ShipToLive · Learning Log
Know it works before you film
How I pick short-video topics now: 對標選題. Find what an audience has already proven it wants, then make your version of it. Making the video is the expensive part; checking demand takes fifteen minutes. Further down: the other three secrets for format, business model, and AI.
The idea in one line
對標 means "benchmark against." Instead of guessing what might go viral, look for topics that have already performed well, more than once and for more than one creator, and make a better or more personal take on them.
Not copying. You borrow the topic and the question it answers, not someone's script, footage, or edit. The underlying logic is the same; your angle, voice, and examples are your own.
A test I ran: two supermarkets
Same format, a supermarket haul / 開箱 video. Which one gets more views: a 全聯 haul or a city'super haul? I guessed city'super, since it's more novel. Then I searched YouTube.
全聯開箱
- Next results: 63萬 and 55萬
- ~1,260 stores in Taiwan
- Over 11 million app members
city'super 開箱
- Next result: about 1,100
- About 10 stores in Taiwan
- Premium, imported groceries
The lesson: audience size wins for reach. Almost everyone in Taiwan shops at 全聯, so almost everyone can relate to the video. Novelty doesn't beat relevance.
But views aren't revenue. A smaller, wealthier audience (the city'super crowd) can be worth more if you're selling something premium or looking for leads. Pick by goal: big audience for reach and followers, niche audience for high-value customers.
Six places to find proven topics
Try it on your own product: the Topic Finder pulls live YouTube and Google searches for what you sell, opens all six places at once, and scores your options.
YouTube search
The fastest demand check. Type the topic and look at view counts.
Sort by view count. Compare views to the channel's subscriber count: 1M views on a 50K-subscriber channel means the topic carried the video. Check the upload dates, and search the Shorts tab separately.
Google search
Shows what people actually ask, in their own words.
Read the autocomplete suggestions and the "People also ask" box. Each question there is a candidate video title.
Keyword planning tools 關鍵字規劃工具
Puts numbers on demand.
Google Ads Keyword Planner (free) gives monthly search volume. Compare two phrasings of the same topic and use the bigger one in your title.
抖音、小紅書
An early look at what's trending.
Formats and topics that are hot there often show up in Taiwan's Reels and Shorts later. Look for repeated patterns across many creators, not one lucky video.
Customer FAQ 客戶 FAQ
Questions your clients ask you again and again.
If three clients asked it, thousands of strangers are searching it. These videos also bring in the right kind of viewer: future customers.
Comments 留言選題
Viewers tell you the next video.
Read the comments on popular videos in your niche, and your own. Questions, complaints, and "can you do X?" requests are free topic research.
The 15-minute topic check
Run this before spending credits or filming time. Tick what's true; three or more ticks means film it.
0 of 6 ticked.
How I'm using this
In my 30-day AI video plan, every episode now starts with this check. Before writing a prompt, I compare phrasings on YouTube and keep the one with proven demand:
| Option A | Option B | What I'm checking |
|---|---|---|
AI 影片 教學 | AI 做短影音 賺錢 | "How-to" vs "make money" framing |
Veo 3 教學 | Kling 教學 | Which tool people are searching for |
AI 短劇 | AI 動畫 故事 | Format demand for the episode series |
Rule for the rest of the month: no topic gets filmed until it passes the check. The 80 minutes of "make" time each day goes to topics with evidence behind them.
Topic is secret one of four
Picking the topic is the first step. The same 對標 logic (benchmark against what already works) applies to the whole process:
| # | Benchmark the… | In one line |
|---|---|---|
| 1 | 選題 topic | Know it will work before you film. |
| 2 | 形式 format | Create the simplest way: 口播. |
| 3 | 商模 business model | Turn views into revenue, and keep the people. |
| 4 | AI | Let AI do the hard parts. |
Secret 2: copy the cheapest proven format
Another A/B test: an on-location vlog (travel, several people, lots of shots) versus one person talking to a phone at night with a strong opening line. The talking-head video got about 750K views, at a fraction of the effort.
For most people, the best format is 口播, talking straight to camera. The value is in what you say, so you can make it with a phone in minutes, and you can make it every week. A repeatable format beats a one-off hit.
- Fancy filming techniques
- Advanced editing skills
- Lots of time per video
- A hired strategist or ghost operator
- A proven topic (Secret 1)
- A hook in the first line
- A phone and a teleprompter
- To start small and keep going
萬事起頭難，起頭要簡單。 Starting is the hard part, so make the start easy. My first video: 30–60 seconds, phone only, auto-subtitles, no AI clips. Publish it, don't perfect it.
Apps: CapCut (剪映 is the mainland version) for subtitles and quick cuts, and 開拍, which is built for talking-head videos with a teleprompter.
Secret 3: copy a proven business model
Don't invent a way to make money; find people like you who already do and copy the structure. Views alone pay little. The money comes after the video:
- Video answers a real question and earns attention.
- Free resource in the bio or comments: a checklist, a tool, a page like this one.
- Own the relationship (把人留著): a LINE OA or email list, so you aren't renting your audience from the algorithm.
- Paid step: a consultation, a product, or a service.
To 對標 a business model, pick two or three creators in your field who sell through short video. Look at their bio link, what they give away free, and where the paid offer appears.
Secret 4: let AI interview you
The best demo of the night was a custom GPT that writes story scripts. It doesn't write right away. It interviews you first, so the script is built from your real experience instead of generic AI text. The method:
- Pick a script type: a story, a teaching video, or a personal-brand plan.
- Answer 8 questions by voice. Talk like you're telling a friend; messy is fine. They map to the arc: where you are now → setback → turning point → second setback → second turning point → lesson.
- Answer the follow-ups. It checks what's thin and asks for four details: a concrete scene, the exact inner voice sentence, the lowest point, and one contrast number.
- Get the one-line arc, five beats. If it doesn't fit on one line, it won't fit in 90 seconds.
- Get the script: one short line per breath, strongest moment first. Read it straight off the teleprompter.
What makes the scripts work: a real quote from the low point, a fake-out where the comeback fails once more, hard numbers instead of "I'm successful," and an ending that turns to the viewer.