Original episode by Singh; creator reports Claude Opus.
What If You Could See Everyone's Number
A 29-second "What if you could see everyone's ...?" episode in flat vector style. A glowing number floats above every head, and the counters roll as the story turns: the rich-looking person has the worst number, the poor-looking one the best, and a twist at the end. Every frame is drawn in code on an HTML canvas, and the lo-fi beat and sound effects are made in numpy. No image, video or voice model.
How it was made
No additional details were supplied.
Models
- Claude Opus (creator-reported) · Anthropic · execution
About
A 29-second "What if you could see everyone's ...?" episode in flat vector style. A glowing number floats above every head, and the counters roll as the story turns: the rich-looking person has the worst number, the poor-looking one the best, and a twist at the end. Every frame is drawn in code on an HTML canvas, and the lo-fi beat and sound effects are made in numpy. No image, video or voice model.
Try it
Copy the prompt and run it with Claude Opus in a coding agent with a shell. It asks for your number idea (or pitches three), the twist and the length, and shows you the shot list before it builds anything. It costs nothing beyond the agent: no voice or paid service. The preview is the creator's actual output, not a Possible-generated substitute.
Creator and prompt
By Singh. See the original post. The copyable text is the prompt as posted, from <inputs> to </build>, with only the post's introduction removed.
What to know
The creator says it cost $0 and that every frame is code. We haven't run the prompt ourselves or inspected his source. The style follows the popular "What If You Could See" channels; this result isn't affiliated with any of them, and the prompt bans their logos.
Model provenance: Claude Opus is reported by the creator, with no version given. Possible verified the source pairing and media presentation, not an independent generation. The preview is the creator's video, copied without re-encoding. Creator media is excluded from the repository's MIT license; the original post remains linked.
Exact prompt
Hidden characters, if present, appear as Unicode escapes<inputs>
Ask me for one "What if you could see everyone's ___?" idea (a number that floats above every head: bank balance, IQ, lies told today, screen time, days left), or pitch 3 and let me pick. Ask me for the twist (the last number that flips the whole story) and the length (25-35 s). Then write the shot list with every caption and every number. Show it to me before you build anything.
</inputs>
<direction>
A 1920x1080 flat vector "what if" episode in the style of the big "What If You Could See" channels.
Characters: a pure black round head with white eyes and a white mouth line, a coloured hoodie with drawstrings and a pocket, dark pants, white sneakers, black hands. Faces change per line: neutral, happy, sad, shock (big eyes + open mouth + brows), smirk (half-closed eyes). Everyone walks, breathes or gestures, nobody stands still.
Above each head a dark rounded pill with a glowing outline and the number in heavy white type: green for good, amber for normal, red for bad. Pills pop in with a 0.3 s overshoot, numbers roll up or down like a counter.
Places are full flat scenes: a bedroom with morning light, a street of shop fronts with signs (CAFE, BANK, PAWN SHOP), a luxury street with a red sports car, a park bench with pigeons. Flat colours, one darker shade for form, no outlines.
A new shot every 3-4 s, a hard cut with a white flash on every new idea, a splat wipe after the title, the camera always drifts 4-12%.
Captions: one line per shot in a dark rounded box at the bottom, the money word coloured.
A red X (two strokes, 0.11 s each) stamps every lie the viewer believed.
Banned: outlines, stock icons, emoji, a still frame longer than 1 s, any channel's logo.
</direction>
<structure>
0-2.7 s the title word by word: "WHAT IF YOU COULD SEE" white, "EVERYONE'S" amber, "[THE NUMBER]?" green, over the blurred street → 2.7-6.3 s day 1, you wake up and your number pops above your head → 6.3-10.3 s the street, everyone has one, 6 pills pop one by one → 10.3-13.9 s the person who looks rich has the worst number, red X → 13.9-17.6 s the person who looks poor has the best number, you react in the foreground → 17.6-21.6 s you do the everyday thing that drops your number, it rolls down on screen → 21.6-25.6 s the twist ties back to the rich-looking guy → 25.6 s the end card asks the viewer a question + "FOLLOW FOR PART 2".
</structure>
<build>
1. Tools: Python, Playwright with Chromium, FFmpeg, numpy. Install what's missing.
2. One HTML file with one 1920x1080 canvas. render(t) draws the exact frame for time t from scratch every call (a pure function of time, no timers), so the render is deterministic.
3. Pre-render anything heavy once (a blurred title background = draw small, scale up).
4. Draw 4 key frames as stills and check them before the full render.
5. Render: Playwright seeks t = frame/30 for every frame, saves JPEG frames, ffmpeg -c:v libx264 -crf 18 -pix_fmt yuv420p.
6. Sound in numpy: a 96 BPM lo-fi beat (kick, snare, hats, Rhodes chords, bass), a pop + ding on every pill, counter ticks while numbers roll, a whoosh on every cut, a thud on the red X, a cash-register ding on the big number, a falling tone when your number drops.
7. Mux video + sound, check the final file plays, send it to me.
</build>What you need
- Coding agent with a shell; Python 3, Playwright with Chromium, FFmpeg and numpy (the prompt installs what's missing).
- No voice and no paid services: the sound is made in code.
Download the result
Technical details and provenance
- Provider
- Anthropic
- Model
- Claude Opus (creator-reported)
- Published
- Publication
- community
- Source
- Original source ↗