The Frame Has to Become the Exit
A Case Study of City of God (2002)
Written by Joshua Allen
Founder and Executive Director, The Atlantis Awards
“The camera does not stop the crossfire. It gives Rocket somewhere to stand outside it.”
The Atlantis Awards Case Studies are not reviews. Their purpose is to examine the craft behind exceptional filmmaking—celebrating successful creative choices and identifying the mechanics behind why they work.
This article discusses the film’s complete plot and ending. It also addresses gang violence, child murder, sexual assault, police corruption, poverty, and the use of children in scenes of extreme violence and coercion.
Rocket’s first escape from the City of God doesn’t take him out of the neighborhood. It takes the film twenty years into the past.
A chicken breaks loose from a barbecue with Li’l Zé’s gang in pursuit, guns drawn, and runs itself into a dead end: a street where the gang has just come face to face with the police, weapons raised on both sides, and one young man frozen between them. Rocket can’t move forward. He can’t retreat. He’s caught between two armed systems that will define this neighborhood for the rest of the film — criminals who control the streets, and police who control which of the street’s crimes officially count.
The camera circles him. His adult body becomes a child’s body. The street opens into an earlier version of itself. The crossfire doesn’t disappear — it becomes a story with a beginning.
That transition establishes the film’s governing mechanism before Rocket has taken a single professional photograph. When the space in front of him offers no viable exit, framing creates another one. His eventual camera isn’t a symbol bolted onto a familiar choice between guns and art. It’s the material form of a power he already has as narrator — deciding where a story begins, stopping a face before violence erases it, opening a present image into the history that produced it. At the end, that same power will let him decide which photograph becomes public and which truth stays out of the frame entirely.
The gun and the badge decide who lives on this street. For the length of one photograph, Rocket decides what the street means.
The Crossfire Has to Become a Story
The opening chase is usually discussed as a compressed statement of the film’s violence — one living thing escapes slaughter only to run straight into a street where everyone else is already armed. Its more specific function is positional. Rocket’s problem isn’t only that he might be shot. It’s that neither direction in front of him belongs to him. Moving toward the police doesn’t mean safety. Moving toward the gang doesn’t mean joining the neighborhood on his own terms. Even standing still leaves him exposed to both.
The film manufactures a third direction by changing time instead of space. The men in front of Rocket become boys. Li’l Zé becomes Li’l Dice. The gang war becomes the Tender Trio robbing gas trucks and splitting the money with the block that shelters them. A confrontation that looked like spontaneous chaos starts acquiring causes. The rewind doesn’t excuse what these men become — it refuses to pretend they arrived that way.
The chase’s most famous accident says something about the production’s own relationship to control. Meirelles has recalled that the chicken chase was largely improvised and produced by chance rather than planned. The production had built the conditions for the scene; the camera stayed free enough to follow the route the animal actually took, and Meirelles kept that footage as the film’s opening. That instinct — build the conditions, then stay free enough to recognize when reality hands you something better than the plan — turns out to be exactly the skill Rocket spends the rest of the film learning to apply to his own life. He can’t manufacture the decisive moments happening around him. His power is in knowing when to hold a frame, when to follow, and when an accident has become the story.
The City Has to Perform Before It Can Be Framed
The film’s drop into the 1960s doesn’t just introduce younger versions of later criminals. It introduces a neighborhood still capable of play. The Tender Trio’s robberies are dangerous, but their bodies move through the streets with an openness the later drug war will erase — running across open fields, staging holdups with comic bravado, sharing the take, flirting, boasting. Rocket watches from the edge. Li’l Dice pushes toward the center, and when the older boys leave him outside the big motel robbery as a lookout, he doesn’t accept the assignment — he converts it into authorship through violence, killing people who’d already surrendered.
That distinction works on screen because the performers aren’t reproducing an outside idea of favela life. They’re building it from the inside. The production’s original plan was to shoot in the real City of God favela — and to test that approach, they first shot a short film, Golden Gate, there with much of the same cast. During that shoot the crew was kidnapped by drug dealers, an experience serious enough that Meirelles has said it traumatized part of the crew and forced the production to relocate the feature to other communities entirely. Fernando Meirelles has said the production interviewed roughly 2,000 young people from Rio’s favelas and brought about 200 of them into a central acting workshop that ran for four and a half to five months. Kátia Lund — who had already spent years around this world, directing the 1999 documentary News from a Personal War about Rio’s real drug-and-police conflict — did much of the work of holding the cast together through that process. Acting coach Fátima Toledo joined to build specific performances. Only one principal actor, Matheus Nachtergaele, came in as an experienced professional.
