A small deterministic renderer
From pixels to pencil strokes—inside your browser.
Sketch2Video is deliberately not an AI service. It combines conventional image processing with canvas animation and browser video recording.
1. Prepare the source
The browser scales large images to a practical canvas size. Edge-preserving smoothing suppresses small camera and texture noise while retaining stronger colour boundaries.
Nothing is uploaded. Refreshing or closing the page clears the in-memory session.
2. Abstract coherent lines
Two Gaussian scales form an XDoG-inspired dark-line response. Adaptive high and low thresholds keep weak marks only when connected to strong contours, then small isolated components are removed.
Clean mode keeps sparse dark contours. Textured pencil mode adds restrained deterministic hatching without an image model.
3. Draw lines, then add colour
The line mask is thinned into connected paths, ordered across the page, and accumulated stroke by stroke. Colour follows as a separate guided, diagonal, or marker-band pass.
The pencil follows the active path. It approximates a readable drawing order without claiming to recover the original artist’s exact sequence.
4. Record the canvas
Canvas capture and MediaRecorder encode the animation. WebM is the reliable option in Chromium and Firefox; MP4 appears only when the browser reports direct support.
Because recording happens in real time, a 10-second video takes roughly 10 seconds to create.
Questions
Good to know
Does it use AI?
No. The current renderer uses local image-processing and browser media APIs only.
Does it need a Worker?
No. The site is static and the rendering work runs on the visitor’s device.