Plan U: AI Short-Video Matrix

Topic library → batch scripts → batch videos → multi-platform distribution → data review

Time
1-2 days / batch
Cost
Medium
Difficulty
Intermediate

Best for

For running several accounts at once on Douyin, Xiaohongshu and Bilibili with one content factory that produces in volume.

Steps

Build a topic library

30-45 min

ChatGPT / Kimi + Perplexity

Have the model generate dozens of niche topics in batch and tag each one (pain-point, trend, tutorial, persona), then check real heat with Perplexity against platform trends and search data. Save the survivors into a ranked spreadsheet.

💡Rank topics by expected click-through and only make the top 20% — don't spread your effort thin.

Batch script generation

30-60 min

ChatGPT / Kimi

Generate voice-over scripts from the topic library with one template: 3-second hook, body, call to action. Tweak tone per platform — fast and punchy for Douyin, deeper for Bilibili, aspirational for Xiaohongshu.

💡Ask for 5 hook variations per script and keep the most conflict-driven one.

Batch video production

1-2 hours

可灵 Kling + PixVerse

Produce visuals or avatar voice-overs in batches with fixed characters, backgrounds and caption styles for a consistent series look. Default to one master version that can be reused everywhere, then make platform-specific cuts.

💡Perfect one full pipeline before going to volume — ten failures at once help no one.

Multi-platform distribution

30 min

Platforms + scheduler

Export adapted versions per platform following best posting times, aspect ratio, hashtags and covers; generate A/B titles and covers with the model, then finalize manually.

💡Change the title and first 3 seconds for each platform — the same video verbatim is the easiest way to get ignored.

Data review

30 min

Platform analytics + ChatGPT

Pull views, completion rate, likes and follower growth; feed the numbers to the model and let it tell you which topics, hooks and covers worked, then feed that back into the library and templates.

💡Review on a fixed weekly cadence and turn winners into reusable formulas.

Tools involved

Pitfalls to avoid

One-size-fits-all distribution

Platforms differ in style, length and copy; adapt each asset per platform.

Volume before validation

Scaling before one flow works means scaling the failures.

No feedback loop

Posting without reading data keeps the topic library stuck at guesswork.

Matrix means more accounts

A real matrix is a hero account with coordinated satellites, not a pile of random accounts.

Related Workflows