Transition Images In Batch

Give this node a batch of images and it builds the in-between frames that transition from one to the next - a crossfade, a slide, a door wipe, a circle iris. It's the "slideshow with transitions" node, and it saves you a trip to a video editor for the simple case. People use it two ways: to make a quick animated slideshow or GIF out of a handful of stills, and to generate transition frames to feed into a video model or an interpolation pass.

It's part of KJNodes, kijai's utility pack. Nothing fancy under the hood - it's compositing math, not a model - which is why it's fast and dependency-light.

How it works

Between each adjacent pair of images in the batch, it inserts a run of blended frames that morph the first into the second using the transition style and easing you pick. The number of inserted frames per transition is up to you, and the easing curve controls how the motion feels - linear is mechanical, ease_in_out is smooth, and the fun ones (bounce, elastic, glitchy) add character. The output is one long image batch: your originals with the transition frames stitched in between.

The inputs that matter

• images - the batch of stills (you need at least two).
• transition_type - the effect: fade (crossfade, the safe default), horizontal/vertical slide, box, circle, horizontal/vertical door.
• interpolation - the easing: linear, ease_in, ease_out, ease_in_out, bounce, elastic, glitchy, exponential_ease_out.
• transitioning_frames - how many frames each transition spans. More = slower and smoother; this is your main "speed" dial.
• blur_radius - softens the transition edge (nice on slides and wipes).
• reverse - plays the transition the other way.
• device - CPU or GPU; push it to GPU if the batch is large.

Output is a single IMAGE batch - send it to a video-combine or save node.

Installing it

ComfyUI Manager → search KJNodes for ComfyUI → install → restart. Or cd ComfyUI/custom_nodes && git clone https://github.com/kijai/ComfyUI-KJNodes, then pip install -r ComfyUI-KJNodes/requirements.txt, and restart.

Where people get burned

• Mismatched image sizes. The blend assumes the images line up. Feed it stills of different dimensions and you'll get stretching or errors - resize them to a common size first.
• Batch explosion. transitioning_frames multiplies. Ten images with 30 transition frames each is a big batch, and every frame is a full image in memory - it's easy to balloon your RAM/VRAM without noticing. Start small.
• It's a compositor, not a generator. The in-between frames are blends of your inputs, not newly generated content. If you want true motion interpolation (invented intermediate frames), that's a job for a frame-interpolation model like RIFE - this gives you clean editorial transitions, which is a different thing.