ComfyUI × PhotoCraft Bridge Nodes: A Two-Way AI Image Workflow

When PhotoCraft Meets ComfyUI

Discussion around PhotoCraft has been relentless in the community lately — some are excited, others skeptical. But the more pragmatic approach is to simply solve the fragmentation problem between tools yourself. A developer shared on Reddit a small project they built with Claude's help: a set of Bridge Nodes connecting ComfyUI and PhotoCraft, filling the gap left by the absence of any official interoperability between the two.

The core value of this project is straightforward: you can do your creative work and prompt writing in PhotoCraft, hand off the actual image generation to ComfyUI, and send the results back to PhotoCraft as a separate layer — all without installing any plugins in PhotoCraft, communicating entirely through its control channel.

ComfyUI is an open-source Stable Diffusion frontend built on a node graph interface, where users construct image generation workflows by connecting functional nodes — highly flexible and extensible, beloved by power users and developers alike. PhotoCraft is positioned more as a professional image editing and creation tool, offering an experience closer to traditional image editors, with built-in prompt writing and layer-based editing. The two tools share a heavily overlapping user base but emphasize different strengths — ComfyUI excels at complex generation pipelines, while PhotoCraft prioritizes the editing experience — which is precisely why a "bridge" is needed. The control channel is a data exchange interface that PhotoCraft exposes externally, allowing external programs to send commands or read state without requiring any extensions installed inside PhotoCraft. This is the technical foundation that makes the zero-plugin approach possible.

How the Two Nodes Work

The solution consists of two ComfyUI nodes with clearly defined roles.

Get Image Node: Pulling Data from PhotoCraft

The Get Image node is responsible for receiving content from PhotoCraft. It returns an 8-bit .png image, a mask, width and height dimensions, and a prompt. Notably, the prompt isn't a separately designed input field — it reuses PhotoCraft's built-in Notes feature. If multiple notes exist, the node automatically concatenates them. This design lets users organize their prompt content within their familiar creative environment without switching interfaces.

The node also supports prompt streaming from within PhotoCraft, meaning it can be embedded fairly flexibly into almost any existing ComfyUI workflow, rather than serving purely as an isolated input endpoint.

Prompt streaming here means the node can dynamically receive prompt updates from PhotoCraft during workflow execution, rather than reading a static input only when the workflow starts. Users can modify prompt content in PhotoCraft's Notes and have those changes take effect without reconfiguring the ComfyUI workflow. This is especially useful for iterative prompt workflows — traditionally, cross-tool prompt changes require manual copy-pasting, whereas the streaming mechanism keeps both sides in near-real-time sync.

Send Image Node: Sending Results Back

Once generation is complete, the Send Image node returns the image and mask to PhotoCraft, placing them on their own dedicated layer. This "pull out — process — push back" loop is exactly what cross-tool collaboration needs most. For users who rely on PhotoCraft for layer-based editing while depending on ComfyUI's powerful generation capabilities, these nodes chain the strengths of both tools together.

Real-World Use Cases

The developer provides two example workflows that directly demonstrate the bridge nodes in action:

• Flux2 Inpainting: Uses the bridge nodes to pass images bidirectionally between PhotoCraft and ComfyUI for a Flux2-based inpainting workflow.
• Krea2 → PhotoCraft: Generates images with Krea2 and sends the results back to PhotoCraft.

These two examples cover both the "bidirectional round-trip" and "one-way return" patterns — essentially the two most common cross-tool collaboration scenarios. Whether you're doing fine-detail inpainting or importing freshly generated images into PhotoCraft for further processing, these nodes eliminate the tedious manual import/export steps.

Flux2 is a recently well-regarded open-source image generation model series known for its generation quality and fine-grained control. Inpainting refers to regenerating only the area marked by a mask while leaving the rest of the image untouched — commonly used to remove unwanted elements, fix artifacts, or swap out specific regions. Krea2 is another AI image generation tool focused on rapid generation and style exploration. Together, these two example workflows cover the most mainstream cross-tool use cases: combining PhotoCraft's precise masking with ComfyUI's model ecosystem for inpainting, and importing outputs from other generation tools directly into PhotoCraft for post-processing.

What to Know Before Installing

This project is still in its early stages, and the author openly acknowledges "it's not finished yet" — so don't expect too much from the example outputs. A few things are worth noting before you dive in:

• You must read the installation instructions on GitHub. The setup process is not one-click — it requires configuring a dedicated directory and passing several command-line arguments to PhotoCraft to establish the connection.
• Pay attention to the limitations and security notes at the bottom of the project page. Any data exchange involving a control channel is worth evaluating on its own security merits.
• The current functionality is more of a working proof of concept. The author explicitly states there's "certainly room for improvement," but it "seems to do the job" for current needs.

The project is open-source on GitHub (the-aSak/ComfyUI-Photocraft-Bridge), and anyone interested can grab the nodes and example workflows to try out.

A Direction Worth Watching

With PhotoCraft not yet offering an official extension ecosystem and community sentiment decidedly mixed, this kind of bridge tool — built rapidly by an individual developer with AI assistance — captures exactly what makes the open-source community vital: don't wait for the official solution, just build it yourself. It may not be mature yet, but it offers a viable path for pairing professional image editing tools with open-source generation engines. As the author put it — "Let's see where this goes" — the evolution of this direction is worth keeping an eye on.