Kokoro
Hexgrad
Kokoro is a breakthrough open-source TTS model (Apache 2.0) from Hexgrad that proves small models can rival commercial APIs. At just 82 million parameters, it runs at 210x real-time speed on a consumer GPU (even on free Google Colab), producing natural-sounding speech that passes casual listening tests against paid alternatives. Apache 2.0 license means zero restrictions for commercial use. English-focused with strong naturalness and pacing. Best for developers and researchers who want production-quality TTS at zero per-character cost, are comfortable with self-hosting, and don't need multilingual or voice cloning features.
Completely free (Apache 2.0 open source). No per-character costs — self-host on any GPU. Available on Hugging Face and GitHub.
Related platforms
Coqui XTTS
Coqui AI (Community)
Open-source multilingual voice cloning model — 17 languages, self-hosted, 6-second voice cloning.
Chatterbox
Resemble AI
MIT-licensed open-source TTS that beat ElevenLabs in blind tests — free self-hosted voice cloning.
Resemble AI
Resemble AI
Enterprise voice cloning with deepfake detection, watermarking, and Hollywood-grade synthesis.
Appsmith
Appsmith
Open-source Retool alternative — self-hosted internal tools with 45+ connectors and zero license cost.