(Quick update: bad things happened during some update. it reset my tiny images to default 512x512, screwing up 2.0. I fixed it. please redownload 2.01 and rerun your tests...should see a huge improvement.)

Version 2: streamlined, added in Hunyuan upscaler option, and made things far easier to follow and understand.

Some testing post upload:
ran a flow from other models (most popular ones) and finding a difference of over 3 times faster to achieve the same end. found that at low resolution, this workflow is not as good as others (some suggestions please) but simply upping the resolution to say, 336x336 hits the right note of quality for the upscaler at the end to nail...still achieving a savings. 60 seconds for mine verses 149 seconds for theirs for an end result of a higher quality outcome with all things the same.

This works folks. give it a try...and give me suggestions that might help the low scale resolutions (seriously, how do the get 209x3somethings look decent? cherry picking?)

Whatever witchery is going on here, it works!

Janky Lightning Workflow
This workflow was a happy accident. While experimenting with ways to improve my other project (Janky Image 2 Video V4), I stumbled onto something that significantly boosted render speeds without sacrificing quality. It might be a new trick, or maybe I just missed the memo—but hey, it works for me, and I wanted to share it in case it helps you too.

What Makes This Different?

By leaning into low-resolution interpolation during decoding before going into hunyuan and pairing it with a simple step process, I managed to shave off a surprising amount of time. Render speeds? Fast. Setup? Easy. Results? Pretty solid for what it is.

A Few Notes:

• This isn’t a polished masterpiece—just an experiment I threw together.

• Use it, tweak it, break it, and share your findings. Let’s figure out together if this is the real deal or just some janky magic.

• If something feels off, drop a comment so it can be refined.

This is free to use and adapt for your own workflows. It’s not about claiming some revolutionary breakthrough—it’s about learning, testing, and sharing discoveries. If it works for you, post your times. If it doesn’t, let’s talk about why.

Thanks for checking this out. Now go make something awesome!