FL Fal Pixverse Transition

The "morph from image A to image B" node: give Pixverse a starting frame and an ending frame, and it generates the video that plausibly gets from one to the other, prompt-guided. It's the transition-shot tool - two product photos, two poses, two locations, and Pixverse fills in the motion between them.

The mechanism

Same v5 Pixverse backend as this pack's other Pixverse nodes, a transition-specific endpoint. first_image and last_image (both optional) anchor the two ends; prompt steers what happens in between rather than describing the whole scene from scratch, since the model already has both endpoints to work from. Like the pack's other Pixverse nodes, batch_size (1-5) runs multiple generations in parallel, each landing in its own frames_N output - running several at once on the same anchor pair is a reasonable way to get a handful of different takes on the same transition to pick from, since the model has real freedom in how it interpolates between two fixed points.

Inputs and outputs that matter

prompt - the default is literally "Scene slowly transition into cat swimming under water," a good illustration of what this expects: describe the transition, not a static scene. first_image/last_image anchor the two ends; you can supply either, both, or neither depending on how much you want to lock down. aspect_ratio covers 16:9/4:3/1:1/3:4/9:16, resolution runs 360p-1080p, duration is "5" or "8" seconds (longer gives the transition more room to breathe but costs more), style offers the same six presets as the other Pixverse nodes (none/anime/3d_animation/clay/comic/cyberpunk), and negative_prompt ships the same sensible long default. nth_frame (1-4) is the frame-extraction stride again.

Outputs: frames_1 through frames_5 (one per parallel batch slot), video_urls, status_msg.

Installing it

ComfyUI Manager: search Fill-Nodes, install, restart. Manually: cd ComfyUI/custom_nodes && git clone https://github.com/filliptm/ComfyUI_Fill-Nodes, then restart.

Where people get burned

Leaving both first_image and last_image empty turns this into a fairly generic prompt-only video generation rather than a controlled transition - if you specifically want the morph behavior, wire up at least one anchor, ideally both. The two anchors being wildly mismatched in composition, subject, or framing is the single biggest cause of a transition that looks like a jump-cut instead of a smooth morph; Pixverse can only interpolate so much, and two images that already share rough framing get dramatically better results than two unrelated shots.

Same billing caveat as the rest of this pack's video nodes applies: batch_size and resolution both scale cost fast, so keep batch_size at 1 while you're finding the right prompt.