Minimum scope for an original story: single scene, few characters, 30–60 seconds
This page tackles the first hurdle of AI short drama topic selection: locking down the scope. Cell 1 of the workflow card, "Scope locked: single scene, few characters, 30–60 seconds," gets checked off right here—before writing the synopsis or touching any tools, use the three hard criteria to cut your story down to a size your first film can truly be finished at. This page involves no software installation and no spending, and it produces just one thing: a five-line topic scope statement.
What you'll be able to do after this lesson
- Use the three hard criteria (1 scene, on-screen characters ≤2, 30–60 seconds) to check any AI short drama topic and judge whether it's suitable as your first practice film
- Use the three methods—cutting time, cutting space, and cutting characters—to narrow an oversized story idea down to the minimum shootable scope (the teaching template of this course)
- Write a five-line topic scope statement that locks down scene, characters, duration, shot count, and aspect ratio all at once
- Explain clearly that 30–60 seconds is this course's practice design goal: the finished film is edited together from multiple short shots, not generated by a model in one go
Where you are right now
- "I have an epic in my head: three factions, five timelines—and I haven't even generated the first shot"
- "The topic keeps growing grander the more I think about it; I've spent a week revising the worldbuilding and still haven't produced one complete film"
- "I did finish writing the story, but it has a dozen-plus scenes and seven or eight characters, and I don't know which image to prepare first"
When to use this technique
When you need to:
- Pick, from a pile of story ideas, an AI short drama topic you can actually finish as your first piece
- Lock down the scope before writing the three-sentence synopsis, to avoid scope creep as you go
And you don't want to:
- Take on multiple scenes and multiple characters in your very first film and end up abandoning it halfway
- Spend your time building a world instead of producing one complete original short video
🎒 Preparation before you start
- Have read Full workflow overview: break the story into shots first, then assemble the film, and know the five stages and the estimation method of "about 3–5 seconds per shot, roughly 6–20 shots for the whole film"
- A pen or a blank document: this page's output is the five-line topic scope statement
- You can have a story idea of any size in your head—this page is exactly about teaching you to cut it down
Core idea
The goal of your first practice film is to "finish it," not to "make it big."
"Single scene, few characters, 30–60 seconds" is the instructional design this course sets for practice films, not a capability parameter of any model. Each of the three criteria governs one thing:
| Criterion | What it controls | Why it's set this way |
|---|---|---|
| Single scene | Asset workload | You only need to prepare 1 scene reference image (see Scene reference images: set the location before rolling), with no need to rebuild the environment between shots |
| Few characters (on-screen ≤2) | Consistency difficulty | Every on-screen character needs a character card and reference images (see Character card: lock down looks, costume, and personality); the more characters, the harder it is to keep their looks consistent across shots—this course does not promise absolute character consistency |
| 30–60 seconds | Total workload | At 3–5 seconds per shot (the course's design value), that works out to roughly 6–20 shots—workload you can estimate and review |
Why can't you count on generating 30–60 seconds in one go? Two grounds:
- A model's single output is inherently a clip of a few seconds: for the open-source model Wan2.2's TI2V-5B, the official documentation states it generates 5-second 720p video (Wan2.2 official model card)
- The official production workflow is also described as "generate multiple short segments, then edit and combine them" (Runway official help documentation)
So setting the topic at 60 seconds refers to the finished film's duration after editing.
| ❌ Going big on your very first film | ✅ Minimum-scope practice film | |
|---|---|---|
| Scenes | 5+ locations, each needing reference images | 1 location, 1 scene image |
| Characters | 5+ faces, maintaining consistency shot by shot | 1–2 face(s) |
| Duration | 3 minutes minimum | 30–60 seconds |
| Shot count | Can't be estimated | 6–20 shots, calculable and checkable |
| Most likely outcome | Abandoned halfway | Finished, and you know what to practice next |
Aspect ratio doesn't change the scope: both vertical short video and landscape can accommodate a 30–60-second practice film. The tooling also supports vertical—for example, Wan2.2 TI2V-5B's 720p resolution, officially documented as available in landscape 1280*704 or portrait 704*1280 at 24 FPS. Pick one aspect ratio in the scope statement first, and finalize it in the shot lessons.
Follow along
Step 1: write down your AI short drama topic (one-sentence version)
Why: A topic always seems small in your head; only by writing it down can you see how big it really is.
Write the story you want to shoot into your document in one sentence. Here's a starting format (a teaching example for this course):
My topic (one sentence):
Who + where + what they want to do + how it ends
Example: Under a pedestrian bridge on a stormy day, a delivery rider tries to protect a birthday cake for his daughter.You should see: A one-sentence topic is now in the document. It may be too big right now—that's fine; that's exactly what the next two steps deal with.
- Mini check: the one-sentence topic is on paper, not still stuck in your head
Step 2: Measure the topic against three hard metrics
Why: Scope isn't narrowed by feel. Measure it first, then you'll know how much to cut.
Answer "yes/no" to each item in the checklist (a teaching template for this course):
Topic scope checklist
□ Setting: the entire story takes place in 1 location(s)?
□ Characters: on-screen characters ≤ 2?
□ Duration: can be told in 30–60 seconds (about 6–20 shots)?All three "yes" → skip straight to Step 4; any "no" → go to Step 3 to cut the topic down.
You should see: All three checklist items have clear answers, and it's obvious at a glance which one fails.
- Mini check: all three metrics measured, and the failing one is marked
Step 3: Cut the topic down—three ways to shrink it
Why: An oversized topic is the number-one reason a first film dies halfway. Cutting a topic has fixed techniques, so you don't have to go by feel (the methods are teaching designs for this course).
