An AI Short Drama for ~$400? Full Pipeline Cost Breakdown

One news story has been flooding creator feeds: Yunnan-based photographer Liu Ziyu produced the AI short drama Zombie Cleaner for about 3,000 yuan (~$420) — and returned more than 2,000x (Sina News, Oct 2026). Three thousand yuan is roughly what a big-city film crew spends on lunch. Now it buys an entire AI short drama.

Most people see a get-rich fairy tale. People in the business should see something else: AI has rewritten the cost structure of short drama production — the real opportunity isn’t “daring to spend 3,000 yuan,” it’s “knowing exactly where that 3,000 yuan should go.”

Especially one fact almost everyone overlooks: in a full AI pipeline, the cost of the subtitle and copy stage has dropped to near zero — and that stage used to be the most labor-hungry part of translation and localization. This piece uses one cost table to take apart every link in the chain, from zero to launch.

I. The Case First: What Does 3,000 Yuan Actually Mean?

Liu Ziyu is a photographer by trade. Zombie Cleaner is a short drama produced with AI tools at a cost of about 3,000 yuan, returning over 2,000x (Sina News, Oct 2026).

Before we dissect it, two cold showers:

1. This is an outlier, not an industry average. Viral-hit return multiples always belong to the tail of the distribution. Budget against them as if they were expected values and you’ll die ugly.
2. Low cost ≠ low barrier. His core assets are visual taste and storytelling ability. AI only crushed the execution cost — judgment is still scarce.

But the case’s lesson is real: once tool costs collapse, what decides success shifts from “budget size” to “pipeline efficiency.” Spend in the wrong stage and 3,000 yuan evaporates. Spend in the right ones and 3,000 yuan is a seed round.

II. Full Pipeline Cost Table (Sample Estimate)

Disclaimer: the table below is a sample estimate, meant to illustrate cost structure — not a quote. Actual costs vary wildly with individual skill, tool subscriptions, and market choice. Use it only as a budgeting frame.

A few notes:

• Subtitles & copy ≈ 0 is the heart of this table. The Flownib Subtitle Extractor is free, needs no signup, and has no usage caps — pull captions from 40+ sites anytime, export SRT/VTT/TXT at will. Rewriting is AI-drafted and human-proofread; the marginal cost is just your time.
• Localization is not a from-zero spend. Draft small languages with YouTube auto-translated captions (150+ languages), then hand-polish the markets that validate. This turns translation from “a fixed per-title expense” into “follow-on investment behind viral hits.”
• Ad spend is a variable, not a constant. For genres that validate on organic traffic, bank that money for the next title.

Run the numbers on this sample table and the cash outlay for a single-market AI short drama lands in the three-to-low-four figures — right in line with Liu Ziyu’s 3,000 yuan.

III. Which Stages Really Approach Zero Cost

1. Subtitle extraction: from manual timing to a 10-second draft

Timing and translating a 2-minute video used to take one to two hours. Now you paste a link and hold the text draft in 10 seconds. Free, nothing stored, fetched on demand — process hundreds a day at no cost. This is the highest-certainty “cost-to-zero” stage in the entire pipeline.

2. Copy rewriting: AI drafts + human gatekeeping

Hook rewrites, dialogue localization, multilingual first drafts — AI can draft all of it. Humans keep only two jobs: proofreading proper nouns and judging cultural context. That’s taste and judgment work, with zero cash outlay.

3. Distribution copy: one draft, many meals

The script draft spins off subtitles, descriptions, posts, and image-text companions at near-zero marginal cost. And sites that mix image-text with video see roughly 3x the AI citation rate of text-only publishers (Digital Arka, Aug 2026) — a combo that costs nothing has no excuse not to be thrown.

IV. Where to Save — and Where Not To

Save on:

• Every “execution” expense: caption timing, first-pass translation, rough cuts, format conversion. AI is already faster and cheaper than people here.
• Early multilingual coverage: blanket small languages with auto-translated drafts, and double down only after validation.

Don’t save on:

• The story itself. The script is the source of everything. Zombie Cleaner broke out on its premise and narrative, not its tools. AI writes serviceable lines — the judgment of what makes a good story is still yours to supply.
• Localization quality in priority markets. Paid conversion happens at the moment of cultural fit. Machine-translated phrasing is a conversion killer.
• Copyright and compliance. Asset licenses, music licenses, portrait rights — spend what must be spent. Tool capabilities like subtitle extraction are for compliant use only: your own content, freely; other people’s content, analysis and learning only; commercial adaptation requires authorization.

V. Don’t Let a 2,000x Return Go to Your Head

Last, the risk talk. The same report carries another number worth remembering: in the first four months of 2026, over 8,000 new short drama companies registered in China, while revenue shares fell sharply (Sina News, Oct 2026).

That tells us two things:

1. Low cost is the whole industry’s entry ticket, not anyone’s moat. When everyone can make a drama for 3,000 yuan, competition snaps back to premise, taste, and operations.
2. A supply surge crushes returns on mediocre content. Costs fell; success rates didn’t. The denominator of the hit rate got bigger — the numerator didn’t.

So the correct posture is: use low cost to run many test rounds, not to bet everything on one title breaking even. Three dramas testing three genres beats one drama at triple the spend. Whichever one breaks out, then add localization and ad budgets. What the full AI pipeline truly changes isn’t “what a drama costs” — it’s “how many times you can afford to be wrong.”

Frequently Asked Questions

Q: Can 3,000 yuan really produce a shippable short drama? Yes — depending on your bar for “shippable.” Liu Ziyu’s Zombie Cleaner at roughly 3,000 yuan with a 2,000x-plus return (Sina News, Oct 2026) has proven the path exists. But an outlier is not an average: budget in your own time and learning costs too.

Q: Can the subtitle and copy stage really be zero cost? At the tool level, yes. Flownib Subtitle Extraction is free with no usage caps and exports SRT/VTT/TXT; AI-drafted rewrites don’t generate fees either. What remains is your proofreading and judgment time — the best-value investment in the whole pipeline.

Q: Does localization translation cost money? Do it in stages. During testing, draft with YouTube auto-translated captions (150+ languages) at zero cost; hand-polish the markets that prove out. Reserving the translation budget for validated viral hits is the cheapest playbook there is.

Q: Are the numbers in the cost table real quotes? No. It’s a sample estimate illustrating each stage’s cost structure and magnitude. Real numbers vary hugely by person, tools, and market. Recalculate against your own situation when budgeting.

Q: What compliance issues matter for AI short drama exports? Three: licenses for assets and music, platform disclosure rules for AI-generated content, and copyright boundaries around other people’s content. Tools like subtitle extraction cover your own content or analysis/learning use cases only; commercial adaptation of others’ work requires authorization.

Conclusion

The thing worth copying from Liu Ziyu’s case isn’t the number “3,000 yuan” — it’s his cost allocation. Money goes to premise and visuals; copy, subtitles, and multilingual versions go to the AI pipeline that’s already approaching zero cost. In an era of collapsed tool prices, efficiency is capital.

So the first thing you do after reading this isn’t opening an AI image generator. It’s finding a benchmark short drama and extracting its complete script. Study your competitor’s cost structure — starting with the step that costs nothing.

3,000 yuan won’t buy you a viral hit — but it will buy you 10 attempts. That’s the real budget sheet of the AI era.

Try it now: Paste any video link and get subtitles in 10 seconds → Extract Subtitles