The cast largely didn’t work from a finished script. Over that same workshop period and the rehearsal that followed, Meirelles gave the performers each scene’s intention and let them improvise toward it; by his own estimate, something like 30 percent of the finished dialogue was scripted in advance. The rest was shaped with the cast.
Leandro Firmino had never acted before playing the adult Li’l Zé, and the performance isn’t built from raw aggression — it’s built from a body that has learned to make everyone else adjust around it. Alexandre Rodrigues builds Rocket in the opposite direction: a body that keeps starting toward an action and then finding a specific person standing where a function should be. That’s the quality that will eventually make him a photographer. Before he owns a camera, the performance has already established that he sees individuals where the other men around him see obstacles or opportunities.
The Story Has to Rewind Before It Can Explain
The Tender Trio’s motel robbery first plays as a mystery. The men flee. Police arrive. Bodies are left behind, more of them than the preceding scene seemed to promise. The film doesn’t explain the gap right away — it lets the massacre pass into neighborhood legend first, and only later rewinds to show what actually happened: Li’l Dice, Rocket’s older brother Goose’s young hanger-on, left outside as lookout and told to fire a warning shot only if police arrived. Unsatisfied with standing watch, he fires the shot himself as a false alarm to send the Trio running — then goes back in alone and kills the guests who’d already surrendered, smiling while he does it.
That delay isn’t a twist for its own sake. It’s the film demonstrating that framing is moral information. The first version gives the audience what the surviving neighborhood could know. The later version gives us what the apparent protagonists never witnessed. The same event changes meaning when the frame finally admits the person it had been leaving out.
The film repeats that structure through its titled digressions — the story of the apartment, the story of Knockout Ned, the story of whoever held a room, a gun, or a grudge before the person now standing in front of the camera picked it up. Each return insists that the present image is incomplete. An apartment isn’t only its current owner. A killer isn’t only the name a newspaper will eventually print.
Editor Daniel Rezende built the film’s rhythm around that same principle. Meirelles reviewed dailies with him throughout the shoot, marking footage with colored ink dots as they went. Rezende has described his own guiding rule directly: speed up and intensify a scene that’s already tense; freeze a face the story will need the audience to remember later; split the screen when both things are happening at once. The editing isn’t decorating the story — it’s doing the same selective, causal work as the narration, deciding when a face needs to become history.
Time Has to Change the Image
The three decades in this film don’t merely change clothes and music. They change the behavior of the frame. Charlone generally used 35mm for wider views and landscapes, then moved to lighter, less expensive 16mm cameras for close work and dialogue — a combination the production had already tested while making Golden Gate with much of the same cast. Charlone’s accounts support the film’s mixed-format strategy and its changing visual grammar across periods.
The look of each period tracks the same shift the story is making. The 1960s read wider and steadier, in a warm palette that allows the Tender Trio’s crimes to retain an almost playful bravado. Across the later periods, the camera moves closer and grows more agitated; by the drug-war frame, the color has cooled toward the blues that opened the film. The favela isn’t only getting more violent. The instrument recording it is losing patience along with it.
The Legend Has to Want Its Picture Taken
Long before Rocket owns a camera, he owns a gun he never fires. Goose hides a revolver in a drawer in the 1960s and tells young Rocket never to touch it. Years later, broke and freshly fired from a supermarket job after his own neighborhood accused him of running with the gangs, Rocket carries that same gun onto a bus, planning to rob the fare collector — who turns out to be Knockout Ned, who talks him out of it entirely. The gun gets carried. It never gets fired. Violence in this film belongs to people who commit to it completely, and Rocket, however desperate, keeps failing to commit.
The camera arrives instead, as a farewell gift from Benny — Li’l Zé’s oldest friend, the one person who can defuse him with a joke — given to Rocket before a party meant to send Benny off to a quieter life. A stray shot meant for Zé kills Benny at that same party, and the camera is what Rocket is left holding.
It doesn’t protect him by making him invisible. It protects him because Li’l Zé wants what it can do. As the war with Knockout Ned escalates into the newspapers, notoriety becomes its own kind of power, and Zé wants the frame the same way Ned already has it. When Rocket photographs the gang, the armed men start performing themselves — arranging their bodies, displaying weapons, waiting for him to decide when they look right. The men who could kill him are, for one instant, holding still for him. When the newspaper runs Rocket’s photo without his permission, he braces for retaliation. Zé is pleased instead — the image has converted him from neighborhood terror into public figure. His gang, notably, can’t operate the camera themselves and doesn’t understand that the film inside it needs to be developed before anyone can see it. The gap in technical literacy is real, and it’s the gap Rocket survives inside.
