Orelon logoOrelon
料金

AI Shot Lists for Cinematic Wedding Films: A Practical Guide

2026年11月14日 · Orelon Team 著

AI動画テンプレートを見る

着想のためにコミュニティ作品をいくつか閲覧し、任意のテンプレートを開いて Orelon で作成を続けましょう。

Learn how to build AI-assisted shot lists that keep wedding films consistent, emotional, and cinematic from pre-production through final delivery.

A wedding film is usually decided before the vows. The couples remember the day, but they judge the film by whether it feels like the day — and that feeling is the product of hundreds of small decisions made in pre-production: where the camera stands during the first look, how the toasts are lit, which reaction shot carries the emotional turn. The strongest wedding cinematographers in competitive markets are not winning on gear. They are winning on planning documents that are specific enough to survive contact with a real, chaotic, schedule-driven wedding day.

That is where an AI-assisted shot list earns its place. Not as a gimmick that generates a wedding video for you, but as a production instrument: a structured, model-readable list of shot intents that you can pre-visualize, stress-test, share with a second shooter, and hand to a generative tool when you need a concept animatic, a risky-shot test, or a stylized sequence that would be impractical to capture live.

The benchmark is a craft standard, not a zip code

When people talk about the top tier of wedding cinematography, they are usually describing a set of expectations rather than a location. Those expectations are consistent almost everywhere:

  • Documentary honesty. Real moments, real audio, minimal staging.
  • Editorial polish. Clean framing, intentional color, controlled contrast.
  • Emotional pacing. A film that breathes instead of a montage that sprints.
  • Technical reliability. Exposure and audio that hold up in a dark ballroom and a bright rooftop in the same afternoon.

In dense, high-expectation markets — New York is the classic example — you also inherit a specific set of constraints: tight venue windows, permissions for shooting on public streets, multi-floor venues with freight elevators, ceremonies that start late, and receptions lit for ambience rather than for cameras. A shot list that ignores those constraints is fiction. A shot list built with them becomes a schedule, a negotiation tool, and a creative safety net.

AI does not change those constraints. It changes how quickly you can explore solutions to them.

What an AI shot list actually is

A traditional shot list is a table of frames. An AI-ready shot list is a table of intents — written precisely enough that a generative video or image model can produce a representative frame, and a human crew can reproduce that frame for real.

Each row should carry, at minimum:

  1. Story beat — what the shot is doing emotionally (arrival, anticipation, vow reaction, release).
  2. Subject and action — who, doing what, in one clause.
  3. Framing — wide, medium, close, insert; subject placement.
  4. Lens feel — wide-angle intimacy, normal perspective, compressed long lens.
  5. Camera movement — locked, slow push, handheld follow, orbit, crane, gimbal walk.
  6. Light — motivated source, direction, quality, contrast ratio.
  7. Duration and priority — how long the edit needs, and whether it is a must-have or a nice-to-have.
  8. Audio note — whether this shot must sync to speech, ambient sound, or music only.

Rows eight columns deep look heavy. They are also the difference between a list your second shooter can execute without a phone call and a list that gets abandoned by 4 p.m.

Build the list in five layers

Trying to write a full wedding shot list in one pass produces a document that is either generic or unmanageable. Build it in layers instead, each one narrowing the previous.

Layer 1 — Story beats

Start with seven to twelve beats: prep details, first look, ceremony processional, vows, rings, recession, portraits, entrance, toasts, first dance, late-night release. Give each a target emotional note in a single word. Every shot you add later must attach to a beat. Shots that attach to nothing are the ones that quietly eat your schedule.

Layer 2 — Coverage and redundancy

For each beat, define the minimum viable coverage: one wide for context, one medium for action, one close for emotion, one insert for texture. Then decide where you cannot afford redundancy because of time, and where you can afford two angles because the moment is unrepeatable. Vows, toasts, and the first dance are almost always the latter.

Layer 3 — Look lock

This is the layer most AI work skips, and it is the one that makes output usable. Write a single paragraph you reuse in every prompt: film stock feel, grain, contrast, color bias, highlight rolloff, aspect ratio, and the specific warm/cool split you use for daylight versus candlelight. Keep it under 60 words and never improvise it mid-project. Consistency across a sequence comes from repetition of the lock, not from cleverness in individual prompts.

Layer 4 — Motion and light language

Decide your movement vocabulary before the day: how many push-ins, how much handheld, whether you orbit during dances. Then describe light in practical terms — a window at camera left, a warm tungsten pool from a single pendant, an overcast sky as a soft top light. Generative models respond far better to a described light source than to adjectives like "beautiful" or "cinematic."

