Origin Flow

For the issue of creative processes growing "invisible" in the age of AI, OriginFlow introduces the idea of "Creative Provenance". Unlike other search tools that find similar images, OriginFlow approaches the investigation from the final work, converting the objects, patterns, composition, and style in the image into investigation clues. It employs AI Agent to autonomously complete identification, search, comparison, inquiry, and reasoning, tracking potential sources and establishing connections among versions, time, and context.

It also generates a Provenance Graph, orchestrating reference material, AI generation, human editing, early versions, and the final work. It harnesses "confirmed", "probable", and "unknown" to distinguish the confidence of provenance links, avoiding direct equivalence between visual similarity and provenance evidence. Consequently, it transforms the previously fragmented and difficult-to-explain creative process into a traceable, verifiable body of evidence, enabling creators to explain the creative process and how AI was involved with greater transparency.

OriginFlow anchors its UI on "investigation process visualization" that is visually centered on the final work, leveraging nodes, edges, thumbnails, and evidence cards to display investigation results in tiers. Users only need to upload a work and select target styles, then follow the workflow of identification, investigation, comparison, connection, and reassembly to complete provenance tracing. Such dynamic status feedback and differentiation of closeness clarify the investigation process and uncertainty. In the end, OriginFlow yields a further verifiable visual report, providing a more transparent creative basis for work publishing, team collaboration, and copyright communication.