← Back

AI UGC Generator

Generating Revenue-Driving TikTok Content Without Hiring Creators

AI UGC Generator overview

The Problem

UGC creators charge $150-500+ per video. We know this stats because we contacted the influencer ourselves. When you're an indie developer trying to validate whether TikTok works as an acquisition channel, that's a non-starter. You can't justify spending thousands on content before knowing if the format converts, especially that we weren't users of TikTok so it was a complete discovery for us and from what we heard (and after test, through partnership, this is very random).

I needed a way to produce TikTok-style hook videos at near-zero marginal cost but enough volume to test hooks, formats, and audiences without the upfront investment of hiring real creators.

What I Built

We started by searching on our niches which content have went viral and tried to duplicate/adapt them. So we organize this phase in a Linear ticket quickly:

Linear board organizing viral content research by niche

And then we quickly created a Next.js web app with 5 specialized generation pipelines, each producing different content types for app marketing:

TikTok Hook Generator: The primary pipeline. Takes a reference face image and generates 4-8 second selfie-style UGC video clips (reactions, movements) using Google's Veo 3.1. The output looks like someone filming themselves on their front camera. All prompts for the Gemini models are based on a famous library that use the correct format to generate correct images/videos.

Batch UGC Pipeline: Mass production mode. Generates N hook videos, then auto-stitches each with b-roll footage using ffmpeg. Outputs ready-to-post TikTok videos (1080x1920) with round-robin b-roll rotation so each video feels different.

Study Scene Generator: Photorealistic overhead study setup images (for SnapQuiz marketing). Configurable location, lighting, camera angle, and subject-specific props. Uses Gemini image generation with a reference photo for identity consistency.

Carousel Portrait Generator: Character-consistent lifestyle images across different scenes/outfits while preserving the same face. Uses explicit image labeling to force the model to maintain identity.

Timelapse Generator: Cinematic study session videos (time-lapse style) for Instagram Reels and TikTok background content. It was used to promote SnapQuiz mainly.

UGC Generator web interface showing the generation pipelines

Results

The UGC videos generated for SnapQuiz have directly driven revenue - real downloads converting to paid subscriptions, acquired through TikTok content that cost essentially nothing to produce. What would have been thousands in creator fees became a Gemini API bill of a few dollars per batch.

TikTok account metrics showing engagement from AI-generated content

Timeline

2026

Stack

Next.jsVeo 3.1Geminiffmpeg

Responsibilities

  • 5-pipeline content generation system
  • Veo 3.1 integration for UGC video
  • Batch processing and ffmpeg automation
  • TikTok content strategy and testing
Visit project