rss2podcast
# rss2podcast <img src="resources/ui.jpg" width=600px></img> A locally hosted, AI generated podcast from an rss feed.      Powered by - <img src="resources/rss.png" width=80px height=80px></img> <img src="resources/ollama.png" width=80px height=80px></img> <img src="resources/coqui.png" width=80px height=80px></img> <img src="resources/kokoro.jpg" width=80px height=80px></img> ## Features - RSS feed parsing and article extraction - Article summarization using Ollama - Text-to-speech conversion using multiple engines: - Kokoro TTS (Recommended) - MLX Audio TTS - Coqui TTS - Podcast generation with customizable settings - Web interface for configuration and control ## Requirements - Go 1.21 or later - Ollama (for article summarization) - One of the following TTS engines: - Kokoro TTS (recommended) - MLX Audio TTS - Coqui TTS ## Installation 1. Clone the repository: ```bash git clone https://github.com/intothevoid/rss2podcast.git cd rss2podcast ``` 2. Install dependencies: ```bash go mod download ``` 3. Configure the application by editing `config.yaml` or using the web interface. ## Configuration The application can be configured using the web interface or by editing the `config.yaml` file. The following settings are available: ### RSS Settings - `url`: The RSS feed URL to parse - `max_articles`: Maximum number of articles to process - `filters`: List of filters to apply to articles ### Ollama Settings - `end_point`: The Ollama API endpoint - `model`: The Ollama model to use for summarization ### Podcast Settings - `subject`: The podcast subject - `podcaster`: The podcaster name ### TTS Settings - `engine`: The TTS engine to use ("kokoro", "mlx", or "coqui") - `kokoro`: Kokoro TTS settings - `url`: The Kokoro TTS API endpoint - `voice`: The voice to use - `speed`: The speech speed (0.25 to 4.0) - `format`: The audio format (mp3, opus, flac, wav, pcm) - `mlx`: MLX Audio TTS settings - `url`: The MLX Audio TTS API endpoint - `voice`: The voice to use - `speed`: The speech speed (0.5 to 2.0) - `format`: The audio format (mp3, wav) - `coqui`: Coqui TTS settings - `url`: The Coqui TTS API endpoint ## Usage 1. Start the application: ```bash go run cmd/rss2podcast/main.go ``` 2. Access the web interface at `http://localhost:8080` 3. Configure the application using the web interface or edit `config.yaml` 4. The application will: - Parse the RSS feed - Extract and summarize articles - Convert the summary to audio using the selected TTS engine - Generate a podcast file ## TTS Engines ### Kokoro TTS (Recommended) Kokoro TTS offers OpenAI-compatible speech synthesis with support for multiple voices and formats. It provides excellent quality with low latency. ### MLX Audio TTS MLX Audio TTS is a powerful text-to-speech engine that provides high-quality speech synthesis with support for multiple voices and formats. It offers additional features like direct audio playback and output folder management. ### Coqui TTS Coqui TTS provides high-quality speech synthesis with support for multiple voices and formats. ## License This project is licensed under the MIT License - see the LICENSE file for details. ## Acknowledgments - [Ollama](https://ollama.ai/) for the LLM API - [Kokoro TTS](https://github.com/kokoro-tts/kokoro) for the TTS engine - [MLX Audio TTS](https://github.com/mlx-audio/mlx-tts) for the TTS engine - [Coqui TTS](https://github.com/coqui-ai/TTS) for the TTS engine ## How it works The application reads an rss feed, extracts the articles and summarises them. RSS + Ollama + TTS = Podcast ### RSS The application reads an rss feed and extracts the articles. Each of these articles are then processed by scraping the article content. ### Ollama The application uses a locally hosted version of Ollama. The Ollama API is used to summarise the article content. Default model used is mistral:7b ### TTS The summarised article content is then converted into an audio podcast using the Coqui TTS API. ## Dependencies This project requires the following dependencies to be installed on your system. ### Ollama You can install the Ollama server by following the instructions on the [official website](https://ollama.com). Ollama needs to be running on your local machine for the application to work. The application is configured to use the default Ollama server URL `http://localhost:11434/api/generate`. This can be changed via the config.yaml file. ### ffmpeg `ffmpeg` is a command-line tool for handling multimedia files. It is used to convert the generated audio files to the MP3 format. #### macOS You can use Homebrew to install `ffmpeg` on macOS: ```bash brew install ffmpeg ``` #### Windows 1. Download the `ffmpeg` build for Windows from the [official website](https://ffmpeg.org/download.html). 2. Extract the downloaded ZIP file. 3. Add the `bin` directory from the extracted folder to your system's PATH. #### Linux The installation command depends on your Linux distribution. ##### Ubuntu/Debian ```bash sudo apt update sudo apt install ffmpeg ``` ### Kokoro TTS (Recommended) Kokoro TTS is a text-to-speech synthesis system that uses deep learning to create human-like speech from text. You can install the Kokoro TTS server by following the instructions on the [official website](https://github.com/nazdridoy/kokoro-tts). #### Docker: Create a docker-compose.yml file and add the following: ```yaml services: kokoro-fastapi-cpu: ports: - 8880:8880 image: ghcr.io/remsky/kokoro-fastapi-cpu:latest # or v0.2.3 for last stable version ``` Start the server by running the following command: ```bash docker compose up -d ``` This will start the Kokoro TTS server on port 8880. The server provides a REST API for text-to-speech conversion. ### Coqui TTS Coqui TTS is a text-to-speech synthesis system that uses deep learning to create human-like speech from text. You can install the Coqui TTS server by following the instructions on the [official website](https://coqui.ai/tts). #### Docker Start the container by using the following command: ```bash docker run -d -p 5002:5002 --platform linux/amd64 --entrypoint /usr/local/bin/tts-server ghcr.io/coqui-ai/tts-cpu --model_name tts_models/en/ljspeech/vits ``` ### MLX Audio TTS MLX Audio TTS is a text-to-speech synthesis system that uses deep learning to create human-like speech from text. You can install the MLX Audio TTS server by following the instructions on the [official website](https://github.com/mlx-audio/mlx-tts). #### Docker As of this writing, MLX Audio TTS needs to be run locally as Docker does not allow GPU access on Apple Silicon. ```bash # Install the package pip install mlx-audio # Create a virtual environment python -m venv venv # Activate the virtual environment source venv/bin/activate # Install the dependencies pip install -r requirements.txt # Run the server mlx_audio.server ``` rss2podcast will automatically request the MLX Audio TTS server to generate the audio file. ## Testing To run the tests, use the following command: ```bash go test ./... ```