Kaira Transcribe Panel

Kaira Transcribe Panel is a web app for local & cloud speech-to-text (STT) using Whisper models and major cloud APIs (using your own API Key). Compare model speed, quality and pricing.

Try it here: https://landonikko.github.io/Kaira-Transcribe-Panel

• Local Models: Tiny, Base, Small, Medium, Turbo, Large (via openai-whisper).
• Cloud Models (BYOK): OpenAI (Whisper-1, GPT-4o Mini, GPT-4o), Deepgram Nova-3 (Summarization & Topics), Gladia (EN), ElevenLabs Scribe v1.
• Outputs: Paragraphs, Segments, SRT, VTT, TSV.
• Media Sync: Player with waveform, real-time highlighting, click-to-seek.
• Themes: E Ink, Nova, Minty, Toffee.
• Queue Support: Batch processing.
• Price Calculator: Estimates API costs.
• Utilities: Copy/download, notifications, stats.

E Ink	Nova
Minty	Toffee

1. Clone: git clone https://github.com/kairauser/KairaTranscribePanel.git && cd KairaTranscribePanel
2. Python & PyTorch: Python 3.8+. Install PyTorch from pytorch.org.
3. FFmpeg:
   • macOS: brew install ffmpeg
   • Ubuntu/Debian: sudo apt install ffmpeg
   • Windows: Download from ffmpeg.org & add to PATH.
4. Dependencies: pip install -r requirements.txt

1. Run: python Launch_Kaira.py
2. Open: http://localhost:5000 in your browser.
3. Cloud API Keys: Paste into UI when prompted (stored in browser localStorage).
   • OpenAI
   • Deepgram
   • Gladia
   • ElevenLabs

• Translation features (API & LLM).
• Additional transcription models.
• In-browser Whisper (ONNX, Transformers.js).
• ARIA compliance.
• Local WaveSurfer.js bundling.
• Modular cloud support.

   

Contributions, bug reports, and feature requests are welcome! Fork the repo, create a branch, make changes, test, commit, push, and open a Pull Request.