This repository contains the model as described in LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM.
cilla-mini
mehdi-hf·released 16 Nov 2026
hf download mehdi-hf/cilla-miniGated, non-commercial (CC BY-NC 4.0): ask for access on this page. Cilla Mini is the 18M model for CPUs, phones and browsers: Cilla shrunk with Kyutai's small-model recipe (8 layers, a 64-dim interface, a one-step drifting sampler head, the 2.6M Mimi decoder), guidance baked in.
How much VRAM does cilla-mini need?
At FP16 / BF16 cilla-mini needs about 0.6 GB of GPU memory.
| Quantisation | Total |
|---|---|
| 4-bitRecommended default for local use | ≈0.5 GB |
| 8-bitLossless in practice, twice the size of Q4 | ≈0.5 GB |
| FP16 / BF16recommendedFull precision, as released | ≈0.6 GB |
Estimates for the weights plus runtime overhead; activations vary with resolution and batch size. Pick a GPU above and the fit column checks every row against it.
Run cilla-mini locally
No runtime support recorded yet. Most models also run through Hugging Face Transformers.
About cilla-mini
Gated, non-commercial (CC BY-NC 4.0): ask for access on this page. Cilla Mini is the 18M model for CPUs, phones and browsers: Cilla shrunk with Kyutai's small-model recipe (8 layers, a 64-dim interface, a one-step drifting sampler head, the 2.6M Mimi decoder), guidance baked in. It speaks Persian in the voice of a short audio prompt. Built on Pocket TTS by Kyutai. Try it in your browser · Run it on a Mac · Phonon runtime · Browser library · Python package Audio made per second of work ("real time" is 1x): speak_fa.py normalises the text, speaks one sentence at a time from the same prompt, trims the silence each piece opens with and joins them with a short pause. Do the same in your own code: cutting a sentence into 50-token pieces (the old Pocket TTS default) made these models repeat words, once the name «ساندر پیچای» twice in a row. The released package is enough (pocket-tts 4.0.1 or newer, checked for this repo). config.yaml sets capitalize_first_letter, append_terminal_punctuation and pad_with_spaces_for_short_inputs to false: leave them, otherwise the package upper-cases or drops the first letters of Persian text. The phonon/ folder has the model for the Phonon runtime (Rust, no PyTorch, no GPU) and the same Persian front-end. Apple silicon: Details, options and notes: phonon/README.md.
Full model card on Hugging Face →Other Cilla models
Whole family →Alternatives to cilla-mini
All alternatives →Kokoro is an open-weight TTS model with 82 million parameters. Despite its lightweight architecture, it delivers comparable quality to larger models while being significantly faster and more cost-efficient.
Russian text-to-speech at 82M parameters, running 9.8x faster than realtime on a laptop CPU with no GPU. Three fixed voices from consented studio actors.
This repository contains the English (eng) language text-to-speech (TTS) model checkpoint.
This repository contains the Hindi (hin) language text-to-speech (TTS) model checkpoint.
MOSS-TTS-Nano is an open-source multilingual tiny speech generation model from MOSI.AI and the OpenMOSS team. With only 0.1B parameters, it is designed for realtime speech generation, can run directly on CPU without a GPU, and keeps the deployment stack simple enough for local demos, web serving…
cilla-mini FAQ
- How much VRAM does cilla-mini need?
- About 0.6 GB at FP16 / BF16.
- Can cilla-mini run on the RTX 4060?
- Yes. It fits at FP16 / BF16 (about 0.6 GB). It fits comfortably on the Apple M4 (16 GB), Radeon RX 6700 XT, RTX 3060 12GB.
- Can I use cilla-mini commercially?
- No. It is released under CC BY-NC 4.0. Non-commercial use only, with attribution.
Rate this model
Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.
0 comments