TranscribeAI is a cross-platform desktop app that uses Large Language Models to transcribe audio files and scanned images/pages. It features:

• LLM-powered transcription (e.g. Google Gemini via google-generativeai)
• Resumable image workflows (skips already-done files, cleans up partial outputs)
• Real-time logs & progress in the UI
• Persistent settings (stores your API key with electron-store)
• Drag-resizable, searchable sidebar for managing transcripts

Note: A “headless” version (no UI) is also available and can be integrated into your own system—see
https://github.com/Minitex/TranscribeAI

Technologies

• Electron: Main process for file I/O and IPC
• React + TypeScript: Renderer UI, bundled with Vite for fast HMR
• Vite: Modern build tool for instant feedback and optimized production builds
• google-generativeai SDK: Interfaces with LLMs for high-quality transcription
• electron-store: Simple JSON storage for your Gemini API key

Setup & Installation

1. Obtain your API keys

   • Gemini (required for audio/Gemini OCR): Sign in to AI Studio and create a key at the Google AI Studio API Key Console. Copy the generated key.
   • Mistral (only for Mistral OCR): Create an API key in your Mistral account—follow the Mistral quickstart at docs.mistral.ai/getting-started/quickstart.

2. Download TranscribeAI

   • Go to the TranscribeAI Releases page.
   • Choose the installer for your OS:
     • macOS: .dmg
     • Windows: .exe
     • Linux: .AppImage or .tar.gz
   • Download and run the installer. Because TranscribeAI is an open-source project and we don’t bundle a paid code-signing certificate, you may see a security warning the first time you run it:
     • macOS Gatekeeper (“Unidentified Developer” or "damaged and can't be opened"): open System Preferences → Security & Privacy, then click Open Anyway next to the TranscribeAI entry.
     • Windows SmartScreen (“Windows protected your PC”): click More info, then Run anyway.
   • Follow the installer prompts to complete installation.

3. Configure your API key(s)

   • Launch TranscribeAI.
   • Click the gear icon in the top-right corner to open Settings.
   • Paste your Google Gemini API key into the “API Key” field.
   • If you plan to use Mistral OCR, paste your Mistral API key into the “Mistral API Key” field.
   • Click Save.

4. Run your first transcription

   • Click the file picker button to select an audio file or image folder, then choose the output folder for your transcripts.
   • Click Transcribe to begin transcription.
   • Monitor progress and logs in real time.

Using Mistral OCR (including batch mode)

• In Settings, enter your Mistral API key (leave Gemini key as well if you use Gemini features).
• In the Image tab, pick the mistral-ocr-latest model.
• Select an input folder of images/PDFs and an output folder.
• Enable Batch mode to process files in batches, then start transcription. The app will call Mistral’s batch OCR and write outputs to your chosen folder.
• Adjust Batch size with the +/− controls (default 50, range 10–500) to balance throughput vs. request size.
• Note: single-image Mistral OCR calls are free; batch OCR requires a paid Mistral account.

Updating to a New Version

• macOS: Download the latest .dmg from Releases, open it, and drag the app to Applications. Choose “Replace” if prompted. If Gatekeeper blocks, right-click → Open once.
• Windows: Download the new installer .exe from Releases and run it; it overwrites the existing install. Tip: after first launch, right-click the TranscribeAI icon on the taskbar → Pin to taskbar so it’s easy to find next time.
• Linux (AppImage): Download the new *.AppImage, chmod +x if needed, and replace your old AppImage file.
• No uninstall needed: Install over the top; your existing data and settings remain.

Screenshots