Layer 5 — Audio and edit intent

Annotate which shots must carry live sound and which will live under music. This single column changes your camera placement: a sync-sound vow shot needs a clean mic line and a stable frame, while a music-only detail shot can be handheld and imperfect. Deciding this after the wedding is how you end up with gorgeous footage you cannot cut.

Continuity: the part AI makes or breaks

Cross-scene continuity is the hardest problem in AI-assisted film work, and weddings are unusually demanding because the same two people appear in every scene under wildly different light.

A practical continuity kit:

  • Character references. Build a small set of approved frames per principal — front, three-quarter, profile, full length — and reuse them as references for image-to-video work rather than re-describing faces in text.
  • Wardrobe anchors. Note the exact outfit, accessories, and hair state per scene, including the transition states (jacket off, veil removed, shoes changed).
  • Location anchors. One approved wide frame per venue that establishes the palette; reuse it as a reference for every shot in that space.
  • Palette rules. Two or three named color relationships — for example, cool daylight neutrals with warm skin, or amber interior with deep shadow — and a rule for when each applies.
  • Aspect ratio and grain lock. Fixed for the entire project. Changing either mid-sequence destroys the illusion faster than a bad face.

The AI image generator is genuinely useful here: generate a handful of reference stills per scene, approve them, then use those stills as the visual anchor for motion tests. Approving stills is fast and cheap; approving motion is slow.

Write prompts like a shot, not like a wish

Most disappointing AI output traces back to prompt structure, not model capability. Four habits fix the majority of it.

One camera move per shot. "Slow push-in on the bride's hands" works. "Push in while orbiting and racking focus" produces mush.

Describe light as a source, not a mood. "Soft window light from camera left, warm bounce from a wooden floor" gives the model something to build.

Separate subject from style. Put the scene description first, then the look lock paragraph. Mixing them causes the model to style the subject instead of the frame.

State duration and pace. A six-second shot needs an action that completes; a three-second shot needs a single gesture. Write the gesture, not the emotion.

Keep a working prompt library per project: the ten prompts that produced approved frames, stored alongside the approved output. It becomes faster than rewriting from scratch, and it keeps a long edit visually coherent. If you want a head start, the prompt library has reusable structures you can adapt to your own look lock.

Pre-visualization that actually earns its keep

Pre-visualization has a bad reputation in wedding work because it is often mistaken for a mood board. It is not. Its job is to answer three questions before the day:

  1. Does the concept read without sound? If the animatic does not communicate the emotional arc muted, the edit will not save it.
  2. Which shots are risky, and what is the fallback? Night exteriors, drone shots in restricted airspace, sparkler exits, rain plans. Test the idea cheaply in AI, then decide whether it is worth the schedule cost on the day.
  3. Does the couple recognize themselves? A short animatic built from approved reference stills is one of the best client-alignment tools available. It surfaces mismatched expectations in week two instead of at delivery.

A useful sequence to target: a 30–45 second teaser animatic built from eight to twelve approved frames, cut to the actual music you plan to license. That is short enough to iterate on and long enough to expose pacing problems.

Wedding-day workflow: where AI helps and where it does not

Be honest about the boundary. AI cannot be at the wedding. The human crew captures the irreplaceable material: vows, reactions, the hand on the shoulder, the audio that makes people cry.

Where AI earns a place in the surrounding workflow:

  • Pre-day: concept animatics, risky-shot tests, lighting references for unfamiliar venues, title and graphics design.
  • On the day: the shot list as a live checklist with priorities, so a delayed ceremony does not cost you the must-haves. Nothing more ambitious than that.
  • Post-day: stylized inserts and transitions, establishing shots you could not legally capture, pickup shots for editorial gaps, social cutdowns, and animated title sequences.
  • Delivery: multiple aspect ratios for social, teaser variants, and trailer-style opens that reuse existing graded footage.

A realistic rule: use AI for anything that would otherwise require a second crew, a permit, or a reshoot. Do not use it for anything the couple will interpret as fabrication of their actual day.

A worked example: the rain-plan reception

A couple plans a rooftop cocktail hour and an indoor reception. Forecast says rain. Here is how the shot list handles it.

Beat 1 — Arrival. Two versions of the same three shots: umbrella walk with wet pavement reflections, and a covered-entry version with warm interior light spilling out. Both are pre-visualized, so the decision is a choice rather than a scramble.

