ImaginationVellum: Generative-AI Ideation Canvas with Spatial Prompts, Generative Strokes, and Id...
summary
ImaginationVellum is a working multimodal canvas that uses the entire 2D surface as an active prompt: spatial arrangement, proximity, and composition of ink, text, images, and intermediate outputs directly modulate generation. The system implements Generative Strokes (semantic, transformation, variation, style), proximity-dependent global vs local intent tags, and mappings from distance to prompt weights, temperature/CFG and control-net strength. It captures timelines via temporal replay and provenance graphs. A small user study shows fast exploratory ideation but notes tradeoffs from folding many parameters into single strokes.
AI-written summary · about
chapters (10)
- 0:00Introduction to ImaginationVellum
- 1:17Spatial prompting and generative strokes
- 4:02Spatial prompting explained
- 6:50Mapping spatial dimensions to generation
- 7:11Semantic generative stroke demo
- 7:32Transformation generative stroke
- 8:14Variation generative stroke
- 8:56Style generative stroke
- 9:59Preserving ideation history
- 11:03Summary and Q&A
