Zero-Edit Daily Short Videos: Automate Hot-Topic Clips with WorkBuddy & MiniMax H3

Why Scheduled + Video

Three facts drive this pipeline:

Hot-topic lifecycles are measured in hours. By the time you manually select, write, and edit, the trend has passed. The only traffic you can capture is the same-day wave.

Short-video platforms favor vertical clips. WeChat Channels, Douyin, and Xiaohongshu give higher distribution weight to short vertical videos, but the production barrier is an order of magnitude higher than text-and-image.

The only solution: automate production, keep human decision and final publish. You don't need editing skills; you only need to approve "publish" when the trend hits.

Overall Architecture: Six Steps End-to-End

The workflow is broken into six discrete steps, each with a defined tool and output:

Scheduled Trigger — WorkBuddy automation uses an rrule to fire at three recommended-feed peaks daily: 09:30, 14:30, 19:30.

Fetch Hot Topics — WebSearch / WebFetch pull the day's trends and grade them by credibility (official confirmation / code evidence / community speculation).

Topic Decision — Cross-reference backend data model (decision-type buying guides ≫ new-product leaks > deployment tutorials > opinion pieces), pick the most viral candidate, output a short-video script.

Draw Frames — Python PIL hand-draws the first and last frames at 1080×1920; titles and keywords are vector-sharp.

Generate Video — Image-to-video (WorkBuddy's built-in video engine, backed by MiniMax H3-class models) produces 9:16 / 10s / 1080p clips.

Deliver & Publish — Export mp4, add platform BGM, upload to WeChat Channels / Douyin.

Key Trick: Draw Frames First, Then Generate

This is the highest-value step where most people fail. Direct text-to-video prompts like "tech-style title animation" produce Chinese characters as garbled blocks because diffusion video models learn pixel motion, not font structure.

The fix uses two layers:

Layer 1: Text via PIL. Python's PIL library renders the first and last frames at 1080×1920 — titles, keywords, and sign-off are all vector-drawn, crisp edges, zero garbling. This matches the official-account cover design language (warm-black background + vermilion), keeping brand consistent.

Layer 2: Motion via video model. Feed those two frames to image-to-video; the model only handles camera push-in, particle float, light sweeps — motion and atmosphere — while text stays perfectly legible throughout.

Trade-off: the look leans "title animation / info card" rather than cinematic live-action. For hot-topic explainer shorts this is fully sufficient — high information density, fast output, batchable.

Scheduling Configuration

WorkBuddy automation uses standard rrule, one line: FREQ=DAILY;BYHOUR=9,14,19;BYMINUTE=30 Three time points map to the morning, midday, and evening recommendation peaks. The task instruction specifies three actions: fetch today's hot topics → grade by source → output one 10-second vertical video → save to designated directory. The rest runs automatically.

Fetching Hot Topics Without Flipping

Reusing the Qwen4 article experience: grade fetched topics first; video scripts only trust verifiable parts; rumors must be labeled with source. If the script presents "community countdown page guesses Nov 1" as "officially announced", the video gets debunked the same day — traffic turns into a PR disaster. Grading awareness is the human judgment most worth preserving in the automation era.

Video Parameter Cheat Sheet (Copy Directly)

Should You Adopt This Stack?

Good fit: daily / twice-daily self-media, hot-topic explainer channels, accounts with a fixed topic matrix (e.g., long-term focus on Mac + LLMs). Automation value: "when the trend arrives, you're already on the production line."

Bad fit: channels requiring live-action on-camera / talking-head, or those demanding cinematic visual quality (image-to-video yields animation style, not film-grade footage).

Cost: mainly video-generation credits; PIL drawing and scheduled automation are free.

Closing

Automation never replaces you; it frees you from repetitive labor: when the trend hits, you only decide whether to publish.

Reference stack: WorkBuddy Automation (scheduling) / VideoGen image-to-video (MiniMax H3 etc. as backend models) / Python PIL (frame drawing); case study: Qwen4 pre-launch 10-second vertical video.