Skip to content
Audio

chatty-tts-316m

Expressive speech synthesis with laughter and emotion support on a laptop

2 min read

The model is based on the 24-layer Pocket TTS architecture by Kyutai and fine-tuned for conversational English with the ability to add non-verbal tags to text. Its main feature is the ability to laugh in the right places via a command or change tone using special tokens. The model runs right on CPU hardware (for example, on a MacBook with an M2 chip faster than real-time) and consumes about 1.8 GB of memory. It is useful for bot developers and local voice synthesis enthusiasts who need lively dialogue without heavy GPUs.

Key features

  • Support for tags like to insert laughter at the beginning or end of a sentence.
  • Runs on a standard CPU, delivering speeds slightly above real-time on an Apple M2.
  • 316 million parameters with a memory footprint of about 1.8 GB.
  • Availability of stylistic tokens to set intonation (for example, [amused]).

What you can do

  • Voice lines for a local conversational assistant or chatbot with lively emotions and laughter.
  • Test speech generation locally via Python scripts or a Colab environment.
  • Experiment with sampling step settings and classifier-free guidance (CFG).

Example

uvx hf download mktj/chatty-tts-316m --local-dir chatty-tts && cd chatty-tts
uv sync
uv run generate.py "That's the funniest thing I've heard all week. " --out out.wav
Read moreDetails
Benchmarks: up to 3–4 successful laughter triggers out of 10 on standard settings.
Hardware: Apple M2 processor (16 GB), about 1.8 GB VRAM/RAM.
Limitations: English language only, sentence-by-sentence generation separately, non-commercial license (CC BY-NC-SA 4.0).
Training details: 2,000 steps on the Expresso and NonverbalTTS datasets with the addition of 27 new tokens.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.6 GB · ready files1.1 GB
  • FP8 / INT8recommendedweights 0.3 GB0.8 GB
  • 4-bitweights 0.2 GB0.7 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # mktj/chatty-tts-316m
    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 mktj/chatty-tts-316m --local-dir ~/models/chatty-tts-316m
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to chatty-tts-316m
  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 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face

About the model

mktj/chatty-tts-316m
Author
mktj
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments