Sayvi
Turns one photo into a talking AI avatar video and publishes it to TikTok, Instagram and Facebook.
- AI video platform
- Products
- Live at
- say-vi.com
What it does
Upload a photo or describe a character. Sayvi builds a reusable avatar, drafts the script, renders a short vertical video and publishes it to TikTok, Instagram and Facebook. Consent and AI-disclosure gates are built into the flow.
How it’s built
Next.js 16 and React 19 over a raw Postgres data layer with HMAC-signed sessions. HeyGen handles avatar generation and lip-sync, fal.ai renders motion and non-human characters, and ElevenLabs supplies multilingual voice. Billing runs on Stripe across three tiers.
The parts nobody sees
The render pipeline is asynchronous and reconciles its own stuck jobs. Social tokens are stored encrypted, per user. A mock-first architecture lets the whole product run end to end with zero API keys, switching to live one provider at a time. A documented public API ships with bearer-key auth and plan-aware rate limits.
