Light tierRuns on CPU

Kokoro-82M

hexgrad · Kokoro

The most-downloaded text-to-speech model on HuggingFace by a wide margin, at 11.6M downloads a month. At just 82M parameters it is small enough to run on CPU alone — no GPU minimum is published, unlike most other speech models here. Apache 2.0 licensed, so it is fully free for commercial use with no restrictions. Ships with 8 languages and 54 voices out of the box.

Params

82M

License

Apache 2.0

Min VRAM

None published — small enough to run on CPU alone

Languages

8 languages · 54 voices

Tier

Light

Install & run locally

terminal
pip install -q kokoro>=0.9.2 soundfile
apt-get -qq -y install espeak-ng

from kokoro import KPipeline
pipeline = KPipeline(lang_code='a')
generator = pipeline("Your text here", voice='af_heart')

Full install and usage docs on the Hugging Face model card.

Will it run on my hardware?

Kokoro-82M needs about None published — small enough to run on CPU alone of fast memory. Here is what in our hardware database clears that bar.

Memory requirement for this model is variable — see the run command and the model notes above.

Hardware links are affiliate links. We earn a small commission if you buy through them — at no extra cost to you. Disclosure

Specifications

Parameters82M
LicenseApache 2.0
Min VRAMNone published — small enough to run on CPU alone
Languages8 languages · 54 voices
Recommended tierLight
Best fortext-to-speechcpu inferencemultilingual speech

Source

Not distributed through Ollama — install and run it directly from the model card.

View on Hugging Face

Popularity trend (illustrative)

Illustrative adoption trend, not actual download counts. We do not publish a download number for any model because we have no honest source for it.

Similar models

LightBroadest language coverage

OmniVoice

k2-fsa · Apache 2.0 (code) / CC BY-NC (pretrained weights)

A zero-shot voice-cloning model from the k2-fsa (Next-gen Kaldi) team with the broadest language coverage of any open TTS model — 600+ languages from a single checkpoint. The code is Apache 2.0, but the pretrained weights are CC BY-NC because of constraints in the Emilia training data, so the released checkpoint is not free for commercial use without separate authorization.

Min VRAM

Not published — tested on server-class GPUs; also runs on Apple Silicon (MPS) and Intel Arc (XPU)

Speed

zero-shot voice cloningmultilingual speechlow-resource languages

Install & run

pip install torch==2.8.0+cu128 torchaudio==2.8.0+cu128 --extra-index-url https://download.pytorch.org/whl/cu128
Heavy

Fish Audio S2 Pro

Fish Audio · Fish Audio Research License (non-commercial)

Fish Audio's flagship speech model, trained on 10M+ hours of audio with fine-grained inline control over prosody and emotion across 80+ languages. Needs a 24 GB GPU for CUDA inference, though a CPU-only install path exists for testing. Free for research and non-commercial use under the Fish Audio Research License; commercial use requires a separate paid license.

Min VRAM

24 GB GPU (CUDA) — CPU-only install also available

Speed

Real-time factor 0.195 on a single H200 · time to first audio ~100ms

voice cloningemotion controlmultilingual speech

Install & run

git clone https://github.com/fishaudio/fish-speech.git
StandardFully commercial

Qwen3-TTS 1.7B CustomVoice

Alibaba · Apache 2.0

Alibaba's Qwen3-TTS with 9 premium preset voices and instruction-based control over tone, emotion and prosody across 10 major languages. Streaming generation gets end-to-end latency as low as 97ms. Apache 2.0 licensed — fully free for commercial use, unlike most other speech models on this page.

Min VRAM

Not published — FlashAttention recommended to reduce memory use

Speed

End-to-end synthesis latency as low as 97ms (streaming)

custom voicevoice cloningmultilingual speech

Install & run

pip install -U qwen-tts