Use agents to draft and render short social videos

A creator used OpenClaw agents to plan, script, voice, and render WeChat videos, then approved each one in Telegram before publishing.

Done with OpenClaw· Intermediate· Oct 2026

What they did
They set up a cron job so one agent sent five topic ideas to Telegram every morning. After they chose a topic, another agent wrote a short narration of about 200 characters, made cloned-voice TTS through MiniMax on fal.ai, pulled timestamps with local Whisper, filled a scenes file, and rendered the video in Remotion. The creator checked the draft in Telegram, gave the go-ahead, and a separate agent posted it.
How it went
Each video took about 15 minutes end to end, with around 10 minutes running without supervision, and cost about ¥0.1. One video got 1,595 views, and the first three days brought 9,018 views total. Before this workflow, their videos usually stayed below 200 views.
Tips
Keep the writing model strong; when they switched from Opus to Sonnet to save money, the scripts got noticeably worse and they switched back within 6 minutes. Also, Whisper timestamps can drift, so review the render before approving anything final.

Try it yourself with OpenClaw

Act as my short-video production assistant for [platform]. Each morning, send me 5 topic ideas for [niche] in Telegram. After I choose one, draft a [length] narration in my voice, create voiceover with [TTS or cloned-voice setup], and assemble a video in Remotion using [visual style]. Show me the draft in Telegram and stop before any publishing step until I explicitly approve it. If I ask for changes, revise the script and render again before publishing. After I approve, publish the finished video to [platform].

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on cloud.tencent.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.