FORMWEFT Motion Workbench

The Chinese-language editor exposes SRT and local-video import, six card styles, a timeline, and landscape or portrait framing. The six styles are an opening title, spoken-line captions, a key point, a data highlight, a quote and a step list. Each card sits on the timeline with its own start and end time.

Since September, the page has grown into a four-step flow. You draft a script from a template or from an AI service you've set up yourself, add your own video, voiceover and music, edit and proofread the captions, then check the range, sound and cover before export. If you already have footage, you can go straight to editing. The tool says it has no AI model built in.

On the canvas you drag captions and cards into place with a mouse or by touch, and each drag can be undone. Background color, opacity, corner radius and text color are adjustable per card, and keyboard shortcuts cover play, stepping by a tenth of a second, and nudging a card by one or ten pixels.

The export panel describes combining video, captions, and original audio on the device. It asks users to keep the page visible during real-time recording and inspect the downloaded result.

No file was imported, so subtitle timing, original audio, export formats, and the claim that processing stays on the device were not tested.

The written instructions spell out the limits. Because export records in real time, it takes about as long as the clip, with a cap of 10 minutes or 256 MiB. It produces MP4 where the browser supports it and WebM otherwise, and there's no MOV option. PNG and image-sequence exports hold only the transparent motion layer, without the video or sound. An SRT file can have up to 200 entries of 800 characters each and run up to an hour. Voiceover and music files can be up to 32 MiB or 10 minutes, and a voiceover lines up from the first second with no automatic lip sync.

Saved project files leave out the video, the audio and the volume settings, so reopening a project means choosing those files again. Refreshing or closing the page clears the finished preview, so download the result before you leave.

Creators who already have footage and subtitle text can work on the caption treatment without starting in a full video editor. The visible timing and frame controls give this a concrete use beyond generating a caption from a prompt.

The real-time export makes long videos slow to finish, so it's best suited to short talking-head clips. The code is MIT-licensed, an offline package is offered beside the online editor, and the watermark is off by default.