Montage

What is a montage in film?

In English-language filmmaking, a montage (or montage sequence) is a self-contained run of brief shots that summarizes action across time or space. A boxer goes from wheezing on the stairs to sprinting up them; a couple meets, marries and grows old. The shots are usually cut to a song or a score cue, with little or no synchronous dialogue, and the whole thing lasts from thirty seconds to a few minutes.

The word causes confusion because it has two meanings. In French and Russian, montage simply means film editing — the whole craft of putting shots together. In English it narrowed to this specific device. And in film theory, Soviet montage names a 1920s school that treated the cut itself as the source of meaning. This article covers the last two separately.

What makes a montage different from ordinary scene editing:

• Time is openly compressed. Continuity editing hides the gaps between shots; a montage announces them. Each cut may skip an hour or a year. It is the most direct tool for showing the passage of time.
• Shots are units of an argument, not of a scene. Each image stands for many similar moments — one sit-up stands for a thousand.
• Music replaces the scene's own sound. The track gives the fragments a shared rhythm and an emotional line.
• There is an arc. A good montage starts in one state and ends in another. The last shot is a payoff, often held longer than the rest.

Structurally, a montage is one sequence bridging two scenes that play in real time.

What is Soviet montage theory?

In the 1920s a group of Soviet filmmakers — Lev Kuleshov, Sergei Eisenstein, Vsevolod Pudovkin and Dziga Vertov — argued that cinema's power lives in the cut, not in the individual shot. Kuleshov's workshop showed that viewers read the same face differently depending on the image placed next to it, the result now known as the Kuleshov effect.

Eisenstein pushed hardest. For him a cut was a collision: shot A plus shot B produces an idea contained in neither, much as two characters combine into a new word. In Strike (1925) he cuts the massacre of workers against the slaughter of a bull; the audience draws the comparison. In Battleship Potemkin (1925) the Odessa Steps sequence fragments one event into dozens of angles — boots, faces, a runaway pram — stretching and shattering real time rather than compressing it. That is the key difference from the Hollywood sense: Soviet montage is about meaning through juxtaposition, not about summarizing weeks of story.

Eisenstein described several methods, from the plainest to the most abstract:

• Metric — cuts at fixed lengths, regardless of content, to set a pulse.
• Rhythmic — cut lengths follow movement inside the frame.
• Tonal — shots grouped by their emotional tone: light, shape, texture.
• Overtonal — the combined effect of the first three over a whole passage.
• Intellectual — images combined to express an abstract idea, as in the bull and the workers.

Vertov's Man with a Movie Camera (1929) applied the same thinking to a city: a day of work and leisure built from hundreds of fragments, with no plot at all. That "city symphony" line runs through Koyaanisqatsi (1982).

In the 1930s Hollywood hired montage specialists such as Slavko Vorkapich to build short transition sequences — spinning headlines, falling calendar pages. The collision theory was traded for a practical job: get the story from here to there, fast.

Why filmmakers use a montage sequence

A montage solves a structural problem: the story needs a change that takes months, and the film has ninety seconds for it.

Compressing a process. The training montage is the familiar case, but the same logic serves building a house, planning a heist, studying for an exam or rising in a criminal empire, like the "Push It to the Limit" run in Scarface (1983).

Summarizing a relationship. The "Married Life" sequence in Up (2009) covers decades of a marriage in about four wordless minutes scored by Michael Giacchino. It works because it repeats motifs — the savings jar, the tie, the hill — and changes them each time, so the audience tracks time and loss at once.

Showing decline. In Citizen Kane (1941) a series of breakfast-table scenes, linked by whip pans, shows a marriage cooling over years: the table gets longer, the talk shorter, until the couple read rival newspapers. In Requiem for a Dream (2000) a repeated set of fast, sound-stamped extreme close-ups — pills, dilating pupils — makes the ritual of addiction feel mechanical.

Releasing emotion that has been held back. The ending of Cinema Paradiso (1988) is a montage of kisses a priest once cut out of the films. The images belong to other movies; the meaning comes entirely from what the audience knows about the man watching them.

The cost is intimacy. A montage keeps the viewer at a distance from any single moment, so it rarely carries a turning point. The decision to train belongs in a scene; the training itself goes in the montage; the fight it prepares for is a scene again. Montage also changes the pace of a film sharply, which is why editors often place one where the story needs a lift after a heavy dialogue stretch.

What makes a good training montage?

The training montage became a genre of its own after Rocky (1976): Rocky's early-morning runs through Philadelphia, cut to Bill Conti's "Gonna Fly Now," end on the steps of the Philadelphia Museum of Art with his arms in the air — the same steps that left him winded and doubled over earlier in the film. Rocky IV (1985) turned it into a cross-cut duel, alternating his farmhouse training in the snow with his rival's high-tech lab. Mulan (1998) staged it as a song, "I'll Make a Man Out of You," where the recruits fail a task at the start and the hero solves it at the end.

