The service requires an NVIDIA GPU and the resulting ts-gpu image is approximately 26GB. It also includes ports for SSH, HTTP, Ollama, and Meilisearch. The web interface features include audio file upload/download, task completion alerts with interactive links, and a HTML5 web player with speed control and transcription highlighting. The service can be customized and troubleshooted through the .env file and Dockerfile. However, the creators warn that this is example code and should not be used in production environments without additional security measures.
Key takeaways:
- Transcription Stream is a self-hosted diarization service that works offline and includes features like drag and drop diarization and transcription via SSH, a web interface for file management, summarization with Ollama and Mistral, and full text search with Meilisearch.
- The service provides SSH drop zones and a web interface for easy use and implementation into workflows, and allows for complex operations on transcriptions using Ollama.
- The service requires an NVIDIA GPU and the resulting ts-gpu image is approximately 26GB in size.
- It also provides instructions for building and running the service, including automated install and run scripts, and details on customization and troubleshooting.