The camera and the gun were never quite opposites. For a while, they need each other — criminal power wants visibility, and Rocket is the only person nearby who can manufacture it.
The Face Has to Be Committed to Memory
As the war with Knockout Ned expands, the film keeps refusing to let people enter it without a history. Ned starts as a body Zé envies and humiliates; after Zé attacks his family and assaults his girlfriend, the film could reduce him immediately to a rival gunman. Instead it interrupts itself to show the sequence — grievance into alliance, alliance into armed resistance, resistance into robbery, robbery into another bereaved child who will later be the one who kills Ned.
That refusal becomes almost unbearable once children start entering the war directly. A pack of small boys — the Runts — begin running their own petty robberies in imitation of everything they’ve watched Li’l Zé do. When he catches them, he doesn’t discipline them like children. He forces two of the boys to choose whether they’ll be shot in a hand or a foot, then orders one of them, a boy nicknamed Steak and Fries, to kill the other. The scene plays as raw and uncontrolled — breath, sobbing, hesitation, bodies too small for the weapons deciding the outcome.
It was, in fact, carefully built. Meirelles has credited Fátima Toledo entirely with the sequence, saying she sometimes asked him to leave the set while she rehearsed alone with the boys and would only call him back once a scene was ready — including the one where a boy shoots his friend and cries. The film’s rawest moment of child violence was produced through structured rehearsal and repeatable performance rather than an uncontrolled breakdown captured by accident. That distinction matters for the film’s own thesis: the frame that can make a fabricated moment look unbearably real is the same frame Rocket will later use to decide what the audience gets to see and what it doesn’t.
Rezende’s freezing of important faces becomes especially consequential here — a pause that isn’t just helping the audience track a crowded cast, but marking the moment before a body becomes another function in the war. The Runts will not stay frightened children for long. By the film’s end they carry guns, execute Zé, and calmly plan who comes next.
The Photograph Has to Choose the Safer Truth
Eventually the film catches up with its own opening image. Rocket is standing between Zé’s gang and the police again — camera in hand this time. The object doesn’t remove the danger. It changes what the men around him expect him to do.
Then the shooting starts before he can finish the posed version of the scene, and Rocket keeps photographing anyway. He can’t stop the war. He can preserve what it does. Afterward, he secretly photographs the police taking Zé’s money and releasing him — a systemic truth more dangerous than any gang portrait, since it shows the apparent opposition between cops and criminals is also a transaction. Then the Runts find Zé and kill him, and Rocket photographs the body too, the film cutting still frame to still frame in time with his own shutter. The shot immediately after Zé’s death is a subjective point-of-view from Rocket’s own eyes — and according to the film’s DVD audio commentary, Charlone insisted that Alexandre Rodrigues, the actor playing him, operate the camera for it himself, teaching him on the spot rather than handing the shot to a crew operator. For one beat, actor and character are doing the same job at the same time.
He goes to the newspaper with two possible stories. One shows police corruption. The other shows the corpse of the neighborhood’s most notorious criminal. Publishing the first could expose the system — and could get him killed. Publishing the second gives the paper a sensational front page and gives Rocket a career, at the cost of suppressing the more dangerous truth he’s already documented.
He chooses the body.
That choice completes the thesis. Rocket doesn’t get out because photography reveals the truth. He gets out because he’s learned exactly what a photograph is allowed to leave outside its own borders. Both images are true. Only one is safe to publish. His exit depends on the omission as much as the evidence.
What the Frame Leaves Outside
The mechanism that gets Rocket out has a cost, and the film — and the production that made it — doesn’t fully absorb it. Both participate in it.
Inside the story, many of the film’s deaths are converted into something else on their way to the audience: a freeze-frame, a chapter title, a photograph, a career. That conversion is part of what makes two hours of a three-decade history of escalating violence watchable at all, and it’s also, unavoidably, a kind of spectacle. Brazilian critic Ivana Bentes applied exactly this argument directly to the film upon its Brazilian release, in a 2002 article contending that City of God converted the favela’s systemic deprivation into a highly consumable, cosmetic spectacle — a direct application of a “cosmetics of hunger” critique she’d developed a year earlier, contrasting the film with the deliberately unglamorous realism of Brazil’s Cinema Novo movement. That’s a real, ongoing critical debate, not a caveat invented for balance. The film’s music, humor, and editorial invention keep producing pleasure in the telling alongside the horror of what’s told, whatever the intent behind them.
