Soblerg/FLUX.1-Kontext-dev-demo

FLUX.1 Kontext [dev] — demo

In-context image editing: upload a picture, describe the change in plain language, and the model applies it while leaving the rest of the scene intact.

• —Model: black-forest-labs/FLUX.1-Kontext-dev
• —Licence: FLUX.1 [dev] non-commercial

How this Space works

This is a static Space. The page runs entirely in the visitor's browser and calls the /infer endpoint of the official black-forest-labs/FLUX.1-Kontext-Dev Space, which hosts the weights on ZeroGPU. No image is stored here.

Visitors can optionally Sign in with Hugging Face, in which case requests are made with their own token and draw on their own ZeroGPU quota rather than the shared anonymous pool.

Why static rather than Gradio + ZeroGPU

Hosting the weights directly needs a Gradio Space on ZeroGPU, which requires a PRO subscription (or a free account older than 30 days with a verified email, or a community grant). A self-hosting version of this demo is kept ready in `zerogpu/`:

• —zerogpu/app.py — Gradio app loading FluxKontextPipeline in bf16, @spaces.GPU-decorated, input resized to a 1280px max side rounded to multiples of 16, MCP server enabled
• —zerogpu/requirements.txt
• —zerogpu/README.md — the frontmatter to use for that variant

To switch: promote those files to the repo root, set sdk: gradio in the frontmatter, move the Space to the zero-a10g flavor, and set an HF_TOKEN secret with access to the gated model.