wtf is clipforge?

ClipForge is a web application built for marketers and small creative teams who need short promotional videos without learning a full video editor. A user describes a project in plain language, breaks it up into scenes, and the platform generates an eight second AI video clip for each scene, then merges the ones you like into a single downloadable video file. The tool lets you mix and match three different AI video generation providers, Google's Gemini Veo Pro, Veo Flash, and MiniMax's Hailuo, choosing per scene which one to use. This means a team can use a cheaper, faster option for early drafts and save the more expensive provider for the one shot that really needs to look polished. Each scene can also include a reference image to guide the video, a toggle for whether the clip should have synchronized audio, and a choice of transition style like a hard cut, fade, or crossfade between scenes. Because these AI video providers charge per generation, the platform enforces daily limits per model, refunds your quota if a generation fails, and shows how many attempts you have left before you use one. Every attempt at generating a scene is kept in a history so different takes can be compared and one selected as the final version. Once scenes are ready, a background worker uses FFmpeg to normalize the clips to a consistent resolution and format, apply the chosen transitions, mix in background music if you upload one, and produce the final merged video, keeping a record of every past export. On the administration side, the first time it starts up ClipForge automatically creates an admin account, and afterward supports role based logins for admins, editors, and viewers, along with an audit log of every generation, merge, and admin action. It also ships with monitoring tools, Prometheus, Grafana, and Loki, already wired together in its Docker setup, plus an analytics page showing generation counts, estimated spending, and which scenes get redone most often. The project is built with React, Express, and PostgreSQL, uses Redis backed background jobs, and stores files in S3 compatible object storage. It can be started locally in about five minutes using Docker Compose, provided you supply your own API keys for the video generation providers. It is released under the AGPL version 3 license.