SWivid/F5-TTS

Health v2 · maintenance only

Adoption not part of the score

15167 stars · 2203 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

F5-TTS is the official implementation of a fully non-autoregressive text-to-speech system based on flow matching with a Diffusion Transformer, trained on 100K hours of multilingual speech. It provides zero-shot voice cloning, code-switching, and fast inference (RTF 0.15) via a Python package, CLI, and Gradio app.

Use cases

• clone a voice from a short reference audio clip
• generate natural speech from text in English and Chinese
• run fast text-to-speech inference on a GPU
• fine-tune a TTS model on custom speech data
• build a voiceover or audiobook generation pipeline
• experiment with flow matching based speech synthesis

When to choose

• you need high-quality zero-shot voice cloning without phoneme alignment or duration models
• you want fast diffusion-based TTS with an MIT license
• you need multilingual or code-switching speech generation
• you want a pip-installable TTS with CLI and web UI

When to avoid

• you need a lightweight CPU-only TTS for embedded devices
• you require real-time streaming synthesis with very low latency
• you need a mature commercial-grade TTS with extensive language coverage
• you want a simple rule-based or parametric TTS without GPU requirements

Facets

library · maturity active

tts machine-learning deep-learning audio-processing speech-processing machine-learning artificial-intelligence python windows cli text-to-speech flow-matching diffusion-transformer zero-shot-voice-cloning non-autoregressive pytorch audio linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="SWivid/F5-TTS")