Try them in order, cutting until all three metrics pass:
| Technique | How to cut | Example |
|---|---|---|
| Cut time | Keep only the most tense segment; cut from "a whole life" to "a single moment" | "A rider's day" → "the last 60 seconds of the rider's final delivery" |
| Cut space | Move all plot points to a single location | "Racing across the whole city" → "a bench under the pedestrian bridge" |
| Cut characters | Keep only 1–2 on-screen characters; turn the rest into voice-overs, back views, or cut them entirely | "Rider, daughter, shop owner, colleague" → "rider + daughter on the phone" |
After cutting, go back to Step 2 and re-run the checklist until all three pass.
You should see: Next to the one-sentence topic there is now a shrunken version, and all three hard metrics pass.
- Mini check: the shrunken topic re-passes the checklist, all items passing
Step 4: Write down a five-line topic scope statement, locked in
Why: If the scope isn't locked in, things quietly creep back in when you write the script and generate shots. These five lines are the "contract" for every lesson that follows: tools, budget, script, and storyboard all defer to it.
Copy the shrunken topic into a scope statement (a teaching template for this course; the example is a teaching example):
My topic scope statement
Topic in one sentence: Under a pedestrian bridge on a stormy day, a delivery rider tries to protect a birthday cake for his daughter.
Setting: under a city pedestrian bridge (1)
Characters: the rider (on-screen), the daughter (phone voice-over only, never on-screen)
Target duration: 30–60 seconds, about 6–20 shots (3–5 seconds per shot, a course design value)
Aspect ratio: vertical 9:16 (or horizontal 16:9, pick one first)You should see: The document contains the five-line scope statement. From the next lesson on, tool choices, budget, script, and storyboard all defer to it; if you want to add something, come back and ask first, "Does it still pass?"
- Mini check: the scope statement is locked in, and the shot count has a concrete range filled in
Checkpoint ✅
- A one-sentence AI short-drama topic is written into the document
- The topic passes the three hard metrics: 1 setting, on-screen characters ≤2, 30–60 seconds
- Used the three methods—cut time, cut space, cut characters—to trim an oversized topic down until it passes
- The five-line topic scope statement is locked in, with the shot count between 6–20
- Can explain that 30–60 seconds refers to the final edited runtime, with the whole film assembled from multiple short shots, not generated by the model in one go
Pitfall warnings
❌ Trying to shoot an "epic" as your first film
Symptom: three factions, five timelines, or opening with "I want to make a 5-minute film."
Cause: every location needs a scene reference image, every on-screen character needs a character sheet and reference image, and every shot must be generated one by one, possibly with retries (course budget: at most 3 per shot, see Shot retry and replacement strategy: at most 3 per shot). Double the scope and the workload far more than doubles; the more on-screen characters, the harder it is to maintain cross-shot consistency.
Correct approach: for your first film, follow the three hard metrics. Log the big story you cut into a "second-film idea list" and bring it out at the graduation review (see Graduation exercise: review this film, plan the next one).
❌ Treating "30–60 seconds" as the model's single-generation capability
Symptom: assuming that once the topic is set at 60 seconds, the model can spit out a 60-second finished film in one go.
Cause: a video model's single output is a clip of a few seconds (Wan2.2 TI2V-5B official notes: 5 seconds at 720P); the official production workflow is also described as generating multiple short clips first, then editing them together.
Correct approach: record the 30–60 seconds as "the target runtime of the final edited film," break it into 6–20 shots at 3–5 seconds each and generate them one by one; for how to break it down, see Storyboard table: breaking the script into a shot list.
❌ The topic keeps growing
Symptom: while writing the script, you "casually" add a second location or a third on-screen character.
Cause: the scope was never locked in, so there's no reference point when adding things.
Correct approach: every time you want to add something, go back to the scope statement and ask, "Do the settings, characters, and shot count still pass?" If not, log it in the second-film idea list—don't cram it into this film.
Advanced tips (optional)
30–60 seconds can't fit a complex plot, but it can fit one twist—that's the most economical shape in short-video story structure: build an expectation up front, then break it in the final seconds (a teaching suggestion for this course). When you can't think of a topic, run "one location + two characters + one small event" and keep the one with a twist. For how to expand the scope statement into a full story, see Story synopsis: tell it in three sentences for 60 seconds.
Lesson summary
✅ At this point, you have:
- Narrowed your AI short-drama topic to a single setting, on-screen characters ≤2, 30–60 seconds
- Used the three methods—cut time, cut space, cut characters—to handle an oversized story idea
- Written a five-line topic scope statement, with the shot count set between 6–20
- Confirmed that 30–60 seconds is the target runtime of the finished film (a course teaching design), with the whole film edited together from multiple short shots
Topic scope quick-reference table:
| Field | Recommended value | Notes |
|---|---|---|
| Scene | 1 location | prepare only 1 scene reference image |
| On-screen characters | ≤2 | Each character needs a character sheet + reference image |
| Final video length | 30–60 seconds | Course exercise goal, not the model's single-generation capability |
| Number of shots | 6–20 | Each shot 3–5 seconds (course design value) |
| Aspect ratio | Vertical 9:16 or horizontal 16:9 | Local route: Wan2.2 TI2V-5B officially supports 704*1280 vertical 720P |
Next Lesson Preview
With the scope locked down, the next lesson helps you choose an execution route for your topic: Cloud tools or local deployment: how to choose between the two production routes.
You will learn:
- The respective barriers of the cloud route and the local route: money, hardware, learning cost
- VRAM requirements for running Wan2.2 locally: how to read the official ComfyUI tutorial and the official script as two separate statements
- Cloud alternative paths when you don't have a dedicated GPU