Seam renders AI video transitions between two shots on your own GPU with LTX-2.5
Seam is a free, open-source app that takes two shots (images, short clips or 3D models) and renders a generated transition between them, such as smoke, water bend or a whip pan, with the open LTX-2.5 model in ComfyUI. It comes as a web studio, a command-line tool and a plugin that lets an AI agent plan and refine transitions.
What it does
Instead of cross-fading two flat frames, Seam asks LTX-2.5's first/last-frame mode for what happens between the shots, so smoke can really hide the subject and a pan gets parallax and motion blur. Ten effects passed the author's quality test (Smoke, Water bend, Whip pan, Lava, Beast, Ink, Light burst, Zoom through, Melt, Wind vortex); takes are 5 seconds, or 10 seconds for Smoke, Water bend and Whip pan. A small local model (Gemma-4 E2B) describes each shot and adds that line to the prompt, which you can correct first. A .glb model can be framed in the browser and recorded as a 3-second orbit, and the transition joins the two orbits. Every take can be reopened as its exact ComfyUI workflow.
Results and limits
The author measured it on an RTX 5090 (32 GB) with ComfyUI Desktop 0.41.3: 5-second takes at 1280×704 took a median of about 41 seconds from submit to finished video, and 10-second takes about 76 seconds. Peak VRAM was about 31 GB.
Worth knowing
It needs an NVIDIA GPU with 32 GB of VRAM (24 GB should work through offloading but is untested), 64 GB of RAM, ComfyUI 0.41.3 or newer and about 44 GB of open-weights models. There is no hosted version. From the command line a take is one line, for example seam make first.png last.png --preset smoke. It was built for Comfy's Open Source Higgsfield challenge; the code is MIT and LTX-2.5 has its own community licence.