Reference
D

Kokoro 82M

Kokoro 82M, a text-to-speech model from DeepInfra.
speechModelOverride: { model: "kokoro-82m-deepinfra" }
Hexgrad·served viaDeepInfra·Released January 2025
Pricing
Per 1M characters
$0.62
Per 1M characters
$0
Specifications
Context window
10,000 tokens
Released
January 2025
Model string
hexgrad/Kokoro-82M
Capabilities
Open sourceLow costMultilingual
Parameters
Model options, passed in speechModelOverride.config.
ParameterTypeDefaultRange / options
voice
Voice to use in TTS. The prefix encodes language and gender.
Selectaf_bellaaf_alloyaf_aoedeaf_bellaaf_heartaf_jessicaaf_kore+48
Voices · 54
American Female — Alloy
American Female — Aoede
American Female — Bella
American Female — Heart
American Female — Jessica
American Female — Kore
American Female — Nicole
American Female — Nova
American Female — River
American Female — Sarah
American Female — Sky
American Male — Adam
American Male — Echo
American Male — Eric
American Male — Fenrir
American Male — Liam
American Male — Michael
American Male — Onyx
American Male — Puck
American Male — Santa
British Female — Alice
British Female — Emma
British Female — Isabella
British Female — Lily
British Male — Daniel
British Male — Fable
British Male — George
British Male — Lewis
Spanish Female — Dora
Spanish Male — Alex
Spanish Male — Santa
French Female — Siwis
Hindi Female — Alpha
Hindi Female — Beta
Hindi Male — Omega
Hindi Male — Psi
Italian Female — Sara
Italian Male — Nicola
Japanese Female — Alpha
Japanese Female — Gongitsune
Japanese Female — Nezumi
Japanese Female — Tebukuro
Japanese Male — Kumo
Portuguese Female — Dora
Portuguese Male — Alex
Portuguese Male — Santa
Chinese Female — Xiaobei
Chinese Female — Xiaoni
Chinese Female — Xiaoxiao
Chinese Female — Xiaoyi
Chinese Male — Yunjian
Chinese Male — Yunxi
Chinese Male — Yunxia
Chinese Male — Yunyang
Use this model
await mindstudio.textToSpeech({
  text: "...",
  speechModelOverride: {
    model: "kokoro-82m-deepinfra",
    config: {
      voice: "af_alloy",
    },
  },
});