iArt.ai
AI motion graphics software — the After Effects alternative anyone can use. Create animations and explainer videos from a text prompt in minutes.
Turn one horizontal clip into 9:16, 1:1 and 16:9 cuts. Rendered locally in your browser; your footage is never uploaded.
MediaTypeScript · JavaScriptGitHub @abaskabatoOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
No updates detected yet. GitHub releases will appear here.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
Flip Cast : Open source 3D flip phone in your browser with thousands of live radio stations and podcasts.
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
Turn one horizontal master into platform-ready 9:16, 1:1 and 16:9 cuts — all
rendered in your browser.
The differentiator, stated up front: your footage never leaves your device.
There is no upload, no render queue, and no cloud transcoder. FFmpeg runs as
WebAssembly in the page, so the free tier costs us effectively nothing, works
offline once loaded, and cannot leak your source footage.Read the full README ↗Turn one horizontal master into platform-ready 9:16, 1:1 and 16:9 cuts — all rendered in your browser.
The differentiator, stated up front: your footage never leaves your device. There is no upload, no render queue, and no cloud transcoder. FFmpeg runs as WebAssembly in the page, so the free tier costs us effectively nothing, works offline once loaded, and cannot leak your source footage.
Cloud clipping tools upload your video to a server queue. That model needs render compute, object storage, upload bandwidth and upload time before you see a single frame. We rejected it: a browser already has a full video encoder built in, and FFmpeg WASM removes every one of those dependencies.
The honest trade-off: WASM encoding runs roughly 8x slower than realtime on a short clip, so a 60-second master can take several minutes. We show real progress and say so in the UI rather than hiding it. The multi-threaded core is used automatically when the browser is cross-origin isolated (the app ships the required COOP/COEP headers) and the device has 4+ cores.
job_outputs row.src/lib/video/geometry.ts
emits an FFmpeg crop= expression that derives geometry from the stream
itself, so there is no ffprobe round-trip and no chance of cropping against a
stale dimension assumption. Verified against real ffmpeg across 4K, portrait,
square and odd sizes.@ffmpeg/ffmpeg) — cores self-hosted in public/ffmpegsrc/lib/auth.ts) + @daveyplate/better-auth-uidocker compose up -d locally)├── drizzle/ # generated SQL migrations
├── scripts/
│ ├── sync-ffmpeg-core.mjs # copies wasm cores + ESM wrapper into public/
│ ├── verify-geometry.mjs # crop math vs real ffmpeg
│ ├── verify-expr.mjs # dimension-free filter expressions
│ ├── verify-probe.mjs # container parser vs ffmpeg metadata
│ ├── verify-render.mjs # real Chromium render, verified by ffmpeg
│ └── verify-e2e.mjs # full signup -> render -> quota flow
├── public/ffmpeg/ # ~63 MB, gitignored, regenerated on install
└── src/
├── db/schema.ts # user/session/account + video_jobs + job_outputs
├── lib/auth.ts # Better Auth server instance
├── lib/quotas.ts # tier limits, max sizes, formatting
├── lib/usage.ts # atomic quota reservation / refund
├── lib/video/geometry.ts # pure crop math (no Node, no DOM)
├── lib/video/ffmpeg-client.ts# render engine
├── lib/video/ffmpeg-loader.ts# unbundled wrapper loader (see note below)
├── lib/video/probe.ts # MP4/MOV container metadata parser
└── components/flipcast/ # dashboard + auth panel
@ffmpeg/ffmpeg's worker loads the WASM core with await import(coreURL).
Webpack statically rewrites that into a build-time module lookup, which cannot
resolve a public path (Cannot find module '/ffmpeg/0.12.10/ffmpeg-core.js') or
a blob URL (Cannot find module 'blob:...'). There is no build-time warning; the
render simply fails at runtime.
scripts/sync-ffmpeg-core.mjs therefore also copies the wrapper's ESM build to
public/ffmpeg/lib, and ffmpeg-loader.ts imports it through an eval-scoped
import() that the bundler leaves alone. public/ is excluded from
tsconfig.json so the vendored .d.ts files are not type-checked.
npm install # postinstall syncs the wasm cores into public/
docker compose up -d # local Postgres
cp .env.example .env.local # set DATABASE_URL + BETTER_AUTH_SECRET
npm run db:migrate
npm run dev # http://localhost:3000
npm run verify:geometry # crop math, 9 cases
npm run verify:expr # filter expressions, 21 cases
npm run verify:probe # container parsing vs ffmpeg metadata
npm run verify:render # real browser render, output checked by ffmpeg
npm run verify:e2e # full flow (needs a server on :3100)
verify:render and verify:e2e need ffmpeg-static, whose postinstall script
npm 12 blocks by default. If the binary is missing:
cd node_modules/ffmpeg-static && node install.js
MIT