Create Engaging Videos with OpenClaw
I showed content creators how an AI assistant can take a video request through generation, editing, and delivery, rather than stopping at a suggested prompt.
Tencent · Singapore · 27 March 2026
I Wanted the Clip, Not Another Prompt
Using a private beta connection to Tencent Cloud, I asked OpenClaw to create a stylized video from a photo reference. The work grew from individual clips into an assembled piece with sound, then into versions suited to different screen shapes. Finished videos came back through Telegram for review.
The talk shows how creation and revision fit together. A request for a square version reused the finished video, while a new facial reference required new footage before assembly. That distinction matters for a creator: some changes are edits to existing work, while others mean making the material again.
Taking a Video Request Through Production
Generation and Assembly Are Separate Operations
The configured tencent-mps skill connected OpenClaw to Tencent Cloud Media Processing Service for generation, audio, stitching, and subtitles. The portrait assembly downloaded and normalized opening, interlude, and closing clips before joining them with a consistent soundtrack. Another version used clip audio from Kling 3 Omni.
Reuse an Edit or Regenerate the Material
A square 1:1 crop reused the completed portrait artifact, preserving the existing clip rather than requesting a new interpretation. A new facial reference required fresh footage: the assistant generated the opening, middle, and closing segments sequentially to avoid concurrency issues, then stitched the completed outputs.
Wait for Outputs Before Delivery
OpenClaw tracked asynchronous cloud jobs and retrieved finished media before starting the next dependent edit. Signed download links allowed access to private outputs before Telegram delivery. The useful handoff was the assembled video returned for review, not merely confirmation that a generation request had been submitted.