Pocket TTS: Clone Any Voice in 5s on CPU
gvij · reddit · 2026-07-06
Kyutai's Pocket TTS is a streaming language model with roughly 100 million parameters. It uses the Mimi neural codec to generate audio tokens, enabling zero-shot voice cloning from just 5 seconds of reference audio without needing a GPU. It is MIT-licensed and easily installed via pip. The author compared it against three CPU TTS models: Kokoro, Supertonic, and Inflect-Nano. Across 180 timed tests, its latency remained nearly constant regardless of text length (RTF 0.69-0.76) and supported streaming output. Although it was the slowest of the four, it offered the most unique features.
Related event: Kyutai Open-Sources Pocket-TTS for CPU-Based Voice Cloning(4 posts)→
More from Multimodal
- Short film 'Odissey of Aeons' made with MiniMax H3 video generation — Musikalandia · 2026-10-27
- Remiqora Video: open-source local studio turns one idea into 4K chained AI videos — inikolax · 2026-10-27
- $5k kale-img-1 fine-tuning contest won with MOSS-Image base, SDXL trails far behind — ostrisai · 2026-10-27
- HN discussion: How far image generation has come since August 2022 — hunterbrooks · 2026-10-27
- FLUX.3 Edit nails a -90° car rotation on first try with just two reference images — EvilBot-666 · 2026-10-27
- Fan-Made AI Series 'Wolverine: Trail of Tears' Episode One Released — ArtificialLightPixel · 2026-10-27