The pattern behind the ones that work:

1. Show the failure first. The opening beats establish what the character cannot do: the dropped weight, the missed punch, the arrow nobody can retrieve.
2. Repeat one test. Return to the same exercise two or three times so progress is measurable on screen.
3. Change the conditions. Rain, night, snow, a crowd joining the run — the world escalates with the character.
4. Pay off with a shot that echoes the first failure. Same angle, same place, opposite result. The payoff shot is usually the longest in the sequence.
5. Cut to the music's structure. Build through verses, save the widest and most heroic images for the final chorus.

Audiences know these steps before they happen. That familiarity lets you skip setup, but it also makes a lazy training montage feel like a cliché instantly.

How to shoot and cut a montage

A montage is written in the edit but won or lost in planning. Treat it as its own short film with a beginning, middle and end.

Write the arc in one line. "She can't run a mile; by winter she runs the harbor wall at dawn." Everything you shoot should serve that before and after.

Budget the shots. A 90-second music cue typically holds 20–40 shots at 2–4 seconds each; a fast-cutting climax may drop to half a second per shot. Shoot about twice as many beats as you plan to use. Beats are cheap to cover and costly to reshoot.

Cover each beat in several sizes. For every action, get a wide for geography, a medium for effort and an insert shot for detail: the taped knuckles, the stopwatch, the calendar. Inserts are the glue; they let you cut anywhere without continuity problems.

Plan a visual progression. Change light, weather and wardrobe across beats so time reads instantly: first-light blue to noon white to golden hour, bare trees to leaves. Keep one element constant — a location, a prop, a framing — so the change registers against it.

Design the transitions. Hard cuts on the beat feel energetic; a dissolve between beats feels reflective and says "time is passing." A match cut — the same gesture in two different years — is the most elegant bridge montage has.

Pick the music early. Temp the montage to the real track, or at least the same tempo, before the shoot, and time key actions to musical phrases. A 120 BPM song gives a beat every half second; cutting every two or four beats keeps the sequence locked to it.

Give it a button. End on a held shot that lands the change — the fighter at the top of the stairs, the finished boat, the empty chair. Without it, the montage just stops.

On set, work from a list of beats rather than scenes, with a small crew, a light handheld or gimbal rig and lenses around 25–85mm.

Montage vs. cross-cutting

Both cut rapidly between different images, and they are often combined, but they do different jobs.

• Cross-cutting alternates between two or more lines of action happening at roughly the same time in different places. Its engine is simultaneity and suspense: will the rescue arrive before the train?
• A montage summarizes one line of action across a long span of time. Its engine is compression: we skip the months and see the change.

The baptism in The Godfather (1972) is the classic hybrid. It cross-cuts Michael at the font with the killings of rival family heads, and because the church organ and the liturgy carry across every location, it plays like a montage — one idea, Michael's double oath, built from fragments.

A quick test: if you could reorder the shots within each strand without breaking the story, you have a montage. If each shot depends on the one before it in the same strand — a fuse burning shorter — you are cross-cutting.

A montage only exists in motion, so watch the cuts rather than any single frame. Notice how short each shot is, what repeats from one fragment to the next, and the one idea the pieces add up to: a ritual, a preparation, a city, a career.

AI will not cut a montage for you, but it is the fastest way to previsualize one: generate each beat as a frame, animate the ones that need motion, and lay them out in order to check the arc before anyone shoots. On a FlashBoards board the reference frames, every take and the beat list sit side by side, so you can see whether the first and last shots really rhyme. The fourth prompt re-shoots a real film shot from a text description alone, and the original plays next to the result so you can compare.

Common mistakes

• One frame, no arc. Models produce a single beautiful image; a montage needs a before and after. Prompt the opening beat and the payoff beat separately, with the same framing and opposite results.
• Character drift. The face and wardrobe change from beat to beat. Paste the same descriptor string (scar, braid, taped hands) into every prompt and keep the best frame on the board as reference.
• Everything in the same light. Without explicit time cues every beat looks like the same afternoon, and time stops passing. Name the hour, the weather and the season in each prompt.
• Asking a video model for the cuts. Prompts like "fast montage of training shots" produce a muddled single shot. Generate one shot per beat and cut them together in the edit, or spell out every shot and its cut time, as the fourth prompt does.
• Stock-photo motivation. "Inspiring athlete" returns clean, posed gym imagery. Ask for effort and wear: sweat, grime, chipped paint, torn tape.

Further reading

Sergei Eisenstein, Film Form: Essays in Film Theory (1949). Karel Reisz and Gavin Millar, The Technique of Film Editing. Walter Murch, In the Blink of an Eye.