selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

mehdi-hf logo

cilla-lite

mehdi-hf·released 14 Nov 2026

CC BY-NC 4.0110M≈0.7 GBgated
Runs on CPU or any GPU≈209 MB of weights
hf download mehdi-hf/cilla-lite
Setup ↓

Gated, non-commercial (CC BY-NC 4.0): ask for access on this page. Cilla Lite is the small 6-layer model for CPUs, distilled from Cilla with guidance baked in (one pass per frame). It speaks Persian in the voice of a short audio prompt. Built on Pocket TTS by Kyutai.

How much VRAM does cilla-lite need?

At FP16 / BF16 cilla-lite needs about 0.7 GB of GPU memory.

--vram
any
--gpus
--quant
QuantisationWeightsTotal
4-bitRecommended default for local use0.1 GB ≈0.6 GB
8-bitLossless in practice, twice the size of Q40.1 GB ≈0.6 GB
FP16 / BF16recommendedFull precision, as released0.2 GB ≈0.7 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.

Can the RTX 4060 run cilla-lite? →

Run cilla-lite locally

No runtime support recorded yet. Most models also run through Hugging Face Transformers.

About cilla-lite

Gated, non-commercial (CC BY-NC 4.0): ask for access on this page. Cilla Lite is the small 6-layer model for CPUs, distilled from Cilla with guidance baked in (one pass per frame). It speaks Persian in the voice of a short audio prompt. Built on Pocket TTS by Kyutai. Speed: 1.88x real time on one thread of a cloud x86 CPU (Intel(R) Xeon(R) CPU @ 2.80GHz), measured while making the samples below. 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 config needs the fork. config.yaml sets capitalize_first_letter, append_terminal_punctuation and pad_with_spaces_for_short_inputs to false, which the released pocket-tts does not know and rejects. The loud error is deliberate: without them the package upper-cases or drops the first letters of Persian text. Write ordinary Persian with punctuation: commas and full stops shape the phrasing. The model reads the ezafe from the context (there are no ezafe marks to type); it still gets some wrong. The voice is cloned from a prompt: a whole short clip (3 to 10 s) that ends at a pause, in a quiet room. Prompts cut mid-word make the model sound broken. prompts/ has two neutral ones, voice1.wav (a woman) and voice2.wav (a man).

Full model card on Hugging Face →

Other Cilla models

Whole family →
mehdi-hf logo
cilla-minimehdi-hf
28MNon-commercial≈0.6 GB↓ 0

Alternatives to cilla-lite

All alternatives →
OpenMOSS-Team logo

MOSS-TTS-Nano-100M

OpenMOSS-Team·TTS

≈0.7 GB

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…

100M32K ctx Apache 2.0
122.3KApr 2026gguf
hexgrad logo

Kokoro-82M

hexgrad·TTS

≈0.7 GB

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.

82MApache 2.0
13.1MDec 2024gguf
zaakirio logo

kokoro-ru

zaakirio·TTS

≈0.7 GB

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.

82MOpenRAIL
72.7KJul 2026
Oomni logo

mithi-v1-flash

Oomni·TTS

≈0.7 GB

A small text-to-speech model for 11 Indian languages, with two voices: Satvik (male) and Charvi (female).

99MApache 2.0
1.4KNov 2026
KittenML logo

kitten-tts-2-mini

KittenML·TTS

≈0.7 GB

82M parameters, 16 languages, 118 voices, 24 kHz. ONNX with int8 weights (82.7 MB; runs on CPU, no PyTorch needed), and the float32 PyTorch weights.

82MApache 2.0
1.3KNov 2026
hexgrad logo

Kokoro-82M-v1.1-zh

hexgrad·TTS

≈0.7 GB

Kokoro is an open-weight series of small but powerful TTS models.

82MApache 2.0
82.9KFeb 2025

cilla-lite FAQ

How much VRAM does cilla-lite need?
About 0.7 GB at FP16 / BF16.
Can cilla-lite run on the RTX 4060?
Yes. It fits at FP16 / BF16 (about 0.7 GB). It fits comfortably on the Apple M4 (16 GB), Radeon RX 6700 XT, RTX 3060 12GB.
Can I use cilla-lite commercially?
No. It is released under CC BY-NC 4.0. Non-commercial use only, with attribution.

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.