Looping help / advice please

[adamreading]

Hi

SO I watched the videos and built a most basic version of what I am doing to test the later stages - where looping would come in.

The idea is there will be 4 start images - each one representing 1 scene of a short movie. The loop will use the scene number as the filename prefix - so you end up with 4 mp4's called 1... 2.... 3... 4...

Before they are compiled - they are trimmed to the sharpest end frame from the last 10 frames (not a public node yet) and the final frame is grabbed for each scene. So far so good with your nodes (thank you!)

BUT - the next phase is to a) build scene extensions from each final frame, in exactly the same way - where they keep the same prefix for the scene they are extending. - So I obviously could just build another whole complete copy of that workflow again to the right - and again for a another extend etc - but I wondered if there was a way to return the last frames back to the beginning of the same loop - in the correct scene position?

B) - what I haven't yet factored, is that each scene will have its own prompt generated (with a series of vision and llm nodes) and the correct prompt has to be fed into the generator each time ( instead of just1,2,3,4), and also each extension last frame will have to go through its own florence and llm series before hitting the generator...

Any help or advice you can give on the best way to achieve this with your nodes would be appreciated.

[adamreading]

[adamreading]

Update - so I think I have cobbled together how to get the complex prompts to the right image - at the right time now...

So its just what the best way to add extending scenes from the last frame of each scene - a) as a new process tagged on the end or b) some way to reuse the middle section again maybe?

[justUmen]

Sorry I don't have time for that.
But you should probably start to organize your images and video in separate folders. (You can use my combine text node for that.)
You can recover images from a folder using one of my node too.
If you need you can save text using my save text node and recover it. You can store filenames or whatever in there.
If you want complex logic, you can try my if/else node.

[adamreading]

Fair enough - I'll find another solution. Thanks for responding anyway. I used to have the individual folder all set up:

But that didn't work for the video combining stage at the end - that has to gather all the videos from every scene and extension and put them all in order as one final video, hence the solution with prefixing the scene names and storing everything in one subdirectory for each project - that can then be simply combined.

[adamreading]

Sorry I don't have time for that. But you should probably start to organize your images and video in separate folders. (You can use my combine text node for that.) You can recover images from a folder using one of my node too. If you need you can save text using my save text node and recover it. You can store filenames or whatever in there. If you want complex logic, you can try my if/else node.

Ok quick question - how would I use your loop node to take the output from this image filter (which could be 3-10 images) and pass them all through 1 at a time? When connected like this only the first gets made.

[adamreading]

Well for anyone else trying to loop images into video... I stayed up all night on this one but it now works:

Dynamic image selection + flexible loop size:
You send all 8 images through ImageBatchSplitter → ImageBatchMulti to keep them accessible as a unified list.
The user-picked indexes ("1,3,4" etc.) from Image Filter are:
🧠 Split into lines via TextSplitByDelimiter
🔁 Looped using LoopAllLines
🔢 Converted to integers
➕ 1 added (for ImageBatchMulti’s 1-based indexing)
🖼 Used to pick images one by one with Bjornulf_SelectImageFromList
Each selected image then flows cleanly through Florence2Run, Resize, and video generation nodes.