How I Built a Free AI Video Generator

#ai #saas #showdev #startup

I'm a mobile dev running a small studio in Israel. We ship React Native and Flutter apps — eSIM, delivery, VPN, VoIP. A few months ago I started building CreatorAI — an AI video editor that actually lets people try it for free.

The Problem

Every AI video tool charges $20-40/month before you even know if it works. Runway, Pika, Kling — all great tech, all paywalled. Small creators and businesses can't justify that cost to test an idea.

What I Built

CreatorAI — a free video editor + AI video generator.

Editor is 100% free — trim, captions, music, transitions, filters, export in HD
10 free AI credits on signup — no card, no trial that auto-charges
Text-to-video — describe a scene, get a clip
Image-to-video — upload a photo, animate it
Web + iOS + Android

Tech Stack

Frontend: React Native (Expo) for mobile, Next.js for web
AI Models: Kling v2.1, Seedance Lite via fal.ai
Backend: Node.js + Supabase
Payments: RevenueCat + Paddle
Hosting: Vercel (web), EAS (mobile)

How AI Generation Works

User submits a prompt → backend queues it → fal.ai processes with Kling or Seedance → video URL returned → user downloads.

Fallback chain: if primary model fails, it tries the next one. Auto-refund credits on failure.

Credit packs:
| Credits | Price |
|---------|-------|
| 10 | Free (signup) |
| 100 | $10 |
| 200 | $15 |
| 300 | $20 |

What I Learned

Free tier drives everything. 5% of free users convert to paid. But without free, nobody tries it.
Model quality matters more than speed. Switched from PixVerse to Kling — generation takes longer but output quality jumped 3x. Nobody complained.
Mobile-first wins. 70% of our users are on phone. Web is for discovery, app is for retention.
Don't build your own model. Using fal.ai APIs saved months of infra work. Focus on UX, not GPUs.