# AI Transcriber plugin for Obsidian This [[Obsidian]] plugin enables speech-to-text transcription using OpenAI's AI models. It converts audio files (MP3, M4A, WAV, FLAC, OGG, AAC) and video files (MP4, M4V, MOV, AVI, MKV, WEBM) into markdown notes directly within your vault. Users can choose between three transcription models: GPT-4o Transcribe for highest quality, GPT-4o Mini Transcribe for speed and cost-effectiveness, or the traditional Whisper model. The plugin includes optional post-processing to improve accuracy, a custom dictionary feature supporting multiple languages (Japanese, English, Chinese, Korean) to correct frequently misrecognized terms, and automatic language detection with manual override. It supports both server-side and local voice activity detection (VAD). The plugin is desktop-only (requires Obsidian v1.8.0+), stores API keys securely using Obsidian's encryption, and has a 500MB file size limit for external files (no limit for vault-stored files). ## References - GitHub: https://github.com/mssoftjp/obsidian-ai-transcriber - Download: https://obsidian.md/plugins?id=ai-transcriber ## Related - [[Large Language Models (LLMs)]] - [[Whisper]] - [[Obsidian plugins]]