Vision and what the app does
We narrate our days in emojis already. A car, a coffee, a rage face, a laptop. That is not a status update, that is a plot. This app reads the emojis as a story and turns them into a movie poster that takes itself completely seriously.
Type 🚗☕😤💻 and you get CODERAGE: MORNING DRIVE, a techno-thriller where "a disgruntled programmer seeks vengeance through vicious code after a bitter morning commute and coffee snub." Four stars from Digital Reel Dispatch, Spring 2025.
The whole card is built to look like a real one-sheet. Festival laurels along the top, a rating badge, a giant emoji as the key art, a star rating, a critic pull-quote, the title in heavy cinema type, a logline, a release window, and a tiny condensed billing block at the very bottom with studio, director, and cast. The comedy comes from how seriously the format treats four random emojis.
How I built it
I started with the model contract. The prompt casts Nova Pro as a movie marketing creative and forces it to return one strict JSON object: title, tagline, genre, logline, rating, runtime, star count, a critic quote and outlet, director, cast, studio, release window, and two festival laurels. Fixing the schema up front meant the poster layout never has to guess what came back. A small parser strips any stray code fences and pulls out the JSON block, so a chatty response still renders.
The backend is a single Python file using the standard library http server. It serves the page and one endpoint, /api/generate , which takes the emoji string, calls Nova Pro through the Converse API, and returns the poster JSON. Page and API share an origin, so there is no CORS to manage. The emoji input is stripped of newlines and length-capped before the call.
The design was a deliberate departure from my earlier entries. One was a dark pitch-deck, another a bright holographic trading card. This one is a cinema one-sheet: a black film-grain background, a gold metallic title in a heavy condensed face, laurels, a rating badge, and a billing block set in tiny condensed caps like the real thing. It is all CSS and emoji. No image generation.
Different emojis pull the genre in wild directions, which is the fun of it.
AWS services and architecture
- Amazon Bedrock (Nova Pro) generates each film as structured JSON through the Converse API.
Browser (local page)
| POST /api/generate { emojis }
v
Local Python server (http.server + boto3)
| Converse API
v
Amazon Bedrock: Nova Pro
|
v
strict JSON -> rendered movie poster
Two files: a server and a page. Nova Pro writes the film, CSS builds the poster.
What I learned
The tighter the JSON schema, the more the result feels art-directed instead of random. Asking for laurels, a billing block, and a critic outlet by name is what makes the output read like a poster rather than a paragraph.
Emojis turned out to be a great model input. They are compact, ambiguous, and full of implied story, so the model has room to be creative while staying anchored to what you typed. Small input, big output.
I also kept the same build pattern as my other local entry: run it on your own machine and let it call the model live, so there is no public endpoint to abuse. The prompt and the layout do all the work, and the whole thing stays tiny.
Link
Loading article