--- license: cc-by-4.0 base_model: kyutai/pocket-tts tags: [testing] --- # tinyPocketTTS-for-testing A slice of the english pack of [kyutai/pocket-tts](https://huggingface.co/kyutai/pocket-tts) for testing purposes only. The first 2 of 6 backbone layers, the first 1 of 6 flow net blocks and the first 1 of 2 layers of each mimi transformer are kept, and only the leading rows and columns of each hidden axis: 1/4 of the hidden sizes and of the seanet channels, 1/8 of the feed-forward sizes. The waveform channel, the 32-d latent, the kernels and the tokenizer are unchanged, and so is the voice cloning encoder (sliced the same way). There are no predefined voices. Built with [slice_pockettts.py](slice_pockettts.py). The output is not meaningful. Do not use this model for anything other than testing a loader or a runtime.