Fal

fal Recast is a one-click character replacement tool for video from fal, the generative media platform that serves image, video, and audio model APIs to developers. You give it a source clip and a reference photo of the new character, and it swaps the person on screen while keeping the original motion, camera moves, cut timing, and soundtrack intact. The tool is aimed at developers and creative teams who want to produce character variations of an existing shot without reshooting or rebuilding the edit. Under the hood, Recast runs on H3 Max, fal's post-trained variant of MiniMax H3 . According to the model page, it replaces the main people in a source video using reference photos, tracks each person across shots, and accepts an optional prompt to control who becomes whom or which details to keep. It supports up to four reference photos and videos up to 30 seconds, with individual shots up to 15 seconds, at 768p or 1080p. The launch matters because character swapping has typically required stitching together separate segmentation, tracking, and generation steps. Shipping it as a single endpoint on fal fits the company's broader positioning as an inference layer for generative media, built by co-founders Burkay Gur and Gorkem Yurtseven , and gives application builders a direct way to add character variation, localization, and ad personalization to pipelines that already run on fal.