https://github.com/rudrankriyam/glosik

Sample project for F5-TTS using MLX Swift
https://github.com/rudrankriyam/glosik

mlx mlx-swift swift text-to-speech tts

Last synced: 12 days ago
JSON representation

Sample project for F5-TTS using MLX Swift

• Host: GitHub
• URL: https://github.com/rudrankriyam/glosik
• Owner: rudrankriyam
• License: mit
• Created: 2024-11-28T19:56:14.000Z (2 months ago)
• Default Branch: main
• Last Pushed: 2025-01-03T14:46:02.000Z (about 1 month ago)
• Last Synced: 2025-01-23T01:46:01.946Z (17 days ago)
• Topics: mlx, mlx-swift, swift, text-to-speech, tts
• Language: Swift
• Homepage:
• Size: 7.35 MB
• Stars: 30
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Głosik

Głosik (pronounced "gwoh-seek") is an example app to showcase the F5-TTS text-to-speech system using MLX Swift. The name comes from the Polish word "głos" (voice) with the diminutive suffix "-ik".

Here is the original repository of the implementation: [https://github.com/lucasnewman/f5-tts-swift](https://github.com/lucasnewman/f5-tts-swift)

https://github.com/user-attachments/assets/ccebcee9-13e8-400b-a189-6df926c6223c

Watch the demo above to see Głosik in action!

## Requirements

- macOS 14.0 or later

- iOS 16.0 or later

- visionOS 1.0 or later

- Xcode 15.0 or later

- Swift 5.9 or later

## Installation

1. Clone the repository

2. Open `Glosik.xcodeproj` in Xcode

3. Build and run the project

## Usage

1. Enter the text you want to convert to speech

2. (Optional) Record or select a reference audio sample:

   - Go to the "Reference" tab

   - Record a new audio sample and provide reference text

   - Save it as a reference sample

   - Select it from the reference picker in the "Generate" tab

3. Click "Generate Speech" to create the audio

4. Use the playback controls to listen to the generated speech

5. Save the generated audio as a WAV file

## Features

### Text-to-Speech Generation

- High-quality speech synthesis using F5-TTS model

- Real-time generation progress tracking

- Generation timing statistics

- GPU memory usage monitoring

### Reference Audio Support

- Record new reference samples with accompanying text

- Manage saved reference samples

- Select reference samples for speech generation

- Play back reference samples

- Support for mono, 24kHz WAV format

### Modern UI

- Native SwiftUI interface

- Split-view navigation

- Dark mode support

- Cross-platform support (macOS, iOS, visionOS)

- Accessibility features

## Project Structure

The project is split into two main parts:

- `Glosik`: Main application

- `GlosikUI`: Reusable SwiftUI components package

# License

This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.