LTXVAddGuidesFromBatchIndexes

LTX Video's keyframe conditioning is its superpower: you can pin specific frames of the latent to specific images, and the model generates the in-between. LTXVAddGuidesFromBatchIndexes is the node that makes that practical - it takes a batch of guide images and stamps them into the latent at the frame indices you specify. It's a refined version of LTX's built-in LTXVAddGuide, rebuilt to work from a batch and to treat larger batches as an image sequence.

It lives in the ComfyUI-Easy-Media pack, which is a strong sign of where this pack's heart is: a bunch of video-media tooling wrapped around a really good keyframe-conditioning helper.

How it works

You feed it positive, negative, vae, latent, and images, plus three controls:

• image_indexes - comma-separated frame indices, e.g. 0,61,121. Each image lands on its index. Leave it empty and the node evenly distributes the valid (non-black) images across the latent length for you - a nice default.
• img_compression - how much compression to apply to each guide image (0–100, default 18). LTX guides work better when the image is compressed to the model's latent-friendly detail level.
• strength - how strongly each guide pins to its frame (0–1, default 1).

The node encodes each image into the latent space, finds the matching latent index, and appends it as a keyframe guide. Nine or more images in the batch flips the behavior: instead of individual keyframes, the whole batch is treated as one image sequence, injected as a run of frames starting at the first index. That's the mode for driving LTX with an actual reference clip rather than a few stills.

Outputs are the three things you put in: positive, negative, and the modified latent (with a proper noise mask). Wire them into your LTX sampler. The code is a descendant of kijai's KJNodes LTX implementation, so if you've used that family, the behavior will feel familiar.

When to reach for it

First/last-frame animation, storyboard-driven clips, and keeping character consistency across a generated sequence. Pair it with LTXVMakeRefVideo (also in this pack) when you want an IC-LoRA-style reference sequence built from a batch of stills.

Install

It ships in ComfyUI-Easy-Media:

cd ComfyUI/custom_nodes
git clone https://github.com/yolain/ComfyUI-Easy-Media.git

restart ComfyUI (or ComfyUI Manager, "Easy-Media"). It's a conditioning node, so no extra models - but it only works with LTX's ComfyUI integration, so have ComfyUI-LTXVideo available and the LTX model files.

The gotcha

Guides that land past the latent's frame count get skipped with a warning - make sure image_indexes stay inside the latent length, or some of your keyframes silently won't apply. Also, the 9-image sequence threshold is a real behavior switch; if you meant five discrete keyframes and feed nine images, you'll get a sequence instead. Know which mode you're in.