🚀 Send to ComfyUI

Send images from the web directly into a ComfyUI workflow with a single right-click.

Send to ComfyUI is a lightweight browser extension + custom ComfyUI node that creates a seamless Web → ComfyUI pipeline.

Instead of manually downloading images and loading them into ComfyUI, you can send them directly from your browser.

No more:

Image → Download → Drag into Load Image → Click Run

Now it's simply:

Right-Click → Send to ComfyUI ✨

And with Auto-Run enabled:

Right-Click → Send → Workflow runs automatically 🚀

---

🎬 Demo

<p align="center"> <img src="assets/demo.gif" width="800"> </p>

Example: Right-click an image on the web and instantly send it into a running ComfyUI workflow.

---

🌐 What This Tool Does

When browsing the web, you often encounter images you want to experiment with inside ComfyUI.

Normally the process looks like this:

1. Download the image
2. Open your downloads folder
3. Drag or upload the image into a Load Image node
4. Click Queue Prompt

This extension removes that entire cycle.

With Send to ComfyUI, images move directly from the browser into your workflow.

New Workflow

Web Image → Send to ComfyUI → Image appears in ComfyUI/input → Workflow loads the image via a Smart Queue

Or with Auto-Run enabled:

Web Image → Send to ComfyUI → Image uploaded → Workflow starts automatically

Fast. Clean. Zero friction.

---

🧠 How It Works

This project consists of two components:

• A browser extension • A custom ComfyUI node

Together they create a seamless bridge between the internet and your AI workflows.

---

🖱️ Browser Extension

The extension adds a new option to the browser's right-click menu:

Send to ComfyUI

When selected, the extension:

1️⃣ Detects the image URL 2️⃣ Acquires the image data robustly (see below) 3️⃣ Converts it into binary format 4️⃣ Uploads it directly to your running ComfyUI server

Robust image acquisition

Many web images can't be read by a naive background fetch — cross-origin/CDN images come back opaque (HTTP status 0), and blob:/data:/lazy-loaded sources aren't visible to the service worker at all. The extension tries, in order:

1. data: URLs — decoded directly.
2. Opaque-tolerant fetch (no-cors) — opaque blobs are still uploadable, so this recovers most cross-origin images instead of failing with "Error 0".
3. Page-context capture — if the above yields nothing, it captures the image from the source page's own context (a page-origin fetch with that site's cookies/CORS, then a <canvas> fallback). This handles hotlink-protected, blob:, and lazy-loaded images.

The image is automatically saved inside:

ComfyUI/input/

Each image receives a unique timestamp filename such as:

web_image_1710243800.png

This guarantees the newest file can always be detected by the workflow.

---

🖱️ Popup Controls

Clicking the extension icon opens a small popup with two controls:

▶ Run Open Tab — manually fires a run in the active ComfyUI tab. Useful when Auto-Run is off or when you want to trigger a run without sending a new image.

Auto-Run on Send — when checked, the workflow starts automatically every time you right-click and send an image. No need to open ComfyUI or click anything.

How Auto-Run fires

Auto-Run uses a two-step approach to trigger the workflow reliably across ComfyUI frontend versions:

1. Sends a Ctrl+Enter keyboard event to the ComfyUI tab
2. If the queue hasn't increased within 500ms, falls back to clicking the Run / Queue button directly (matched by text, aria-label, or title, so it works across frontend versions)

Send and Auto-Run are independent

Capture, upload, and auto-run each run in their own error boundary. A failed auto-run never blocks or masquerades as a failed send: if the image uploaded but the run couldn't fire, you get a distinct "Sent ✓ but auto-run …" toast, and the image is already in your input folder ready for a manual Run. This is the fix for the old behavior where a single hiccup (e.g. a blocked tab injection) would silently kill auto-run.

Site access required for Auto-Run. Auto-Run injects the run trigger into the ComfyUI tab, which Chrome only permits if the extension has site access to it. With unpacked extensions Chrome often defaults this to "on click", which silently blocks background injection and produces a "Cannot access contents of the page" error. Fix it once: chrome://extensions → Send to ComfyUI → Details → Site access → On all sites.

---

🤖 Custom ComfyUI Node (Smart FIFO Queue)

The included Load Image Queue (FIFO) node is designed specifically for automated pipelines. Instead of blindly grabbing the newest file in your folder, it acts as a smart queue.

The node:

• Tracks Sessions: Records exactly when ComfyUI starts, ignoring all images from previous sessions. • Processes in Order (FIFO): If you rapidly send 3 images from your browser, it lines them up and processes them in the exact order they arrived (Image 1 → Image 2 → Image 3) without skipping any. • Graceful Fallback: If the queue is empty, it acts like a normal node and loads the newest image in your input folder so manual workflows never break. • Safe IS_CHANGED: Returns a stable string instead of NaN when the input folder is empty, preventing ComfyUI from looping unnecessarily. • Race-safe reads: Auto-Run can fire the instant the browser finishes uploading, so the file may briefly be missing, zero-byte, or half-written. The node retries with backoff until the file size is stable and PIL fully decodes it (img.load()), instead of throwing "image not found" / a truncated-image error on a freshly landed file. This is what makes Auto-Run reliable.

