๐Ÿ”Š KokoroTTS-HF

Full 70-voice text-to-speech catalogue running on Space CPU. Generate your audio above, audition every sample file below.

70 voices11 languagesCPU inference24 kHz output

๐Ÿ”Š Tแบกo giแปng ฤ‘แปc

Synthesise any text on Space CPU. Type text, pick a voice, press Generate.

109/300
Mode
Language filter
Voice A
0.25 4

Blend mixes two voices of the same model family into a new synthetic speaker โ€” a KokoroTTS-HF exclusive. Pick Blend mode, choose voice B, and set how much of B to mix in.
  • Short sentences synthesise fastest.
  • Keep speed between 0.9 and 1.1 for natural pacing.
  • Vietnamese, German, Japanese and Chinese voices load their model pack on first use โ€” the first request takes longer, later ones are fast.
Try an example
Text to speak Mode Voice A Voice B (must share A's model family) Speed Mix: weight of voice B (0 = pure A, 1 = pure B)

๐ŸŽง Nghe thแปญ giแปng mแบซu

Pick a voice from the list โ€” its bundled sample plays below instantly, no compute used. Then use it in the generator above.
Language filter
Voices (click to audition)

Now auditioning: af_heart
Based on Hangry Labs KokoroTTS (Apache-2.0) and hexgrad Kokoro. German voices use kikiri-tts checkpoints; Vietnamese uses the ContextBoxAI checkpoint.