Are you a N00b like me?

UPDATE: 14:44 PDT 1 Nov 2025

Just uploaded v1.2. Took in the suggestions from all the commenters - and THANK YOU for that! The speed and quality both got much better after that! Y'all rock!

All these workflows seem complex as hell, right? What nerd knobs do I turn? Why did I break all my shit suddenly.

I spent several days playing with Coyotte's Moviemaker - GREAT workflow, fun to use (if you're smarter than me) - and decided I needed to dumb it way the fuck down for, well, me. You might benefit from it, too.

How the workflow operates:
You supply a starting image and a prompt. Just look for green boxes.

There are some things you can tweak, but the ones in the purple (they look pink to me, but whatevs) should be touched only if you have some idea what you're doing. The black boxes are for model loads - those are probably best to leave to advanced people. The reds DO NOT TOUCH, or you'll fuck it all up.

It uses JoyCaption to analyze your image, and build the first part of a combo prompt, uses LoraManager to pull the trigger words from your LoRAs, and merges your prompt into it. It does some magic with samplers (again, I'm new, you're new, neither of us knows how that shit actually works, so we call it magic) and drops that down to your "working" video. Keep an eye on that; if you don't like how that turns out, don't waste GPU time upscaling and adding sound to it.

USING LORAMANAGER:

First, make sure you used Comfy Manager to install LoraManager.

Now, go to http://[yourIPAddress]:8188/loras

If you've never used LoraManager before, it may take a minute to index all your LoRAs. That's fine. You'll find yourself wanting to reorganize them, anyway.

Now, select a LoRA, and, when you mouse over it, you'll see a little "Send" arrow.

Pick whichever LoRA loader seems most appropriate for what you're doing right now.

Then switch back to your ComfyUI tab. You'll see it in the list, and the trigger words next to it.

Those trigger words will automatically be added to the prompt. Use the slider bar to determine what LoRA strength you want, and click any of the trigger words you do NOT want, and it'll deselect them.

And that's it. It'll spit out some videos for you, and you can have a good time with it.

On a 5090 (that's what I have, so I didn't test with anything else), it takes 30 minutes to do a ten-second, non-looping video with sound. That's not bad, considering all the stuff it does on the way there. 

On cards with less VRAM, you might want to try switching the JoyCaption settings WAY WAY down. You'll have to unpin and expand it in the "String Ops" section in the lower right. And think 3-4 seconds max if you've got < 24GB of VRAM.

INSTALLATION:

If you're new to Comfy-UI, install ComfyUI manager. No, I won't teach you how to do that; that's what Google is for. Once that's installed, open this workflow. Now, open Comfy-UI manager, and select "Install Missing Nodes". It'll do that, and you restart Comfy-UI from within ComfyUI Manager. Once that's done, reload the page, open Manager again, go to Model Manager, change your filter to "In Workflow", and then install all those models. THIS MAY TAKE A WHILE. They're huge.

Now, that MIGHT not install the MMAudio models. If not, download models from here: https://huggingface.co/Kijai/MMAudio_safetensors/tree/main

Put them into: ComfyUI/models/mmaudio

THANKS KIJAI!

Nvidia bigvganv2 should be downloaded automatically, but if it's not - https://huggingface.co/nvidia/bigvgan_v2_44khz_128band_512x

Put everything from nvidia's huggingface into: ComfyUI/models/mmaudio/nvidia/bigvgan_v2_44khz_128band_512x

Good luck! Happy moviemaking! 

If you think my workflow sucks, you won't hurt my feelings. Tell me what I can improve. This is literally my first workflow. 

If you like it, well, hey, that's awesome!