---

🔄 Reset Queue Toggle

The node includes a Reset Queue toggle widget directly on the node:

| State | Behaviour | |---|---| | Normal FIFO queue (default) | Processes images in the order they were sent | | Reset: grab latest | On the next Run, clears the queue state and loads the absolute latest file in the input folder |

This is useful when an offset has developed — for example if you sent images while ComfyUI was not running. Simply toggle it on, hit Run once to resync, then toggle it back off to resume normal FIFO order.

---

⚡ Auto-Run Mode (Zero-Click Pipeline)

The extension includes an optional Auto-Run mode toggled from the popup.

When enabled, the workflow starts immediately after the image is uploaded.

Manual Mode

Right-Click Image → Send to ComfyUI → Open ComfyUI → Click Run

Auto-Run Mode

Right-Click Image → Send to ComfyUI → Workflow runs automatically 🚀

This keeps execution fully visible inside the ComfyUI interface:

• nodes lighting up • progress bars • workflow execution

---

🔔 Visual Notifications

The extension provides feedback directly on the webpage using toast notifications.

Examples:

✅ Sent to ComfyUI! 🚀 Sent & Auto-Queued! ❌ Could not connect — is ComfyUI running? ⚠️ Sent ✓ but auto-run blocked: grant this extension "On all sites" access. ⚠️ Couldn't read image: …

These notifications confirm whether the transfer succeeded, and — because the stages are independent — distinguish a failed send from a failed auto-run.

---

🖼️ Image Handling

Images downloaded from the web often contain orientation metadata.

The custom node automatically:

• corrects EXIF rotation • converts images to RGB • normalizes pixel values • converts images into Torch tensors

This ensures compatibility with ComfyUI's internal image format.

---

📦 Installation

1️⃣ Install the Extension

Open your browser and navigate to:

chrome://extensions

Enable Developer Mode.

Click Load Unpacked and select the extension folder.

Extension Permissions

The extension uses the following Chrome APIs:

• contextMenus — adds the Send to ComfyUI right-click option • tabs — locates the active ComfyUI tab for Auto-Run • scripting — fires the run trigger in the ComfyUI tab, and captures the image from the source page when a direct fetch is blocked • storage — saves the Auto-Run preference • declarativeNetRequest — allows requests to the local ComfyUI server • host permissions (<all_urls>) — needed to read images from arbitrary sites and to inject the run trigger; remember to set Site access → On all sites (see Auto-Run)

---

2️⃣ Install the Custom Node

Clone or copy this repository into:

ComfyUI/custom_nodes/send-to-comfyui/

Restart ComfyUI.

You should now see a node called:

Load Image Queue (FIFO)

---

⚙️ Configuration

The extension assumes your ComfyUI server is running at the default address:

http://127.0.0.1:8188

If your ComfyUI instance uses a different port or address, update the endpoint inside:

background.js
popup.js

---

🧪 Example Workflow

A minimal pipeline might look like:

Load Image Queue (FIFO)
        ↓
Preview Image
        ↓
Caption / Generation Node

Any image sent from your browser will immediately queue up and appear in the workflow.

---

⚠️ Known Limitations

Queue offset when Auto-Run fires with the popup closed — the extension fires a run immediately on send, but if ComfyUI's input folder already contains unprocessed images from a previous session, the node will process those first. The visual strip (if enabled) may not reflect the actual processing order in this case. Use the Reset Queue toggle on the node to resync.

No guaranteed sync across sessions — the node's queue state is in-memory and resets on ComfyUI restart. Images sent before a restart will be treated as new files in the next session.

Auto-Run needs tab site access — there is no ComfyUI REST endpoint to run the currently loaded graph, so Auto-Run must inject into the ComfyUI tab. If you see "Cannot access contents of the page", set the extension's Site access to On all sites (see the Auto-Run section).

---

🛠 Requirements

• Chrome or Edge browser • ComfyUI running locally • Default port 8188

Tested with:

• ComfyUI 0.16.3 • ComfyUI Frontend 1.39.19

---

💡 Example Use Cases

This tool can support many workflows:

🧠 Caption generation 🎨 Image-to-image experimentation 📚 Dataset creation 🔍 Reference image collection ⚙️ Rapid AI workflow prototyping

---

🔮 Possible Future Improvements

Potential ideas for extending this tool:

• automatic workflow selection • batch image sending • storing original image URLs as metadata • drag-and-drop browser integration • multi-workflow routing

---

❤️ Why This Exists

Working with reference images should feel effortless.

Instead of constantly downloading files and manually loading them into nodes, this project turns the internet into a live input source for ComfyUI.

Right-Click → Send → Create.

---

🙏 Acknowledgements

Parts of this project were developed with the assistance of AI tools, including Anthropic Claude, OpenAI ChatGPT and Google Gemini, which were used for brainstorming, debugging, and refining parts of the implementation.

The overall design, implementation, and integration of the extension and ComfyUI node were carried out by the repository author.

---

📜 License

MIT License

---

⭐ Enjoy!

If you build interesting workflows using this tool, feel free to extend it further.

The bridge between the web and ComfyUI opens up many creative possibilities.