The project provides a high-performance, distributed API for automatic speech recognition using OpenAI’s Whisper model and its faster variant. It leverages an async producer-consumer architecture with a thread pool for multi-GPU parallel inference, enabling scalable transcription and translation workflows. Integrated crawlers automate fetching audio from social platforms like Douyin and TikTok into the pipeline. Future plans include LLM integration (e.g., ChatGPT) to extend beyond transcription into natural language processing.