GETGOYNG

Text in.

Motion post out.

A browser-native motion studio that turns a headline, a link, or a premise into a finished, animated social post — imagery, voiceover, cinematic motion and an MP4.

No video-generation API

No backend

No build step

Stuttgart hackathon · July 2026

THE PROBLEM

Every creator has the same Tuesday problem

40

minutes

The gap between having something to say and having to post it. Not enough time for a motion designer. Not enough time to learn After Effects. So you open a template — and the post looks like a template.

1

Templates read as templates

Every account on the feed is drawing from the same twelve layouts.

2

Motion is the missing craft

Stills are cheap now. Movement is what still looks expensive — and stays out of reach.

3

Video models bill you to guess

Wait minutes for a clip, then pay again to find out you don't like it.

WHO IT'S FOR

The people shipping today's post themselves

1

Solo creators

Posting to Reels, TikTok and Shorts on their own schedule, with no editor and no studio budget.

2

Small brand accounts

One person holding the whole feed, expected to look like a team of five.

3

Newsletter writers

Turning a paragraph they already wrote into the post that carries it to a new audience.

getgoyng is not a scheduler, not an account manager, and not a video generator. It makes the post. You decide when it goes out.

THE INSIGHT

Generation solved the picture. It never solved the motion.

What the market built

• Call a video model with a prompt.
• Wait minutes for a clip to come back.
• Pay again for every version you try.
• Resolution, timing and type are baked in at generate time.

What we built

• A real-time WebGL motion engine, running in the tab.
• Stills become cinematic motion instantly, at any resolution.
• Restyling touches no server and costs no API call.
• Type, timing and camera stay editable to the last frame.

Generation is used only where it belongs — planning, imagery and voice. The movement is ours.

HOW IT WORKS

Paste something. Get a finished post.

01

Paste

A headline, an article URL, a story premise or a product description — then pick a mode: News, Story or Product.

→

02

Generate

Claude plans the slides and writes the narration. fal renders the imagery and a per-slide voiceover. About thirty seconds.

→

03

Compose & export

It opens straight into the studio. Change preset, palette or timing live, then export an MP4 with the voice already in it.

News headline or article URL → Focus Reel, urgent read, photoreal imagery

Story a premise in a few lines → Kaleidoscope, warm narrator, illustrated

Product a URL or description → Coverflow, confident brand voice, clean shots

ARCHITECTURE

Three layers, one seam

Layer 1 — Generate

compose.html

text or URL + mode → Claude plan → fal imagery → fal voiceover

↓ project bundle · .goyng.json

Layer 2 — Compose

studio.html

WebGL motion engine · presets · camera · timing · card styling

↓ live studio instance

Layer 3 — Export

export modules

MP4 · GIF · HTML embed · After Effects script · voiceover muxed in

The contract

• Every layer runs, and demos, on its own.
• Layer 2 makes no network calls and holds no API keys.
• Layer 1 never touches Three.js.
• A bundle made once re-loads forever, offline.
• The seam is a file — so any layer can be replaced without touching the others.

THE DIFFERENTIATOR

Restyling is free

Nothing re-renders on a server. Nothing costs another call. Iterating toward a post that feels like yours costs seconds, not minutes and not money.

Video-generation pipeline

getgoyng

Change the preset

re-render the clip · minutes

instant

Change the palette

another API call

instant · no call

Retime the deck

regenerate from scratch

a slider drag

Fix a typo

the type is pixels — start over

live text, any frame

Most tools make you wait minutes to find out you don't like it.

THE DETAIL THAT SHOWS

The text is not baked into the picture

Most tools composite your headline into the generated image. From that moment the type is pixels: fix a typo and you regenerate the frame, export larger and the edges soften.

getgoyng keeps live type all the way through. Titles are baked to texture at export resolution, so the same deck is razor sharp at 720p, at 1080p, and at whatever comes next.

Edit a headline mid-playback and the change lands on a moving frame.

Aa

720p

Aa

1080p

Aa

4K

Same deck. Same type. No resampling, no soft edges, no regeneration.

THE ENGINE

Ten presets on one tone axis

Stack Cascade

cards peel off a deck

Vertical Stack

a continuous feed

Halo Arc

the hero layout

Coverflow

a flanked centre hero

Orbit Ring

a tilted ring

Focus Reel

the news default

Portal Zoom

a recursive Droste zoom

Slide Stack

diagonal drift

Kaleidoscope

a blooming mandala

Spiral Ring

fogged at depth

SERIOUS

PLAYFUL

one slider moves easing, timing and camera together

SYNC

One clip per slide. Sync is true by construction.

The voiceover is generated as one clip per slide, and each slide's screen time is that clip's measured duration. There is no alignment pass, so there is nothing that can drift.

The script is narration, not a screen reader. The titles are already on screen — reading them back is the fastest way to make a post feel machine-made, so the voice carries what the titles cannot.

Export duration is pre-filled from the voiceover, so narration is never silently cut.

TIMELINE

Slide 1

3.4s

Slide 2

2.8s

Slide 3

4.1s

Slide 4

2.4s

Slide 5

2.9s

slide screen time = clip duration

ECONOMICS

Iterating costs nothing after the first pass

$0.01

per image at the default quality tier — the console exposes the tier, so cost is a choice, not a surprise.

0

API calls to change a preset, a palette, a camera move or the timing of the whole deck.

1

generation pass per project. Everything after it is local compute you already own.

Chart is structural, not measured: it counts billed generation calls, not dollars.

OUTPUT

Four ways out, no infrastructure

MP4

Deterministic, frame-by-frame through WebCodecs at 30fps, supersampled then downscaled. The voiceover is muxed in as an AAC track.

GIF

Exactly looping at 20fps, quantised per frame. Rendered from the timeline, never screen-captured.

HTML embed

A standalone autoplay, loop, muted, playsinline document — copied to the clipboard and downloaded.

After Effects

An ExtendScript that rebuilds the deck as real AE layers, for when someone wants to take it further.

Static files. Vanilla ES modules and an import map, served as-is. No bundler, no npm install, no backend for the studio — the only server in the project holds the API keys for Layer 1.

STATUS

What is running today

3

layers, shipped

10

motion presets

297

tests, 9 suites

4

export formats

✓

End to end — Paste a headline in the console and an MP4 with voiceover comes out the other side.

✓

Per-slide voice — Narration split by slide, each slide timed from its own measured clip.

✓

Live text motion — Word, line and block level animation with emphasis tiers, driven by the tone axis.

✓

Offline-safe demo — Bundles and assets persist in IndexedDB; a saved project replays with no network at all.

Next: more presets, brand kits, and a small server-side render path for teams that want it.

GETGOYNG

Text in.

Motion post out.

Forty minutes becomes thirty seconds — and the post still looks like someone made it on purpose.

1

Watch it — A headline becomes a finished, narrated post in one take.

2

Break it — Change every preset mid-playback. Nothing waits on a server.

3

Ship it — Static files on any host. It goes live the day you decide.