pocket-tts-int8

An INT8-quantized version of a lightweight autoregressive text-to-speech model.

kyutai/pocket-tts-without-voice-cloning, by Kyutai

Channel-wise INT8 quantization by Idle Intelligence.

CC-BY-4.0

Channel-wise INT8 quantization of Kyutai’s Pocket TTS, shrinking the weights from 225MB to 132MB. An earlier quantization than the GGUF Q8_0 build (pocket- tts-gguf) the live tts-web demo actually loads.

Type
model
Runs
browser, WASM
Size
100M params, 132MB (channel-wise INT8, 41% smaller than the original)
Source
idle-intelligence/tts-web
HuggingFace
idle-intelligence/pocket-tts-int8
Status
maintained

← Home