Beat 2 — Toast reactions. Priority flagged as must-have. Two angles specified, one on the speaker, one on the couple, both sync-sound. No AI involved.

Beat 3 — Room establishing shot. The venue's ballroom is beautiful but too small for the wide the edit wants. A single AI-generated establishing shot, built from an approved still of the room and matched to the palette lock, fills the gap in three seconds of screen time. Because it is built from a real reference, it reads as the same room.

Beat 4 — Late-night. The couple wants a sparkler exit that the venue will not permit. The alternative is a generated stylized sequence clearly framed as an artistic epilogue, with the real exit footage still included. Transparent, and it protects the relationship.

The lesson: the shot list's value is not that every row gets generated. It is that every row has a decision already made.

Mistakes that quietly ruin AI-assisted wedding work

  • Over-listing. Ninety shots in a five-hour window means forty get skipped. Rank ruthlessly; twenty must-haves beat ninety maybes.
  • No priority column. Without it, the crew shoots what is convenient, not what the edit needs.
  • Improvising the look lock. Each new prompt drifts. Six shots later, the film looks like three different projects.
  • Ignoring the audio column. Sync-sound moments require different framing, different mics, and different time. Decide early.
  • Treating AI output as finished. Generated frames need grading, grain matching, and usually a slight crop to sit beside real footage.
  • Using AI where authenticity matters. Reactions, vows, and family moments should never be reconstructed. Audiences forgive stylistic inserts; they do not forgive fabricated emotion.
  • No fallback per risky shot. Every ambitious row needs a "if this fails, we shoot that" line.

Choosing tools: decision criteria that matter

Tool selection for wedding work is less about model rankings and more about whether the tool fits a real production pipeline. Evaluate on:

  • Image-to-video control. Can you start from an approved still? Continuity depends on it.
  • Duration per generation. Wedding inserts need three to eight seconds; anything shorter multiplies your edit work.
  • Aspect ratio and resolution. Vertical cutdowns are not optional anymore.
  • Iteration speed. Fast, cheap drafts matter more than a perfect final render, because most of your time goes into selection.
  • Motion realism on people. Hands, hair, and fabric are the tell. Test on your own footage style before committing a project.
  • Export and grading headroom. Flat or log-friendly output saves hours in post.

Compare options against your actual workflow rather than spec sheets — Orelon's video generator is built around idea-to-motion iteration, and the templates are useful for title and teaser structures when you do not want to design from zero. If you are weighing platforms, our alternatives comparisons focus on workflow fit rather than feature counts.

FAQ

Do I need an AI shot list if I already have a shot list?

Not a separate one. Convert your existing list into the eight-column format. The gain comes from writing light, motion, and audio intent explicitly — the same discipline that makes generative tools usable also makes a human crew faster.

How many shots should a wedding shot list contain?

For a full-day, single-crew wedding, 40–60 rows works for most filmmakers, of which 15–25 are flagged must-have. Multi-crew teams can go higher. Beyond that, the list becomes a wish list nobody reads on the day.

Can AI generate an entire wedding film?

No, and it should not try. The parts that make a wedding film matter — vows, reactions, real sound — must be captured live. AI is strongest for pre-visualization, stylized inserts, establishing shots, and title or teaser sequences.

How do I keep the same couple looking consistent across generated shots?

Use approved reference stills per principal and per location, keep a fixed look-lock paragraph, and lock aspect ratio and grain for the whole project. Re-describing faces in text every time is the most common cause of drift.

What is the biggest time-saver in this workflow?

Approving stills before generating motion. A frame takes seconds to judge and a sequence takes minutes; approving the frame first removes most of the wasted render time.

Is it ethical to include generated shots in a wedding film?

Yes, with transparency. Use AI for stylistic, clearly authored sequences and for shots that could not otherwise exist, and never to reconstruct moments the couple believes were captured. Tell clients where it is used; most couples find it interesting rather than objectionable.

Start with the list, not the render

The filmmakers who get the most from AI in wedding work are not the ones with the most tools. They are the ones whose planning documents are specific enough that any tool — or any second shooter — can execute them. Build the eight-column list, write the look lock once, rank the must-haves, and pre-visualize only what is genuinely risky.

When you are ready to test an idea in motion, Orelon is built for exactly that: turning a described shot into a cinematic frame you can approve, iterate on, and cut into a real edit. Start with a single beat from your next wedding, generate three variations, and see how much faster the day goes when the decisions are already made. For more workflow breakdowns, the Orelon blog covers pre-visualization, prompting, and post-production craft.