Transcribe Panel

Transcribe Panel is an interface for local and cloud transcription models for comparing their output speed, quality and pricing.

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

The hosted version is a local interface that works by using your own API keys and caches the data in your browser.

For testing local models, see the Python Setup.

Features

Local Models (OpenAI Whisper)

• Tiny, Base, Small, Medium, Turbo, Large

Cloud Models (Bring Your Own API Key)

• OpenAI: Whisper-1, GPT-4o, GPT-4o Mini
  • 4o supports custom prompting for translations and localization
• Deepgram: Nova-3
  • With AI summarization & topic detection
• Gladia: Gladia v2
• ElevenLabs: Scribe-v1, Scribe-v1-Experimental

Smart Features

• Example Files: Multi-language samples (EN, FR, JP)
• Batch Processing: Queue multiple files
• Media Player: Audio/video preview with waveform visualization
• Live Sync: Transcript segments are highlighted by audio position
• Multiple Formats: SRT, VTT, TSV, plain text, numbered segments
• Transcript Editing: Edit transcripts
• Upload/Import: Import existing transcript files
• Cost Calculator: API pricing estimates per file
• UI Themes: E Ink, Nova, Minty, Toffee
• Statistics: Word count, character count, processing time

API Keys (Cloud Models Only)

All cloud providers offer free starting credits, with some offering monthly refills:

• OpenAI
• Deepgram
• Gladia
• ElevenLabs

API keys can be stored in your browser the same way your passwords are saved.

Python Setup (For Local Models)

Requirements

• Browser, Python 3.8+, (PyTorch pytorch.org), FFmpeg

Installation

1. Clone the repository

   git clone https://github.com/LandoNikko/Transcribe-Panel.git
   cd Transcribe-Panel

2. Install Python dependencies

   pip install -r requirements.txt

3. Install FFmpeg (required for media processing)

   • Windows: Download from ffmpeg.org
   • macOS: brew install ffmpeg
   • Linux: sudo apt install ffmpeg

4. Run the application

   python Launch_Transcribe_Panel.py

   Open http://localhost:5000 in your browser

Tips

• Local models: No internet required, complete privacy. Relies on your hardware (GPU vram), so transcriptions can be slow.
• Cloud models: Fast, advanced features, requires API keys. Costs, but very cheap. Can be used heavily with just free credits.
• File support: Most audio/video formats (MP3, WAV, MP4, etc.)
• Batch mode: Select multiple files and process them sequentially.
• Export options: Copy to clipboard or download in various formats.

License

LGPL-2.1 License - see LICENSE file.