Show HN: A canvas to explore AI image models (open-source, BYOK) (koubou.app)

Koubou is an open-source, BYOK (Bring Your Own Key) AI canvas for generating, editing and compositing images with any image model you choose to connect. It’s a lightweight web-based workspace that supports text-to-image and image-to-image workflows while keeping API keys and model access under your control — a key privacy and compliance advantage for teams that don’t want to hand secrets to third-party UIs or be locked into a single provider. On the usability side Koubou offers an “infinite” pan & zoom canvas, multi-image selection, drag-handle resizing, delete/Download (right-click), image upload and clipboard paste to quickly compose and iterate visuals. Technically it’s model-agnostic: you provide the model endpoint or API key (self-hosted or cloud) and the UI handles generation/editing requests, making it easy to prototype interfaces for Stable Diffusion, OpenAI/Replicate endpoints, or private inference servers. For the AI/ML community this matters because it provides a transparent, extensible playground for stitching together model-driven image pipelines, doing dataset curation or iterative editing without surrendering key material — a practical, auditable base for research, production prototyping, and creative workflows.