Search our docs, check out common issues, or ask the community
From zero to your first transcribed meeting in 5 minutes
Get Signifyr running on macOS or Windows
Walk through transcribing sign language in a test meeting
Grant permissions and connect your meeting platforms
How Signifyr processes sign language in real-time
Fine-tune detection accuracy and customize your experience
Camera placement, lighting, and model tuning tips
Choose between on-device and server-side transcription
Save meetings as text, PDF, or structured JSON
Adjust how live captions appear during meetings
Build with Signifyr's WebSocket-based Python API powered by a highly trained deep learning CNN
Connect to real-time transcription streams via WebSocket. The API processes video frames through a Convolutional Neural Network (CNN) trained on extensive ASL datasets to recognize hand shapes, movements, facial expressions, and spatial relationships.
Learn about Signifyr's deep learning architecture. The CNN analyzes video frames in real-time, extracting features from hand positions, finger configurations, facial expressions, and body language to accurately transcribe ASL gestures into text.
Self-host Signifyr's Python backend with the CNN model. Deploy the inference server locally or on your infrastructure for maximum privacy and control.
Submit PRs, report bugs, and improve ASL models. Help train and refine the CNN model with new datasets and techniques.
Extend Signifyr to support regional ASL variations. Fine-tune the CNN model with custom datasets for specific dialects or signing styles.
Quick fixes for the problems people actually encounter
Grant camera permissions and troubleshoot access
Check lighting, camera angle, and model configuration
Optimize performance on older machines
Reduce latency and improve real-time processing speed
We're an open source project built by volunteers. Join Discord to chat with maintainers and other users, or open an issue on GitHub.