LTX - Multi images - Timeline Control (prompt relay) - custom workflow

Added: 2026-05-10

[00:00:12]FLAT BURSTS AHEAD. BRACE YOURSELVES AND PUSH HER THROUGH.

[00:00:25]COMMAND. What the hell was that?

[00:00:34]Hi, today I want to show you a new workflow for LTX where I have mixed the multi-image loader from What Dream Cost, a modded LTX sequencer with a timeline input for automatic sync with the timeline of the prompt relay node which has a timeline output for the sequencer.

[00:00:50]It allows you to use the timeline on the prompt relay node without having to calculate and re-enter time value in the LTX sequencer. The first image will be added at the beginning of the first segment. The second image will be exactly at the beginning of the second and so on. It is pretty flexible. You can also completely change whatever you want. For example, inserting only one image and make multiple prompt, multiple images with one prompt or even put a reference sheet as I will show you on this video. It is very flexible for making your own animation or even a short film. I'm going to explain to you different method with this workflow I have used to make the three scenes of the intro animation. So, as I have mentioned, this workflow has two modded nodes to sync the timeline prompt relay to the LTX sequencer. So, if you want to use it, I will leave a link in the description. Let's dive in. The first sequence is the three planes appearing from off- camerara view and then they just cruise in formation across the sky.

[00:01:44]Initially, I wanted them to appear from the left, but the camera was always making a rotation, which caused the plane to have issues with the propeller duplicating. It was a bit too much for the model. But I've just realized that I have actually written the wrong side in the prompt right instead of left. And that is the reason I didn't get exactly what I wanted. But in the end, I got a shot I quite like, which is this one.

[00:02:06]Anyway, it is just for demonstration. So for this shot, I've used two images, one without the planes and the other with.

[00:02:13]There is only one prompt in the segment prompt area. Since I have only one prompt and two images, I have only the first image synced. The second image is set manually in the sequencer. You can also tweak the strength so the original image will have more or less influence.

[00:02:28]If you want your original image to be as close as possible in your generated shot, set the strength TO ONE.

[00:02:35]>> FLACK BURSTS AHEAD. BRACE YOURSELVES AND PUSH HER THROUGH.

[00:02:45]Flack burst. For the second sequence, I've used two images as well, but with two different subprompts for the two segments and one master prompt that describes the overall atmosphere of the shot. There is a sync log text here, so you can check the sync of your images and various info. We can see here that the sync is enabled, but for the third shot, I'll use the manual mode, which means that I will disconnect the timeline connection between the prompt relay node and the LTX sequencer so I can insert a reference sheet.

[00:03:18]Command. What the hell was that?

[00:03:30]>> Command. What the hell was that?

[00:03:32]>> So, the third shot. As you can see here, the log says manual mode. It's because for this shot, I have disconnected the timeline. As you can see, the reason is I need the X-wing to appear from off camera while the planes are moving. At first, I've tried to put the complete image with the planes and the X-wing in the image like this, but it just does not work. So, with the sequencer is in manual mode or with the input timeline disconnected if you prefer, I just enter manually the frame number that correspond to the segment. And for the X-wing, I've set the frame to minus2 to use it as a ref for the second prompt.

[00:04:06]And it worked. It's not perfect, but not too bad. It picked up the X-wing as ref when I mentioned it in the second prompt instead of morphing plane like I have with no ref. So, as you can see, the workflow can be used in some different ways. And I find LTX to be very capable for an open- source model of this size.

[00:04:23]In this workflow, I also used a Laura that adds a bit of detail called the soft enhanced Laura with the FP8 model for speed. So, that's it for today. See you.