This looks interesting.
Yeah, that does look very interesting. I’ll be curious to hear a report back if/when anyone here tries it in Flame. Thanks for sharing.
Louis was showcasing this on the IBC showfloor with Flame. Quite impressive, especially if you don’t want to spend time setting up Comfy locally. It is available for Mac, you can download and use it for free if you’re running your own ComfyUI server. You can use their token system and use their server if you don’t want to fire it up locally.
Super nice work.
Even a thumbless monkey can install and run this toolset in minutes.
I’m typing this with index fingers.
Wish I had thumbs, and an imagination.
I have no idea why I would use it.
No Linux option at this point then?
Word is they’ll have a beta going and plan to release in November.
I look forward to the day when there is a legally trained AI model that is willing to indemnify its user base that doesn’t produce images that suck.
Not sure why TVCs escape the same legal minefield that features and episodic need to tippy toe through.
Hahahaha. Yeah. But I mean, we’ll have world peace, zero food insecurity, and cured cancer before this happens.
I think TVCs get away with it because of their short lived nature and the scale of potential targets. Even IF a court case was brought and won, it’s against one company out of thousands. The sued company says, “our bad, won’t happen again. We’ll stop airing that ad.” (or more likely, “we had no idea and it was our VFX vendor who did these crimes”), pays a nominal fine and stops airing the ad. The ad has likely been off air for months or years by the time the case is settled. It’s all very low risk.
What’s that phrase that I learned n America?
Better to ask for forgiveness than permission.
What’s that phrase I’m learning during my citizenship?
Nobody gives a fuck and the law is available for purchase…
FOMO and
replacing a runnign ad is cheap now if anyones sues so the risk is nothing
Also applies to Europe.
Ain’t no thing but a chicken wing kiddoes.
So I was curious about iNodes and signed up for their trial. I’ll throw €7 in the pot to see it first hand.
But they handle the Linux version differently than the Mac version. So you have to sign up for different plans.
Update: Made it through the Flame install and started running some prompts. Still working through other aspects, including ComfyUI integration.
Got the ComfyUI side working with some caveats:
Here running the LTX ComfyUI workflow directly from Flame and getting the matte back.
It’s working, but considering a beta solution is probably a fair assessment. For the ComfyUI part. For the other models, its usable as is.
If you need a quick NanoBanana T2I or T2V workflow, this is a lot simpler than some alternatives.
Hi Jan, thanks for testing this.
iNodes is not a beta anymore, including the ComfyUI part.
With this kind of workflow, you can get different types of output back in Flame.
What caveats did you hit? Happy to help here.
Hi Benjamin,
The mentioned caveats were the studio setup and all the things we discussed.
Looking forward to re-installing 1.1 and seeing how this compares now. In the queue.
OK, thanks allklier.
We’ve been having a lot of fun with something else: Lightricks’ LTX-2.5 Alpha Gen, launched right from Flame Batch through our iNodes Comfy bridge. You give it a plain RGB clip, no green screen, no mask, no prompt, and you get back a premultiplied RGBA EXR sequence (half float, ACEScg) with the matte in the alpha. There’s an example below.
This one is 1920x1080, 89 frames, just under 4 minutes on an RTX 5090 on our ComfyUI server. The source is a generated clip, so real plates are next on our list.
It’s not magic, of course. It picks the foreground on its own and you can’t steer it, it tops out at 1920x1088 and 145 frames, soft edges keep a bit of background colour, and we got a faint matte on the ground around the feet. Also, the weights are under the LTX-2.x Community License, which means a paid licence above $10M in annual revenue.
If you like it, we’d be really happy to share the Comfy templates and a walkthrough in a dedicated post. Just tell us!
Oh, and we have the same thing for LTX-2.5 SDR to HDR: an SDR clip goes in, an ACEScg EXR sequence comes back.
Cheers,
Benjamin,
Yes, LTX Alpha Gen has taken our world by storm, and lots of folks have been experimenting with it. We had a very active thread (LTX 2.5 Alpha gen) and numerous native implementations that came from it.
You can nudge the model with a region of interest. We also found ways of solving the drift that happens in fast-moving footage.
Thanks for sharing! ![]()
I’m sure you’ve already done some great work with LTX in ComfyUI!
My intention was simply to show how easy it is to connect your own ComfyUI workflows to iNodes, using a simple input/output setup.
LTX Alpha is just one example. With iNodes, you can integrate any model, including EXR sequences, alpha channels, and much more.
The idea is to let artists experiment freely with any model, directly in their favorite software, without having to develop a specific OFX plugin for each one.
I was actually taking the opportunity to build on your LTX Alpha experiments and share some of our first iNodes templates!
It’s also a nice way to show how much the integration has improved, with support for multiple inputs, EXR sequences, MP4 files, and OpenClip.
I thought your example was a perfect opportunity to share these new possibilities and open up the discussion!
Absolutely. iNodes is a great solution to bring a variety of models and workflows in Flame, as a fully supported tool, and I have it installed as you know.


