Text-to-speech
Kokoro 82M requirements, license & speed
By hexgrad · released 2024-12-26 · 82M parameters
Commercial useYes
Min memoryNot stated
Runs on CPUYes
StreamingUnknown
Key facts
- License
- apache-2.0 source
- Commercial use
- Yes sourceApache-2.0; card welcomes production/commercial deployment
- Parameters
- 82M source
- Minimum memory
- Not stated by source
- Runs on CPU
- Yes sourceONNX build (kokoro-onnx) runs on CPU; measured here on an 8 vCPU x86 VM
- Streaming
- Unknown
- Voice cloning
- Unknown
- Languages
- en-US, en-GB, es, fr, hi, it, ja, pt-BR, zh sourceHF metadata lists only en; lang codes from the kokoro package README
- Released
- 2024-12-26 sourceHugging Face repo creation date (proxy for release date)
- Gated on HF
- No
Measured speed & memory
| Hardware | Runtime / settings | RTF | Time to first audio | Peak RAM | Date |
|---|---|---|---|---|---|
| Cloud x86 VM, 8 vCPU Intel Xeon, 16 GB RAM, no GPU | kokoro-onnx fp32 (onnxruntime CPU) voice=af_heart | 0.1234 (8× real time) | 1.416 s | 933.9 MB | 2026-10-21 |
| Cloud x86 VM, 8 vCPU Intel Xeon, 16 GB RAM, no GPU | kokoro-onnx int8 (onnxruntime CPU) voice=af_heart | 0.1480 (7× real time) | 1.686 s | 744.0 MB | 2026-10-21 |
Median of 5 runs after one warm-up, ~10 s English passage. Hardware is labelled exactly: a cloud x86 VM is not an N100 or a Pi. Full method on the CPU only page. Peak RAM is whole-process RSS. Methodology
Accuracy
We don't run accuracy evals. For quality and intelligibility, check the Hugging Face Open TTS Leaderboard.
Comparisons: Kokoro vs Piper · Kokoro vs Chatterbox · Kokoro vs Qwen3-TTS
Similar TTS models
Spotted an error? Report it. Every fact links to the model card or repo it came from.