors

Application Scenarios of Breeze TTS 2

• Virtual Streamers and AI Live Streaming: AI streamers adjust their tone in real time based on danmaku and complete voice performances; a single livestream can last 1-2 hours, with hundreds of sessions completed cumulatively.
• Game NPCs and Role-Playing: Generates unique voices for thousands of in-game NPCs, supporting real-time dialogue and emotional changes to enhance immersion.
• Audio Content and Radio Dramas: Creators use natural language to design multi-character voices and produce long-form radio dramas (such as The Three-Body Problem fan creations), which hit the trending charts on the first day of release.
• AI Agents and Intelligent Customer Service: Low-latency real-time voice interaction enables AI Agents to proactively report progress, accept interruptions, understand user intent, and continue collaborating.
• Interactive Entertainment and Virtual Companionship: Gives virtual characters and AI companions unique voice identities, supporting long-term, character-based emotional interaction and storytelling experiences.