{"id":71315,"date":"2026-02-15T11:48:17","date_gmt":"2026-02-15T11:48:17","guid":{"rendered":"https:\/\/youzum.net\/meet-kani-tts-2-a-400m-param-open-source-text-to-speech-model-that-runs-in-3gb-vram-with-voice-cloning-support\/"},"modified":"2026-02-15T11:48:17","modified_gmt":"2026-02-15T11:48:17","slug":"meet-kani-tts-2-a-400m-param-open-source-text-to-speech-model-that-runs-in-3gb-vram-with-voice-cloning-support","status":"publish","type":"post","link":"https:\/\/youzum.net\/th\/meet-kani-tts-2-a-400m-param-open-source-text-to-speech-model-that-runs-in-3gb-vram-with-voice-cloning-support\/","title":{"rendered":"Meet \u2018Kani-TTS-2\u2019: A 400M Param Open Source Text-to-Speech Model that Runs in 3GB VRAM with Voice Cloning Support"},"content":{"rendered":"

The landscape of generative audio is shifting toward efficiency. A new open-source contender, Kani-TTS-2<\/strong>, has been released by the team at nineninesix<\/strong>.ai. This model marks a departure from heavy, compute-expensive TTS systems. Instead, it treats audio as a language, delivering high-fidelity speech synthesis with a remarkably small footprint.<\/p>\n

Kani-TTS-2 offers a lean, high-performance alternative to closed-source APIs. It is currently available on Hugging Face in both English (EN<\/strong>)<\/a> and Portuguese (PT<\/strong>)<\/a> versions.<\/p>\n

The Architecture: LFM2 and NanoCodec<\/strong><\/h3>\n

Kani-TTS-2 follows the \u2018Audio-as-Language<\/strong>\u2018 philosophy. The model does not use traditional mel-spectrogram pipelines. Instead, it converts raw audio into discrete tokens using a neural codec.<\/p>\n

The system relies on a two-stage process:<\/strong><\/p>\n

    \n
  1. The Language Backbone:<\/strong> The model is built on LiquidAI\u2019s LFM2 (350M)<\/strong> architecture. This backbone generates \u2018audio intent\u2019 by predicting the next audio tokens. Because LFM (Liquid Foundation Models) are designed for efficiency, they provide a faster alternative to standard transformers.<\/li>\n
  2. The Neural Codec:<\/strong> It uses the NVIDIA NanoCodec<\/strong> to turn those tokens into 22kHz waveforms.<\/li>\n<\/ol>\n

    By using this architecture, the model captures human-like prosody\u2014the rhythm and intonation of speech\u2014without the \u2018robotic\u2019 artifacts found in older TTS systems.<\/p>\n

    Efficiency: 10,000 Hours in 6 Hours<\/strong><\/h3>\n

    The training metrics for Kani-TTS-2 are a masterclass in optimization. The English model was trained on 10,000 hours<\/strong> of high-quality speech data.<\/p>\n

    While that scale is impressive, the speed of training is the real story. The research team trained the model in only 6 hours<\/strong> using a cluster of 8 NVIDIA H100 GPUs<\/strong>. This proves that massive datasets no longer require weeks of compute time when paired with efficient architectures like LFM2.<\/p>\n

    Zero-Shot Voice Cloning and Performance<\/strong><\/h3>\n

    The standout feature for developers is zero-shot voice cloning<\/strong>. Unlike traditional models that require fine-tuning for new voices, Kani-TTS-2 uses speaker embeddings<\/strong>.<\/p>\n