CURRENT PROJECTS
"If you're going to try, go all the way. Otherwise, don't even start. - Bukowski"
🎬 ABOUT THE NEXT PROJECT
NOTE: Research info has been moved to the Research page
I finished "Footprints in Eternity" in late June which was a huge undertaking of 80 days for just 10 minutes of footage.
After that, I took some time out to test new software and upgrade my ComfyUI setup.
Once I finish the current research phase, then I will decide on the next project. Maybe "The Highwayman stageplay" that I have been using for research test examples.
Current thinking is that I will begin to focus on the next project around early October 2025.
I'll also post my ongoing side-projects here. This will include anything related to the AI film-making process. Check them out below.
AIMMS - Automated Intelligent Movie Making Software (Vrs 1.0)
(I hope to have version 1.0 ready to release as an "image & video storyboard management system" under freeware licensing after further testing and before the end of 2025)
I found it really hard to stay organised on "Footprints In Eternity" as the takes mounted up, and I had folders full of images and videos with different versions of shot names and lost track of much of it by the end.
Though I was organised with the naming and tracked the original ideas in a csv, it still got surprisingly chaotic. That was for just 120 shot names. Imagine a full movie with over 1200 shot names but multiple takes of not just each base image but videos, mask videos, character swap videos, lipsync videos etc... That is a lot of footage to keep track of.
The lesson from that was - as Comfyui gets better at making video for short films, I need to get better at tracking what I am doing, and seeing at a glance what I have to work with. No more hunting through folders trying to find a shot I did two weeks ago, but can't remember what I called it, or where I saved it.
So, meet AIMMS...
It's a front end that can interact with Comfyui in the backend, while keeping all the takes organised and easily visible in one place.
I'm currently working on version 1.0 and hope to include an AI "helper" option for things like improving video prompts. For that, I am currently trialling using llama.cpp and a quantized Deepseek r1 Q4 GGUF LLM model installed locally, which works fine on my 3060 RTX 12GB VRAM in initial tests.
I hope to push a freeware version of the prototype when that's completed. I'll see how far I get with it, how much work it takes, and what I'm legally permitted to license for use. It's making progress, but whether it would be viable for some else to install and use, I'll have to decide down the road.
GITHUB PROJECTS
A COMFYUI CUSTOM NODE - "SAVE IMAGES AS VIDEO (FFMPEG)"
Something I needed for Comfyui workflows to function properly with AIMMS, was a video export node that wouldn't automatically add numbers to the filenames, and would allow for a folder name to be more easily defined. I couldn't find one suitable, so ended up forking a Github project I found and adapting it.
I published the node to github as AIMMS - Comfyui Save Images As Video node.
IMAGE 2 REVERB
This takes a still image of an environment and creates an IR wave file for use in any IR reverb to create the sound environment "reflection" size, based on the image.
I've barely touched this other than to test it. It works in Python, so I shared it "as-is" after I tweaked it a bit.
I have plans to turn it into a ComfyUI custom node at some point. Maybe also provide an open source IR reverb of some sort with it and inject it into the audio chain just before the video output. Also providing the IR result as a .wav file that can be used in post-production. A quick check with Grok confirmed this should easily be possible in a lightweight approach see code example here
For image2reverb progress check the github page here image2reverb