EvoLink Topaz Video Upscale (Official)

Every generation node in the EvoLink pack hands you a video, but it's a compressed API export - fine for a look, not for delivery. Topaz Video Upscale is the cleanup step: the same platform, running Topaz's AI video upscaling, to sharpen and bump the resolution. It's the node that turns a "generate → save" graph into a "generate → edit → upscale" pipeline, which is exactly how the pack's README frames the arrangement.

How it works

The notable thing about this node is what it doesn't take: a prompt. The tooltip says it flat out - Topaz is a processing model, it doesn't read prompts. What it needs is video_urls, a required multiline field of video URLs to process, and the intended wiring is dead simple: take the result_urls output from whatever EvoLink generation node made your clip and feed it in here. result_urls is literally formatted for this - one URL per line, which is exactly what video_urls expects.

Then pick upscale_factor: 1 enhances without enlarging, 2 and 4 scale the resolution up, and the tooltip notes the factor affects billing. The output is a VIDEO you save with the native Save Video node, same as any generation. Timeout defaults to 1800 seconds because this is the slow, expensive-looking part of the pipeline - processing takes real time.

The inputs that matter

• video_urls - required; wire from an upstream node's result_urls, or paste public URLs one per line.
• upscale_factor - 1 (enhance only) / 2 / 4, billing scales with it.
• prompt - required by the schema but unused; the tooltip tells you not to bother filling it. Leave it empty.

Outputs

The standard EvoLink five: video (downloaded, → Save Video), task_id (receipt for the job log), status_info, result_urls, response_json. Because it emits result_urls itself, you can stack two passes - though honestly, if 2× then another 2× is what you're planning, do the math on the billing first.

Install and gotchas

Part of the comfyui-evolink pack. Manager → search EvoLink → Install → fully restart ComfyUI (close the console window; a refresh doesn't count). Or:

cd ComfyUI/custom_nodes
git clone https://github.com/deeplearning-goethe/comfyui-evolink

Only requests as a dependency; no models, no GPU. API key from evolink.ai/dashboard/keys, fill once, blank afterward.

The classic mistake: trying to feed this node an image/video tensor directly. It's a URL-in, URL-processed node - the video_urls text field is the interface, and your upstream node's result_urls output is the natural source. Also remember result links die after 24 hours, so don't build a workflow that stalls for a day between generate and upscale. Otherwise pack convention applies: 401 bad key, 402 no credit, 429 rate limit, free failed tasks, cached-and-unbilled reruns.