Back to the course
Lesson 06Intermediate9 min

Camera language: shot types, movement, and format

A working film-school vocabulary for HappyHorse-1.0 — shot sizes, camera movements, lens language, and how to choose aspect ratios per platform.

Updated 2026-07-17

HappyHorse-1.0 was trained on footage described the way filmmakers describe footage. So the fastest way to lift your results from "AI clip" to "shot" is to use the real vocabulary — the same fifteen or so terms a director calls across a set. This lesson is that vocabulary, organized for prompting.

Shot size (how close are we?)

  • Wide shot / establishing shot — geography first, subject second. Use to open a sequence.
  • Full shot — the whole body; best for action and dance.
  • Medium shot — waist up; the conversational default.
  • Close-up — the face; where lip-sync scenes (Lesson 5) live.
  • Extreme close-up — an eye, a ring, a droplet. High drama, short duration.

One size per clip. "Wide shot that becomes a close-up" is legal — but say it as a camera move: "slow push-in from wide to close-up."

Camera movement (does the frame move?)

TermWhat it doesFeels like
Static shotNo movementComposed, observational
Slow push-inDolly toward subjectRising tension, focus
Pull-backDolly awayReveal, release, loneliness
Tracking shotFollows a moving subjectEnergy, momentum
Orbit / arc shotCircles the subjectShowcase, awe (great for products)
Crane up / downVertical sweepScale, grandeur, endings
HandheldUnstabilizedDocumentary, urgency, realism
Aerial / drone shotHigh and movingScope, geography

Two rules. One movement per clip — 3 to 15 seconds fits exactly one camera idea; "orbit then crane up" produces mush. Match movement to emotion — push-in for intimacy, pull-back for context, static when the subject's own motion is the show.

Lens and depth language

You're not choosing glass, but the model knows what glass does:

  • "Shallow depth of field" — subject sharp, background melted; the instant "cinematic" switch.
  • "Deep focus" — everything sharp; documentary and landscape energy.
  • "35mm film look" — grain, halation, organic color; pairs well with handheld.
  • "Anamorphic widescreen" — oval bokeh, horizontal flares; the blockbuster look.
  • "Macro lens" — for the extreme close-ups above.

Aspect ratio is a distribution decision

HappyHorse renders natively in five ratios — pick by where the clip will live, and say it in the prompt as well as the settings (the wording nudges composition):

RatioPlatformPrompt tag
16:9YouTube, web heroes, TV"widescreen cinematic"
9:16TikTok, Reels, Shorts"vertical short-form video"
1:1Feed posts, product tiles"square format"
4:3Retro / editorial"classic 4:3"
3:4Pinterest, portrait feeds"portrait format"

Composing for 9:16 changes more than the crop: subjects sit lower, faces bigger, motion more vertical. If you need both a widescreen and a vertical cut, generate both natively rather than cropping one from the other — the model recomposes the shot properly for each.

Putting it together: a three-shot sequence

A coffee brand spot, one product, three generations:

  1. Establish: "Aerial drone shot descending toward a mountain-side coffee farm at sunrise, mist in the valleys, warm golden light, cinematic photorealism, serene — widescreen"
  2. Feature: "Orbit shot around a matte-black coffee bag on a wooden table, beans scattered, shallow depth of field, soft window light, premium and calm"
  3. Human close: "Close-up of hands cradling a steaming mug, slow push-in, morning light through linen curtains, quiet contentment, soft acoustic underscore"

Same lighting family, one camera idea per shot, escalating intimacy: wide → product → human. Cut together, it reads as directed — because structurally, it was.

Next lesson leaves the director's chair for the terminal: calling HappyHorse-1.0 from code, with real per-clip prices.