VRGDG Meta AI Browser Image

Every image model worth using has an API, a key, and a billing page - until you get to the consumer apps. Meta AI's image generator is genuinely good and genuinely awkward to use from ComfyUI: there's no official API. VRGDG Meta AI Browser Image solves that the only way that works: it automates a real Chrome tab. It opens Meta AI, pastes your prompt, optionally attaches images, waits for the generation, downloads the result, and hands it back as a normal ComfyUI IMAGE. No key, no token, no rate-limit dance - just a browser session.

How it works

The node drives Chrome through the remote-debugging protocol on a local port - debug_port, default 9224. The README's own framing is "Browser AI options where supported," and this is the image half of that feature set: the pack uses the same trick for storyboard images and GPT/Flow-style automation, which is why the default port is deliberately separate from those. Launch Chrome with remote debugging enabled on that port, log into Meta AI once, and the node takes it from there.

The prompt is the text that gets pasted and submitted. The interesting part is image_count: it controls how many of the optional image1…image50 sockets get attached before the prompt is submitted. Set image_count to 0 and you get text-only generation; set it to 3 and the first three image sockets you've wired in get uploaded via Meta AI's plus/upload control, then your prompt runs with them as context. This is how you feed the character-reference workflow into a model that has no official image-input API.

timeout_seconds (default 600) is how long the node will wait for a generation before giving up - Meta AI can be slow under load, and 60 seconds isn't enough. reuse_open_project defaults on, meaning it keeps reusing the current conversation tab instead of navigating back to the Meta AI start page every time; that's what makes batch runs tolerable.

The single output, image, is a standard ComfyUI IMAGE - decode-ready, wireable into anything downstream.

The inputs that matter

• prompt - what gets submitted to Meta AI.
• image_count - how many image sockets get attached; 0 for text-only. The image sockets are image1 through image50.
• debug_port - the Chrome remote-debugging port; keep 9224 unless you're already using it.
• timeout_seconds - raise it if slow generations time out.

Install

Pack-wide: ComfyUI Manager → search vrgamedev (or add https://github.com/vrgamegirl19/comfyui-vrgamedevgirl), restart, hard-refresh. Manual: git clone https://github.com/vrgamegirl19/comfyui-vrgamedevgirl into custom_nodes, pip install -r requirements.txt. On Windows portable, install Cython and scikit-build-core first.

Troubleshooting

The node can't open Chrome for you - you need a Chrome instance running with remote debugging. The standard incantation is launching Chrome with --remote-debugging-port=9224 (and a separate user-data-dir so your normal profile isn't locked), and you need to be logged into Meta AI in that instance. If the node hangs forever, check that the port matches exactly. If images come back without context, check image_count against how many sockets you actually connected - a socket wired but not counted gets ignored. And this is a browser automation node: it depends on Meta AI's web UI, so a UI redesign can break it between pack updates. Treat it as a fast creative loop, not a production pipeline.