F5-TTS Server

This project provides a web server and a user-friendly interface for the F5-TTS model, allowing you to easily perform text-to-speech synthesis with voice cloning capabilities.

How to Use

1. Clone the repository:

   git clone https://github.com/jeantimex/F5-TTS-Server.git
   cd F5-TTS-Server

2. Create and activate a virtual environment:

   python -m venv .venv
   source .venv/bin/activate

3. Run the start script:

   This will install the required dependencies and start the server.

   ./start.sh

   The server will be running at http://127.0.0.1:8000.

Features

• Web Interface: A simple and intuitive web UI to generate speech without using the command line.
• Voice Cloning: Use your own reference audio files to clone voices for TTS.
• Customization: Adjust speech speed, NFE steps, cross-fade duration, and more.
• File Management: Upload and delete custom reference audio files directly through the UI.
• REST API: A comprehensive API for programmatic access to all TTS functions.
• Real-time Cancellation: Cancel TTS jobs that are in progress.

API Endpoints

Here are the available API endpoints with curl examples.

Get Frontend

Serves the main HTML page.

curl -X GET http://127.0.0.1:8000/

List Reference Audios

Get a list of all available reference audio files.

curl -X GET http://127.0.0.1:8000/ref-audios/

Upload Reference Audio

Upload a new reference audio file.

curl -X POST http://127.0.0.1:8000/upload-ref-audio/ \
     -F "file=@/path/to/your/audio.wav"

Upload Reference Text File

Upload a .txt file corresponding to a reference audio. The filename of the text file should match the audio file's name (e.g., my_audio.wav and my_audio.txt).

curl -X POST http://127.0.0.1:8000/upload-text-file/ \
     -F "file=@/path/to/your/text.txt"

Delete Reference Audio

Delete a custom reference audio file. Note: You can only delete files from the custom folder.

# Replace 'custom/your_audio.wav' with the actual file path
curl -X DELETE http://127.0.0.1:8000/delete-ref-audio/custom/your_audio.wav

Serve Reference Audio

Get a specific reference audio file.

# Replace 'default/basic_ref_en.wav' with the actual file path
curl -X GET http://127.0.0.1:8000/ref-audios/default/basic_ref_en.wav -o ref_audio.wav

Text-to-Speech Generation

Generate speech from text. This is the main endpoint.

Basic Example

This example uses the default settings, including the default English reference voice.

curl -X POST http://127.0.0.1:8000/tts/ \
     -H "Content-Type: application/json" \
     -d '{
           "gen_text": "Hello, world! This is a test of the F5-TTS server."
         }' \
     --output basic_speech.wav

Multi-language Example

You can generate speech in different languages by specifying the appropriate reference audio.

English Example

curl -X POST http://127.0.0.1:8000/tts/ \
     -H "Content-Type: application/json" \
     -d '{
           "gen_text": "This is a test in English.",
           "ref_audio": "default/basic_ref_en.wav"
         }' \
     --output english_speech.wav

Chinese Example

curl -X POST http://127.0.0.1:8000/tts/ \
     -H "Content-Type: application/json" \
     -d '{
           "gen_text": "这是一个中文测试。",
           "ref_audio": "default/basic_ref_zh.wav"
         }' \
     --output chinese_speech.wav

Customization Example

This example adjusts the speech speed and quality (NFE steps).

curl -X POST http://127.0.0.1:8000/tts/ \
     -H "Content-Type: application/json" \
     -d '{
           "gen_text": "I can speak faster or slower depending on the settings.",
           "speed": 1.2,
           "nfe_steps": 40
         }' \
     --output custom_speech.wav

Voice Cloning Example

To clone a voice, you first need to upload a reference audio file, then use its path in your TTS request.

Step 1: Upload a reference audio

curl -X POST http://127.0.0.1:8000/upload-ref-audio/ \
     -F "file=@/path/to/your/voice_sample.wav"

The server will respond with a JSON object containing the filename, for example: {"filename": "custom/voice_sample.wav"}.

Step 2: Use the uploaded audio for TTS

Important Note: For the best voice cloning results, it is highly recommended to provide the transcript of your reference audio. You can do this by uploading a .txt file with the same name as your audio file (e.g., voice_sample.wav and voice_sample.txt). If you don't provide a text file, the server will attempt to automatically transcribe the audio, but the quality may vary.

Use the filename from the previous step as the ref_audio in your request.

curl -X POST http://127.0.0.1:8000/tts/ \
     -H "Content-Type: application/json" \
     -d '{
           "gen_text": "This text will be spoken in the voice from the uploaded audio file.",
           "ref_audio": "custom/voice_sample.wav"
         }' \
     --output cloned_speech.wav

Cancel TTS Generation

Cancel a running TTS generation process. You need the request_id which is returned in the headers of the /tts response.

# Replace 'your_request_id' with the actual request ID
curl -X POST http://127.0.0.1:8000/cancel-tts/your_request_id

Get TTS Status

Check the status of a TTS generation request.

# Replace 'your_request_id' with the actual request ID
curl -X GET http://127.0.0.1:8000/tts-status/your_request_id