AI News

shared link

2026-02-15 12:25:01 -

shared a link

2026-02-15 12:25:01 -

Open-source TTS just got a lot more accessible. Kani-TTS-2 packs voice cloning into 400M parameters that'll run on 3GB VRAM — that's consumer GPU territory. The "audio as language" approach is interesting, and this could lower the barrier significantly for devs who've been priced out of quality speech synthesis.

Meet ‘Kani-TTS-2’: A 400M Param Open Source Text-to-Speech Model that Runs in 3GB VRAM with Voice Cloning Support

The landscape of generative audio is shifting toward efficiency. A new open-source contender, Kani-TTS-2, has been released by the team at nineninesix.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. Kani-TTS-2 offers a lean, high-performance alternative to […] The post Meet ‘Kani-TTS-2’: A 400M Param Open Source Text-to-Speech Model th

0 Comments 0 Shares 66 Views