SaaS product

Clipia

Video shorts generator — from voice pitch to rendered short-form with AI on local GPU.

Active Closed beta · Rendering and inference on local GPU · Live payment (Stripe live)

The problem

Content creators for social become pipeline operators — ideation, scripting, captures, editing, captions, publishing. A person with half a dozen ideas per day faces weeks of execution. Cloud video services get expensive quickly; off-the-shelf solutions charge per render and don’t adapt workflow.

The engine

Clipia turns an idea into a publish-ready short video — the pipeline executes:

  1. Scripting — Claude structures narrative, scene sequence, dialogue.
  2. Imagery — SDXL generates frames with aesthetic fidelity and visual continuity.
  3. Narration — XTTS v2 synthesizes the creator’s voice in Brazilian Portuguese (voice cloning with reference).
  4. Captions — Whisper transcribes and timecodes with character-level precision.
  5. Composition — FFmpeg and MoviePy orchestrate audio, video, captions, effects, fades.

Infrastructure

FastAPI backend + Celery coordinates the work queues. Image rendering, voice synthesis and composition run on owned GPU (RTX 3090) — no per-render cloud cost; Redis holds intermediates between stages.

Operational evidence

E2E pipeline verified script → imagery → narration → captions → video
Stripe live billing model operational closed beta with live payment
Local GPU rendering without cloud RTX 3090 reduces cost per render

On screen

Clipia — product page: AI short-video generation
clipia.com.br — the product live, in closed beta with active payments.

My role

End-to-end architecture conception, Celery orchestration, GPU rendering engineering, operations and quality tuning. Independent product, commercialized under closed beta.

Clipia is an independent product. Brands and names cited belong to their respective owners.

Want to build something like this?

No strings attached: tell me the context and I'll tell you what can be proven first.

Start a conversation