Private AI media workspace
FrameFlow generates short video and images with xAI, accepts your own uploads, stitches clips together, and publishes the result to your connected TikTok, YouTube, and X accounts — through each platform’s official API, on a server you control.
FrameFlow is a single-operator workspace with no public sign-up. The studio is private; everything that explains it is on this site.
The pipeline
Every stage lives in the same studio, and every attempt is recorded in SQLite on your own server — nothing is fired off and forgotten.
Write a prompt and generate a clip or still with xAI, optionally anchoring it to a source image so the shot starts where you want. Or upload MP4, WebM, MOV, JPEG, PNG, WebP, or GIF straight from disk.
Stitch short generations into a longer sequence with ffmpeg, extract a last frame to continue from, and play everything back in the browser before it goes anywhere. Captions and titles are written in the same view.
Pick destinations, set each platform's own options, then publish immediately or park the post until a chosen time. A background worker runs the upload, so a large file never races an HTTP timeout.
Destinations
Each connection uses credentials you issue for your own account. FrameFlow does not scrape cookies, automate passwords, or route around platform approval.
Native Content Posting API with PKCE authorization. Creator settings are read before every post, videos upload in chunks, and status is polled until the post completes.
Official resumable upload through the YouTube Data API, defaulting to private visibility so nothing becomes public before you have looked at it.
OAuth 1.0a request signing, chunked media upload with processing polls, and post creation on the v2 API — for video, image, and text posts alike.
How it behaves
Publishing to someone's real audience is not a place for surprises, so the defaults are the cautious ones.
Clips made by a model are labelled as AI-generated when they are posted, because that is what the platforms require of them.
Every provider credential is encrypted in the local database and is never returned to the browser or written to client storage.
A publication is only ever created by pressing Publish in the form. Failed posts wait for you to retry them rather than re-firing on their own.
A scheduled post sits in the queue until it is due, and can be cancelled or removed at any point before it runs.
Generated and uploaded files live in a data directory on your own server, served only to an authenticated session.
There is no FrameFlow cloud. The app runs on hardware you administer, against accounts you already own.
Questions
The support page covers access, data removal, and revoking a connected account — or write to zhangyang8605@gmail.com.