VoxCPMSpeech generation foundation model with emotional expression control and zero-shot voice cloning from audio clips.
Apache-2.0 License · Free weights and inference code
Voice & transcription
Kokoro
Free optionOpen sourceLocal
Generate spoken audio from text using a local speech model.
What you can use it for
Generate spoken audio from text using a local speech model.
- Covers the core everyday task of ElevenLabs, HeyGen and Synthesia: generating natural speech from text, locally and without an API bill.
- Creators report using it extensively (accessibility products, ebook reading, Home Assistant, side projects) and finding quality close to paid ElevenLabs voices for everyday use.
Workflow coverage and comparison notes
Generate text-to-speech locally without a paid subscription or metered API. The core voice-over slice only; no hosted voice cloning, avatars or managed studio features, and no video generation.
Before you switch
Tiny 82M model: narrower language coverage and weaker on very short or highly emotive utterances. No hosted voice cloning, avatars or studio features; you run it yourself and, for video, pair it with a separate tool.
Alternatives to these paid apps
Each comparison covers the workflow described above, rather than every paid feature.
More for the same workflow
See alternatives →VoxCPM
PiperConvert text into speech locally using downloadable voice models.
Free and open source (GPL-3.0 licence per GitHub).
VoiceStudioTurn scripts into spoken audio and narrate personal audiobook projects on your own computer.
Free software · AGPL-3.0 · Default OmniVoice model: noncommercial only · Your hardware and storage