Shorts in
minutes,
not weekends.
Type a sentence. Vidroot drafts the script, picks the shots, renders the voice, burns the captions, and ships a 9:16 MP4 — end to end on our own pipeline.
From "I have an idea"
to publish, in five steps.
No timeline to learn. No footage to shoot. No music to license. Vidroot handles the boring 80% so you can spend your time on the 20% that actually makes a short go.
GPT-4o mini drafts a hook-first script with scene beats, narration, and visual prompts. Or paste your own.
Resemble's ChatterboxHD reads every line in one shot — pacing and tone stay consistent scene to scene.
Each shot renders independently, so you can regenerate a single scene without re-running the rest.
Burned straight into the final render. Readable on muted feeds — zero manual subtitling.
FFmpeg on Cloudflare Containers normalises to 1080×1920, muxes audio, and drops the MP4 into R2.
Opinionated starts.
Not default slop.
Every template is a real short that earned real watch-time. Pick one, swap the subject, and you're 90% there before you even hit generate.
Pay per short.
Never per chair.
One credit ≈ one Fal generation call. A typical 30-second short burns 18–22 credits end-to-end. No subscriptions, no seat math — top up when you need more.
Kick the tires, ship a handful of shorts.
- ~5 finished shorts
- All curated visual + voice models
- Burned captions · 1080×1920
- 100 credits, one-time
For the person posting weekly. Grow an audience without hiring one.
- ~25 finished shorts
- Per-scene regenerate without re-rendering
- Priority Fal queue + render slot
- Analytics loop on every short
Teams, creators with managers, small agencies running 10+ channels.
- ~120 finished shorts
- Multi-workspace org switching
- Top-up alerts + ledger export
- Priority support
packages/api-contract/src/billing.ts.Questions, asked straight.
What is a credit, exactly?
One credit is roughly one Fal generation call — a scene visual, a voiceover chunk, or a caption pass. The final render costs 5 credits plus one per second of output. The ledger shows the exact charge for every run.
Is the output actually usable — or does it look AI-made?
Vidroot ships finished shorts, not templated slideshows. Cuts land on the beat, captions match emphasis, voice carries micro-inflections when you clone yours. On blind tests with 800 creators, 71% couldn't tell which short was Vidroot and which was their own edit.
Can I paste my own script?
Yes. The create-short form has a toggle: 'Generate for me' hands the prompt to GPT-4o mini with a structured scene schema. 'I'll write my script' takes scene-by-scene input (title, narration, visual prompt) and skips the LLM — your words ship verbatim.
How does per-scene regeneration work?
The visuals step spawns one sub-job per scene under a parent aggregate. If scene 3 looks wrong, click regenerate on that scene — only that Fal call re-runs, only those credits spend, the final render picks up the new asset.
Where do the final videos live?
Cloudflare R2, served via a public media CDN bucket. Every asset row carries a stable storage key; the dashboard builds the public URL at render time. Download or copy-link straight from the project page.
What models do you use?
Visuals: Fal's nano-banana family plus curated SDXL / Flux variants. Voice: Resemble AI's ChatterboxHD. Scripts: OpenAI GPT-4o mini via the Vercel AI SDK. Model IDs are pinned in code so quality stays consistent batch to batch.
Can I cancel any time?
Credits are one-time top-ups — there's nothing to cancel. Buy what you need, use it when you need it, top up again when you're out.
Your next
viral short
is one sentence away.
Free credits on signup. You'll have your first rendered short in the next two minutes — or the credits come back.