whatmodel.sh / tts / Kokoro 82M v1.0

Kokoro 82M v1.0

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Kokoro 82M v1.0 is an open-weights text-to-speech model from hexgrad, released January 2025. Its published size is 0.082B parameters. Licence: apache-2.0.

52.2±2.9#59 of 118 text-to-speech modelsfrom 7 benchmark results across 4 sources

Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.

$ how it compares

The text-to-speech models ranked either side of Kokoro 82M v1.0, out of 148 with a score.

#modelscorewhereRAMreleased
71Lightning v3.1 Smallest.ai53±3api $17.50/M chars2026-03
72Gepard 1.0 nineninesix.ai53±4comparison only2026-06
73Kokoro 82M v1.0 hexgrad52±3api $0.62/M chars +local~0.3GB2025-01
74LEMAS-TTS (w/ prosody, multilingual_prosody) LEMAS Project52±6local~1.5GB2026-01
75Neuphonic neu_hq Neuphonic52±3comparison only2025-08

see all text-to-speech models · find the ones that fit your machine

$ what it's good at

other tests7 not in a task · 97% of its score
testresultvs the others in its rundifficultycountswhy
EmergentTTS · 6-category preference25.9% ✎▅▅▃▃█▅▅▃·▃·▅ beats 8/16hard · avg 30%▇▇▇▇▇▇▇▇▇▇ 26%at its level
carun tts-bench · 56-sentence English corpus · Whisper large-v3 WER2.3% ✎█····█·····█ beats 2/2avg 2.9%▇▇▇▇▇▇▇▇ 20%small run
Paradee small-TTS comparison · 200 WikiText-103 sentences · UTMOS (utmos22_strong)4.52 / 5 ✎█······█···█ beats 2/2easy · avg 4.39 / 5▇▇▇▇▇▇▇ 16%small run
RW-Voice-EQ TTS · Voice Identity (human)3.50 / 5 ✎█·····▄····█ beats 2/4medium · avg 3.08 / 5▇▇▇▇▇▇ 14%at its level
Paradee small-TTS comparison · 200 WikiText-103 sentences · Whisper-base WER5.7% ✎▄··········█ beats 1/2avg 5.6%▇▇▇▇▇ 12%small run
EmergentTTS · 6-category WER13.4% ✎▂▄··▂▄▄··█▄▆ beats 8/16avg 16%▇▇▇ 6.3%blunt
RW-Voice-EQ TTS · Long-Form Stability (human)3.54 / 5 ✎█······█·███ beats 4/4medium · avg 3.42 / 5▇ 2.0%blunt · at its level

$ where to run it

Through an API

api · DeepInfra · $0.62/M chars · deepinfra.com ↗ · price sourceUSD per 1M input characters: DeepInfra lists hexgrad/Kokoro-82M at $0.62 / 1M characters (page pricing data: input_character_length billing). Product: DeepInfra hosted inference. API model: hexgrad/Kokoro-82M. Settings: standard serverless. Price basis: published list rate. Currency: USD. DeepInfra official model page, observed 2026-09-25.

Other ways to run it

version

  1. Run it as published · needs ~0.7 GB, Mac only · from github.com/Blaizzy
    pip install mlx-audio misaki && mlx_audio.tts.generate --model mlx-community/Kokoro-82M-bf16 --text 'Hello, world!' --voice af_heart

Official weights and documentation: huggingface.co/hexgrad ↗

$ how the score was worked out · EmergentTTS-Eval six-category… 32%, Paradee: Distilling Kokoro-82… 28%, carun/tts-bench: intelligibil… 20% and 1 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
EmergentTTS-Eval six-category leaderboard2▇▇▇▇▇▇▇▇▇▇ 32%Apache-2.0
Paradee: Distilling Kokoro-82M into an 8M-Parameter Single-Voice TTS Model (arXiv:2610.06817v1), Table 3 small-model comparison2▇▇▇▇▇▇▇▇▇ 28%CC BY 4.0
carun/tts-bench: intelligibility benchmark of four open TTS engines (published run, 2026-09-10)1▇▇▇▇▇▇ 20%Apache-2.0
RW-Voice-EQ Bench (Hume AI)2▇▇▇▇▇ 16%CC BY 4.0
prior · pull toward the middle▇ 3%
weighted average= 52.2

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json