Flashback: The Rotoscoped Sequel-That-Wasn't

Ask a room of people who grew up with an Amiga what Flashback was, and a good half of them will tell you it was the sequel to Another World. It wasn’t. It shares a publisher, a year band, a rotoscoped protagonist and a side-on camera, and that has been enough to keep the misattribution alive for over three decades.

Here is the actual record. Another World was Eric Chahi, working largely alone, published by Delphine Software in 1991. Flashback: The Quest for Identity was led by Paul Cuisset with a proper team at the same studio, released in 1992. Cuisset had come off the Future Wars and Operation Stealth adventure games. He was interested in different things. The resemblance is a house style and a shared building, and treating the second game as a follow-up to the first has cost it thirty years of being read on its own terms.

Because on its own terms it is one of the best-controlling games of the sixteen-bit era, and the reason is that it makes you commit.

Controls that cost something

Conrad B. Hart does not have an analogue run. He has states, and moving between states takes time and cannot be interrupted.

Tap a direction and he takes a single measured step. Hold it and he transitions — through a windup animation you must pay for — into a run. A standing jump goes about a body-length. A running jump goes further and cannot be aborted mid-flight. Rolling, drawing the pistol, holstering it, hauling yourself up a ledge: every one of these is an animation with a duration, and the game will not let you cancel out of it because you changed your mind.

Modern design orthodoxy hates this. Responsiveness is the first commandment; input latency is a bug; the character should be a cursor with legs. And there is a whole class of game where that is correct. Flashback is arguing the opposite, and it wins the argument, because the cost of the animation is what converts a platform into a decision.

A gap in Flashback is a question: from here, do I have room for the windup? An enemy round the corner is a question: pistol out first and lose two beats, or roll past and hope? You are not steering Conrad. You are issuing him orders and then watching him carry them out while the world keeps moving. Every failure is legible as your own bad planning, which is precisely what makes retrying feel fair rather than arbitrary.

The rotoscoping is what sells it. Cuisset’s team filmed reference and traced it — the same technique Mechner used for Prince of Persia and Chahi used for Lester — and Conrad has more animation frames than anyone had business affording in 1992. He hangs off ledges with his weight in the right place. He staggers when hit. The animations are long because they were traced from a real body doing a real thing, and the mechanical commitment falls out of the fidelity. The design and the art asset are the same object, which is the mark of a game that knows what it is.

The bit everyone forgets: it has a plot

Another World refuses to explain itself. It has no dialogue, no HUD, and no text during play, and its ending resolves nothing. This is where the two games actually part company, and it is why the sequel framing does Flashback such damage — judged as a follow-up it looks garrulous, and judged as itself it is doing something Chahi had no interest in.

Cuisset built a science-fiction thriller with an actual structure. Conrad wakes on a jungle moon of Saturn with his memory wiped and a holocube from his own past telling him to get to Titan. The plot from there runs through a shapeshifting alien species, a corrupted human government, a televised murder game show, and an escalation to a genuine climax. There are cutscenes. There is exposition. There is a twist.

It works because the delivery is disciplined. The cutscenes are short, polygonal, and staged like film — cut, hold, cut — and the game never stops for a conversation you have to click through. The pre-rendered backdrops give each act a distinct visual grammar: the green tangle of the jungle, the vertical neon of New Washington, the sterile institutional grey of the Death Tower. You always know where you are in the story from the wallpaper.

The Death Tower section is the one people remember and it is the best piece of design in the game — the fiction and the mechanics collapse into each other. Conrad needs money and papers, so he goes on a lethal TV game show, and suddenly the arena-crawl level design is diegetic. You are being watched because you are literally being broadcast. That is a trick The Running Man did in cinema and games have been doing badly ever since; Flashback did it cleanly in 1992.

Where it creaks

I am not going to pretend the whole thing holds. The mid-game runs on a class of task the era loved and I have no patience for: talk to the man, fetch the object, bring it back, receive the next fetch. It is padding, it is transparently padding, and it exists because the game needed hours. The commitment-heavy movement that makes combat tense makes backtracking a chore, since you are paying the full animation cost on every traversal of ground you already understand.

The combat itself is thinner than the movement deserves. Draw, shield, fire. Most fights resolve to the same drop-and-shoot rhythm once you have it, and the game keeps deploying it long past the point of discovery. Chahi solved this by keeping his game to about two hours and retiring each idea after one good use. Cuisset had a commercial product to fill.

It filled it well enough. Flashback sold north of a million copies and became, for a stretch, the best-selling French-made game ever — a Guinness entry, no less. That commercial success is exactly why it got the sequel (Fade to Black, 1995, in polygons, and a much rougher piece of work) and why its reputation is bigger than the smaller, stranger, better-directed game it keeps getting attached to.

The port problem

Flashback travelled further than almost anything else off that studio’s floor, and the travelling tells you what the game actually is.

The Amiga original is the reference. The MS-DOS conversion is close. The Mega Drive version is the one most people outside Europe actually played, and it holds up remarkably well because Conrad’s animation survives the transfer intact — the frames are the game, and the Mega Drive had the sprite budget to carry them. The SNES version runs slower and the difference is immediately legible in the movement, which is the clearest possible proof that the timing of those animations is load-bearing rather than decorative.

Then come the CD releases — Sega CD, 3DO, CD-ROM PC — which bolt on an anime-styled animated intro and voice acting. They are worse. The voice acting is 1994 CD-era voice acting, which is to say it explains things the polygon cutscenes had already staged perfectly well, and the animated opening replaces a piece of tense wordless direction with a cartoon. Given more storage, the instinct was to fill it, and every megabyte of filling made the game slightly less confident. Chahi’s portable bytecode approach had the opposite effect on his game’s ports: constrained by design, they stayed themselves.

Why it holds up

Play it now and the thing that survives is the weight. Everything else in that generation of Amiga design has aged into a museum piece — the parallax that sold hardware reads as a tech demo today, and most of the era’s platformers control like a hovercraft. Conrad still feels like a body.

The lineage is easy to trace and worth tracing. The original Tomb Raider has Flashback’s grid-and-commit movement in three dimensions — you step, you measure, you jump, and you accept the consequences. Blade Runner’s Westwood adventure inherits the same genre-cinema borrowing instinct. Limbo and Inside took the deliberate movement and dropped the fetch quests. Every soulslike that makes an attack animation an unrecoverable commitment is running Cuisset’s logic with a bigger budget: your input is a promise, and the game will hold you to it.

The verdict: a genuinely great control system attached to a good thriller and about ninety minutes of busywork, permanently misfiled as somebody else’s sequel. Read it as Cuisset’s game and it gets substantially better, because the thing it does that nobody else was doing — making movement expensive on purpose — is the thing that outlived the entire generation it came from.

Where to play it: the 1992 original runs on anything with an emulator, and the Amiga and Mega Drive versions are the ones to want. There is a 2023 remaster that adds a rewind and difficulty options; the rewind quietly dismantles the commitment that makes the game work, so use it sparingly or leave it alone.

Spoilers below

The identity plot lands: Conrad’s wiped memory turns out to be self-inflicted. He erased himself deliberately, because he had discovered the Morphs — a shapeshifting alien species quietly infiltrating human government — and knew they would read his mind if they caught him. The holocube is a message from Conrad to Conrad, which recasts the whole opening as a man following his own instructions without knowing why.

The finale sends him to the Morph homeworld and he blows it up, which is a much more conventional resolution than a wounded man being carried away over a ruined city. That gap is the entire difference between the two games, and it is why only one of them gets called a sequel to the other.