a set11 / 14
Short film
Storyboard, quote per shot, generate in order — and the shots stay separate.
apipod-short-filmSKILL.md on GitHub
When to use it, and when not to
when to use it
- The request is a sequence, a scene, an ad, or a short film rather than one clip.
- Continuity across cuts matters: the same character look, wardrobe, prop or location.
- The spend needs to be agreed before five generations run.
not for
- apipod-generate-video — One clip, no continuity problem.
- apipod-plan-video — Only the plan is wanted; nothing should be produced yet.
procedure
6 skillsWhat the skill does, in order.
- 01
Plan in text, not in video
Three to five shots. For each: what happens, camera angle and movement, duration, and what carries continuity. A five-shot 1080p sequence can cost more than fifty images.
- 02
Name the continuity anchor explicitly
Pick the one element that must survive across cuts — the character look, the wardrobe, the prop, the location — and decide how: a first frame reused as image_urls, or a fixed description pinned in every prompt.
- 03
Show a storyboard table and stop
Shot, beats, seconds, model. Do not generate yet. A storyboard the user has not approved is not authorization to spend on five generations.
- 04
Cost the whole film per shot
Read duration and resolution enums from get_model_params rather than guessing which durations a model supports, then estimate_cost per shot and state the total, naming the shots that dominate the cost.
- 05
Shot 1 alone, confirm the look
A wrong look multiplied by five is the most expensive mistake in this skill. Later shots pass the approved frame as the first frame and restate the anchor. check_job every 10 s.
- 06
Deliver the shots as shots
Permanent URL per shot, in order, with duration and cost_usd, plus a total. If the result does not hang together, name the specific cut that breaks — colour, motion direction, speed of action.
try it
Send it from here
example request
Use APIPod to plan a three-shot espresso sequence — grinder, pour, first sip — as a storyboard with per-shot costs. Generate only after I approve both.
The buttons open a new chat with this prompt prefilled. If APIPod is not connected yet, the prompt tells your assistant to stop and help you connect first. In Claude Code or a terminal, copy the prompt instead.
Not set up yet? Wiring guides for Claude, ChatGPT, Codex and more →
call trace
get_model_params (duration · resolution enums)
→ no guessing numbers
estimate_cost (per shot)
→ 3 shots · $— total · waits for a yes
generate_video (shot 1)
→ vid_f1 · later shots reuse its frame
check_job (every 10 s)
→ 3 · completed · no stitching
Illustrative: the tools this skill runs, in order, for the request on the left. Prices appear the way estimate_cost returns them live — this page itself quotes nothing.
what it will not do
The rules it holds itself to.
01
APIPod does not stitch clips. Do not offer a finished single file; say the shots are separate and the user or an editor assembles them. Only give an ffmpeg command if asked how to concatenate.
02
A failed shot gets its error_code, error_message and refund reported, and a question — not an automatic retry.
03
Silently downgrading the model is never the way to fit a budget. Change length or resolution and re-quote.