Comfy-Flame

Flame Tools and Tech
,
1

Hey everyone :waving_hand:

I’ve been working on a little side project that I wanted to share with the community: a bidirectional integration between Flame and ComfyUI.

The idea is simple — right-click a clip in the media panel, pick a workflow, and Flame exports the frames to ComfyUI, processes them, and automatically imports the result back into your library. No round-tripping through Finder, no copy-pasting paths.

How it actually works

  1. Build your workflow in ComfyUI using two custom nodes I made: LoadFromFlame (which reads the frames exported by Flame) and SendToFlame (which drops the output into a watched folder). You build your AI pipeline between those two nodes — upscaling, style transfer, ControlNet, whatever you want.

  2. Save that workflow as a .json in a dedicated folder (flame_comfy_workflows/). From Flame’s Settings dialog you can mark your favorites, which will pin them at the top of the right-click menu for quick access.

  3. Export from Flame: right-click a clip → ComfyUI → select your workflow. Flame exports the clip as a PNG sequence into a staging folder, then automatically loads the workflow in your browser with the correct clip already wired into the LoadFromFlame node.

  4. Run the queue in ComfyUI — the frames get processed and dropped into the output folder.

  5. Auto-import back into Flame: a background watcher detects the new files and imports them directly into your library, organized in a timestamped folder (CLIP_NAME_HHMM).

Two more things worth knowing:

First, workflows can run fully automatically — no queue button, no browser interaction. Flame exports, ComfyUI processes, Flame imports. You just wait for the result to appear in your library. The only difference is that you save your workflow in API format instead of the standard one (Settings → Enable Dev Mode Options → Save as API Format in ComfyUI).

Second, the LoadFromFlame node supports multi-clip workflows: just select several clips in Flame before triggering the export, and each one gets wired into its own node instance. Useful for any workflow that needs to compare or combine multiple sources — a background plate, a matte, a grain reference, whatever you need.

It works both locally (Flame and ComfyUI on the same Mac) and over a network (Flame on Mac, ComfyUI on a remote Linux workstation via Tailscale).

Fair warning: I’m not a developer. I’ve been building this with a lot of trial and error (and a lot of AI assistance :sweat_smile:), so the code is probably not the cleanest or most elegant thing you’ve ever seen. But it works, and that’s what matters to me right now.

What I’m really looking for is feedback — things that don’t make sense, stuff that could be done better, edge cases I haven’t thought of. And beyond the technical side, I’d love to build a small community around this where we can share workflows, tricks, and ideas for how to use AI tools inside a proper Flame pipeline.

I’ve probably forgotten a few details here and there, so if anything is unclear or you have questions, don’t hesitate to ask — happy to help!

The project is on GitHub: GitHub - gasparmb/Comfy-Flame: A bidirectional workflow bridge between Autodesk Flame and ComfyUI, enabling AI-powered image processing directly from Flame's media panel with automatic result import.

Would love to hear your thoughts. :folded_hands:

PS: I have used Flame 2026 to work on this, so I don’t thinks it work on 2025

26 Likes
2

Blow Your Mind Wow GIF by Product Hunt

1 Like
3

Very cool project, thanks so much for sharing your efforts.

4

@finnjaeger might be interesting for your setup

5

Does it support only mac for local workflow?

6

The auto-load in Safari is Mac-based (AppleScript), so that part won’t work on Linux. I don’t have a Linux machine running Flame to test and adapt it — if someone wants to tackle that part, contributions are very welcome!

7

nice! but PNG only ?

8

@kily I tried to change the code to add support for Flame on linux, here is the file (you need to install the others from the Github and just replace this one)

You also need to install Chrome and Playwright. Don’t know if it’s working but as soon as I have a Linux in hand, I would try.

comfy_integration.py (101.1 KB)

Screenshot 2026-02-18 at 08.44.40
Screenshot 2026-02-18 at 08.44.401508×584 35.8 KB

1 Like
9

Thanks, yes for the moment I only manage to do PNG export (even if the settings say you can export in EXR). I haven’t used ComfyUI much yet — I’m just starting to integrate it into my workflows, so I’ll update that kind of feature when the need arises.

1 Like
10

Thanks Gaspar. I’ll give a try.

11

Hey everyone,

First off, big thanks to @gasparm for the ComfyUI ↔ Flame integration — great concept and implementation.

Over the last 2 days, I adapted his script to run on Linux (tested on Flame 2025.2.2) with no external dependencies — runs directly on Flame’s built-in Python and is more tailored to my workflow.

Still testing, but here’s a summary of what changed:

Zero dependencies — works with what Flame already ships (PySide6 + native libs), no pip install needed

Flame-native UI — styling consistent with Logik Portal scripts

Prompt editing — when running a workflow (API Auto) with text nodes, a dialog opens to edit the prompt before sending to ComfyUI

Pipeline path tokens — for export/result paths (, , , etc.) with symlinks into ComfyUI

Auto-import — to Batch Schematic Reel “ComfyUI Results”, with timestamped names so previous results aren’t overwritten

Frame range selection — choose between all frames, current frame, or custom range before exporting (for now only tested with image edit workflows)

Export caching — if the clip was already exported, asks whether to reuse or re-export

Improved FlameLoad node — forces re-execution every run without needing to restart ComfyUI; not sure if this existed before, but when trying to re-run the workflow it would error out on the loader node

Once I validate on more Flame versions and other workstations, I can share these modifications with everyone. I’m also using Vibe Coding, so I’m not sure if some of these changes are tied to my specific install environment — still need more testing.

Wilton

15 Likes
12

Really well done

1 Like
13

I knew this guy would do something like this.

Well done mate.

FYI: There’s a workflow for minimax remover that has blockswap which is useful for running longer or highrez videos in case you are lowvram user. AFAIK This BMO minimax doesnt let you do blockswapping.

1 Like
14

@wiltonmts - Wilton, you monster… :rofl:

1 Like
15

Thanks a lot! Really cool, I wasn’t aware of this more optimized MiniMax workflow.

I have a VRAM limitation since I share Flame and ComfyUI on a single workstation with a 4090. Because of that I’ve developed the habit of cropping regions, doing downscale/upscale inside my ComfyUI workflows and using them on shots with fewer frames. If possible, could you point me to where I can find that MiniMax version you mentioned? Really appreciate it!

16

Thank you for what you did on the project. I’m not exactly sure how GitHub works, but I could give you access to the project so you can update it (or maybe create a Linux branch if this version isn’t compatible with Mac, and I’ll update the Mac version to include the improvements you made).

Great job in any case! Having the prompt directly in Flame was on my to-do list.

2 Likes
17

@gasparm - this is a great idea.
@wiltonmts is a great collaborator.
I’m happy to join in and do code-sharing on non-commercial endeavours.
Let me know

2 Likes
18

Thanks for the invitation @gasparm , really appreciate it. Happy to help out however I can. And @philm , thanks for the kind words mate — you’re the man!

19

@gasparm To keep the group thread clean, I’ve sent you a DM to sort out any details. Cheers.

20

what can I say?
it looks great.

1 Like