Vectorize Pro

If you've ever tried to turn a PNG logo into a clean SVG, you know the pain. Auto-trace tools (Potrace, Inkscape's trace bitmap) shred flat artwork into hundreds of anchor points - a circle becomes a knot of tangled paths and the colors come out muddy. People who do this for a living will tell you vector conversion is a fiddly manual job, minutes of parameter-tweaking per file at minimum. Vectorize Pro (class AugmentPNGToSVGPro) is a ComfyUI node that outsources that job to a paid cloud API and hands you back an SVG you can actually open in Illustrator.

The name is the honest one. This is not a local model. There's no checkpoint, no VRAM hit, no model download - you feed it an image and an API key from Augment Studio (the same folks behind the upscale-pro pack), it uploads your image to their servers, and a few seconds later you get SVG bytes back. It's the "AI vectorizer" backend that services have been selling for years, exposed as a node in your graph.

How it works

Mechanically it's a straightforward client. The node converts the incoming image tensor to PNG bytes, POSTs it to augmentstudio.app/api/process with your key as a Bearer token, and gets back a request_id. Then it polls the job status every second - up to sixty of them - and when the job reports done, it fetches the finished SVG and returns it. If the request times out or the API returns an error, the node raises and your queue stops, so a hung job is loud rather than silent.

Inputs and outputs that matter

Three things worth knowing:

• api_key - paste your Augment Studio key. It's the one input you'll actually fiddle with, and it's what unlocks (or blocks) everything.
• image - any IMAGE from ComfyUI. Load Image, a KSampler output, whatever's at the end of your wire.
• svg output - a custom AUGMENT_SVG type that only plugs into this pack's Export SVG node, not into generic nodes. That's the point: it's a specialized pipe, not a universal one.

The outputs are svg (the vector file), json_result (a small JSON blob with the node name and byte size, handy if you're scripting), and trigger (a "done" signal for chaining steps). Wire svg straight into Export SVG and you're finished.

One real gotcha: the node only processes the first image in a batch - image[0]. Feed it a batch and everything after the first frame is silently ignored.

Installing it

This is one of the rare ComfyUI nodes where the hard dependency is a credit card, not a model file. Install is easy:

• ComfyUI Manager: search for "augment Vectorize Pro" and hit install.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/augment-lib/vectorize-pro.git augment-vectorize-pro, then restart ComfyUI.

Dependencies (numpy, Pillow, requests, torch) are all things ComfyUI already ships - no extra wheels, no local weights. What you do need is an API key from augmentstudio.app/pricing, and that costs money; the source literally comments "Paid Node, credit values may change." The README suggests storing the key once with the Variable Set/Get nodes from the separate augment-ComfyUI repo, so you're not pasting it into every workflow.

Where people get burned

• Privacy: your image leaves your machine. For client logos and unreleased work, that's a real consideration - check their pricing and terms before uploading proprietary assets.
• Timeout: the client gives up after 60 seconds. Large or busy images can exceed that on a loaded API, and you'll get a RuntimeError instead of an SVG.
• Cost: if you vectorize once a month, a free tool or a redraw in Illustrator is probably the right call. This is for people converting logos and icons constantly and needing output clean enough to hand off without cleanup.

Honest take: for logo work this is the one I'd reach for over local auto-trace, because the quality gap is real - that's the whole pitch, and the pack is honest about being a service. Just know what you're signing up for before you make it a workflow dependency.