The same tension runs behind the camera, and here the film’s own methods offer less protection than the fiction does. In reporting tied to the 2013 documentary City of God: 10 Years Later, Alexandre Rodrigues recalled being offered a choice after filming — a guaranteed payment of 10,000 reais (roughly $4,500 at the time) or a percentage of the film’s eventual box office. He took the certainty of the upfront payment, a rational decision that looked very different once the film became an international success; other cast members have described being offered a similar choice. The documentary doesn’t reveal one uniform outcome. Some performers, including Seu Jorge and Alice Braga, built durable careers; others remained in precarious circumstances or returned to lives largely unchanged by the film’s success. Shadow and Act’s reporting on the documentary found that outcomes broke sharply along color lines — that lighter-skinned cast members found real opportunity afterward while most of the Black actors did not.
That divide isn’t the whole story. Meirelles and Lund set up an ongoing school for the cast after the film, which grew into the organization now known as Cinema Nosso, and several performers went on to the television series and film City of Men, made by the same directors with much of the same cast. It would be inaccurate to say the production simply abandoned everyone once filming ended.
The more defensible cost is that the exit widened unevenly. The performers’ improvisations became part of an internationally celebrated cultural object, and the professional and financial benefits of that object did not return to everyone in comparable measure — nor did they track anything about who had contributed the most. Rocket’s frame creates a route outward for one witness. The production’s frame created real opportunity for some of the people who built it, and far less for others, split along lines that had nothing to do with what they’d given the film.
The Exit Has to End at the Edge
City of God begins with a chicken escaping the place prepared for its death. The production didn’t force the animal back onto the planned route — the camera followed. Rocket spends the film learning the human version of that same skill: not mastery of the street, but the instinct to notice when the expected path has failed and a different one has opened.
At the beginning, that path runs backward, through narration — trapped in the crossfire, he escapes into history long enough to explain how everyone reached the street. At the end, it runs forward, through publication — trapped between two photographs, he chooses the one that can carry his name into a future. Neither escape is complete. The rewind returns him to the same crossfire it started from. The front page leaves him carrying the knowledge of what he chose not to publish.
But the frame has given him a position violence never assigned him. He’s no longer just the boy standing between two sets of guns. He’s the person deciding where the image stops.
Rocket cannot control what happens in front of the camera. He survives by controlling what remains inside the frame. His exit is the edge of the photograph. The City of God keeps going beyond it — for the Runts still composing their list, and for the performers whose contributions to this film never translated into the same durable mobility it imagines for Rocket.
A Note on Sourcing — City of God was directed by Fernando Meirelles and co-directed by Kátia Lund, adapted by Bráulio Mantovani from Paulo Lins’s 1997 novel, with cinematography by César Charlone and editing by Daniel Rezende. Casting and workshop details, the film’s improvisation ratio, the Golden Gate/kidnapping incident, and the account of the Runts sequence are drawn from Meirelles’s 2024 BFI interview and an earlier Nitrate Online interview. Kátia Lund’s prior documentary work is drawn from published biographical reporting on her career. Charlone’s accounts support the film’s mixed 35mm/16mm strategy and its changing visual grammar across periods; the essay’s more specific description of color and emotional effect in each period is Atlantis’s reading of the finished film. Editing details are drawn from Meirelles’s account of his review process with Rezende and from Rezende’s own published remarks on his editing philosophy. Alexandre Rodrigues’s role operating the camera for the subjective shot following Zé’s death is drawn from the film’s DVD audio commentary, as reported in subsequent retrospective coverage. The account of cast compensation is drawn from Alexandre Rodrigues’s own recollection in reporting tied to the 2013 documentary City of God: 10 Years Later; the racial disparity in long-term outcomes is drawn directly from Shadow and Act’s reporting on that documentary. The account of Cinema Nosso and the City of Men projects is drawn from published interviews with the documentary’s co-director. Ivana Bentes first formulated the broader “cosmetics of hunger” concept in a 2001 Jornal do Brasil essay and applied it directly to City of God in her 2002 article “Cidade de Deus promove turismo no inferno,” corroborated in subsequent academic and press coverage of the film. The governing argument — that the film converts framing itself, through photography, editing, narration, and chronology, into Rocket’s only viable route out of the violence around him, and that this same mechanism describes the production’s own relationship to the people who built it — is Atlantis’s interpretation of the finished film.