Topaz Diagnostics

Every node in the ComfyUI-TopazVideoLocal pack depends on one fragile thing you have no direct visibility into: whether this machine has a usable, licensed Topaz Video installation and the models to go with it. Topaz Diagnostics is the pack's answer to "why doesn't anything work" - it scans for the installation and reports everything in one readable string, without running a workflow and without needing any other node wired in. The pack's own instructions say it plainly: start here, before opening an example workflow.

What it reports

Feed the report STRING output into a text preview node (or any node that displays text) and you get a structured dump:

• Installation - where Topaz Video was found, its ffmpeg path, model directory, version, and whether its ffmpeg actually provides the tvai_* filters. That last check is the whole ballgame: a folder only "counts" as a Topaz install if its ffmpeg has tvai_up.
• Codecs - most notably whether a software H.264 decoder exists. Topaz compiles its ffmpeg without one, which is expected here because this pack feeds raw frames through a pipe and never needs a decoder. The report tells you this so you don't go hunting a problem that isn't one.
• Models - the model directory and what's available.
• Licence - the current licence/lock status.
• Other apps - the CLI-lock status of Topaz Photo and Gigapixel (both locked behind enterprise licences; the report surfaces it so that finding doesn't have to be rediscovered).
• Config path - where the pack's own config.json lives.

The two switches

• refresh (default on) - re-scan for the installation and rebuild the model list. Leave it on unless you're diagnosing something and want a cached view.
• check_license (default off) - run a real one-frame Topaz job to verify the licence actually works, not just that the config file says so. Takes a few seconds. The pack caches licence checks, so a check that times out counts as unknown, never as invalid - but when you genuinely suspect licensing, this is the definitive test.

The node is marked so it always re-runs when you trigger it - it's a diagnostic, not a pipeline stage, so there's no caching of results.

When to reach for it

The intended moment is the very first run after install. Wire report into a text preview, hit run, read the output. Most first-run failures in this pack reduce to one of the lines in that dump: no installation found (then set install_path on the Engine Settings node), models missing weights (then enable allow_model_download, or run the model once in the Topaz app), or a licence problem (then open Topaz Video and sign in). Every example workflow in the pack ships with this pair already wired up, grouped as "Before the first run".

Install

It ships with the pack:

git clone https://github.com/donangel85/ComfyUI-TopazVideoLocal.git

from ComfyUI's custom_nodes folder, restart ComfyUI. No pip step beyond numpy. ComfyUI Manager finds it by searching "TopazVideoLocal". The node itself needs nothing extra - it just inspects your system - but everything else in the pack still needs Windows plus a licensed, signed-in Topaz Video installation.

Common issues

If the report itself says "NOT FOUND" for the installation, that's the pack's auto-discovery failing, and the fix is on the Engine Settings node: an explicit install_path pointing at your Topaz Video folder. If you see a model you expected but marked [download required], the weights aren't on disk. And if a licence check comes back unknown rather than valid, don't immediately blame the pack - open Topaz Video, sign in, and re-run with check_license on.