Skip to content
Audio

Pocket-TTS-LiteRT-FP16

On-Device Text-to-Speech and Voice Cloning

2 min readupdated

This repository contains 22 Pocket TTS model packages converted to the LiteRT format (a lightweight runtime for running models on mobile and edge devices) with FP16 precision (16-bit floating point). Each build generates 24 kHz mono audio using a short reference WAV file for voice tuning. The models cover 26 languages, including official variants from Kyutai, community packs, and a general Indian multilingual set. This release is useful for mobile app and embedded system developers who need local text-to-speech with voice cloning capabilities.

Key features

  • Support for 26 languages, including English, German, French, Japanese, and others.
  • Model weights saved in FP16 format with FP32 inputs and outputs for stable CPU execution.
  • Each package includes four .tflite graphs for audio encoding, generation, and decoding.
  • A detailed manifest (pocket-tts-index.json) with text preprocessing rules and checksums.

What you can do

  • Integrate local text-to-speech with voice cloning into mobile applications running on LiteRT.
  • Create a speech engine to generate audio files from a reference sample without sending data to a server.
  • Run the text and audio processing pipeline using the included Colab notebook.

Example

To run the pipeline and generate speech from a reference audio file, use the dedicated Google Colab notebook from the repository. The host code handles text tokenization, generation cache management, and final audio assembly using the package files.

Read moreDetails
Output audio signal: mono, 24 kHz.
Weights format: FP16 (graphs), FP32 (inputs/outputs).
Architecture: packages include a tokenizer, weights, and four .tflite files for reference encoding, generation, and two stages of Mimi decoding.
Notes: Korean and Welsh packages use larger 24-layer models.
Discuss Telegram 0 345

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Serkali-sudo/Pocket-TTS-LiteRT-FP16
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Serkali-sudo/Pocket-TTS-LiteRT-FP16 --local-dir ~/models/Pocket-TTS-LiteRT-FP16
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Pocket-TTS-LiteRT-FP16
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Pocket-TTS-LiteRT-FP16 (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

Serkali-sudo/Pocket-TTS-LiteRT-FP16
Author
Serkali-sudo
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

3450

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments