AI short dramas: the bottleneck isn't generation, it's the pile of files afterward

Ladies, let me start with something counterintuitive. Right now everyone talking about AI short dramas is talking about models, visuals, how to make a film in one sentence. After two months of tinkering with Jimeng AI and Kling AI, what actually blocks me was never generation, it's the pile of files that don't match up after generation.

There's a comparison in the material that stuck with me. In the first 8 months of this year 430,000 micro short dramas went live, AI dramas over ninety percent, the pie is genuinely huge. But on the other side:

Per-episode cost compressed from tens of thousands to a few hundred, cycle from one or two weeks to one day, the rate of breaking 100 million views is about 0.47%, the proportion that recoup costs is said to be 1.3%

Those two numbers side by side are pretty jarring. There's also that case in Jiang'an District, Wuhan, the province's first AI micro short drama copyright infringement, 47 episodes, one hour total, sued just because the source of materials wasn't sorted out.

I can't do a pie that big. Little Bao is two, during the day I only have the two hours of his nap, and after I coax him to sleep at night I can squeeze out another hour or so. A one-minute-plus period-style short, I need four or five sessions to finish. The problem is exactly those four or five sessions.

Once a film is generated, the folder is messier than my brain

Jimeng AI delivers per clip, same with Kling AI. A one-minute finished film behind it is thirty-some video clips, seven or eight voiceover versions, a pile of prompt screenshots, plus the cover and info sheet needed for platform filing. Filenames are all a string of numbers the system gave, I can't even be bothered to rename them. The baby cries and I have to drop everything, come back and sit down, forget which is which.

Before I relied on my memory, by the third day I couldn't remember.

Later I threw all of it at WorkBuddy. I've used it about three months, before mainly to organize that pile of side-hustle courses in my Taobao favorites, wrote a piece about it too, this time a different use. A so-called ledger, plainly put, is just a checklist you can look things up in.

Hands-on: three steps to build a short drama material ledger

My setup is a Windows laptop, WorkBuddy desktop version, may not apply to everyone.

• Create folders: first layer locally, four levels project name/scene/material type/finished film, scenes numbered S01, S02, don't write "Act One" in Chinese, sorting gets messed up later
• Upload: open WorkBuddy, click New Task on the left, in task type choose File Organization (in my version it's under the batch processing column, the middle of the interface is a dashed box), drag the whole project folder in, it scans first, the right side shows the number and types of files recognized
• Write rules: in the dialog box below write the requirements, don't use jargon, use plain language

The first time I dragged in 62 files, it scanned for a dozen seconds. The rule I wrote was: extract each video file's filename, duration, resolution, sort by the scene number in the filename, output a CSV table, columns in order scene number, filename, duration, resolution, usage note, leave the note column blank for me to fill. CSV is a comma-separated table file, Excel opens it directly.

Click Start, there's a progress bar on the right, when done click Export Table, a file appears on the desktop.

The effect is like this. Before, finding a clip took five minutes of digging, now open the table and search the scene number, two seconds. Sunday night after the baby sleeps I run an incremental pass, throw in the newly generated ones, done in ten minutes.

Pitfalls I hit and a bit of looking ahead

The biggest pitfall is saying too much at once. At first I wanted to say organize, batch rename, and generate cover copy in one go, result: the table columns all misaligned. Later I learned my lesson, one thing at a time, organize first, then rename, two passes is actually faster.

One more detail, uniformly prefix filenames with the date, 20261028_S01_xx, then it won't scramble when sorting. Also run it every Sunday night, don't let it pile up, pile up to month-end and that one run takes twenty minutes, baby wakes up and it's ruined.

On permissions, I don't have a team on my side. I set the exported table to read-only and shared it with my husband, he can check progress anytime, can't edit. If I collaborate with my girlfriends later, I plan to give read-only too, edits I'll do myself.

Looking ahead, I think what AI short dramas compete on next isn't who generates faster. Douyin already opened a pre-check tool for copyright holders and directors in August, Hongguo is also cracking down on homogenization and material violations. Compliance will become a hard threshold, whoever has clear project files can keep going. Those films piled up from a bunch of garbled files, one incident is enough to sink you.

My ledger isn't fancy, but for someone like me with only three hours a day, only what you can manage counts as an asset, what you can't manage is all a burden.