DEV Community

Cover image for I Built a Tool to Generate Video Ads with AI
jester
jester

Posted on

I Built a Tool to Generate Video Ads with AI

tags: [ai, video, automation, startup]

Creating a 30-second video ad used to require six manual steps: scripting, image sourcing, voiceover generation, converting images to video clips, syncing audio, and editing in CapCut. Each step involved different tools and constant context switching.

So I built Ad Forge - an AI-powered tool that streamlines video advertisement creation.

How It Works: 7 Stages

  1. Sketch: Describe your ad concept, target audience, and tone
  2. Scenes: Gemini breaks sketches into individual scenes with descriptions and mood
  3. Style: AI generates visual consistency guidelines including color palette and character descriptions
  4. Images: Fal.ai generates images for each scene, maintaining consistency through reference images
  5. Videos: Images become video clips with camera movements (pan, zoom, dolly)
  6. Audio: Text-to-speech generates voiceover with customizable character voices
  7. Merge: Combines video and audio with optional lip-sync using FFmpeg

Tech Stack

  • Next.js 16 with React 19
  • Google Gemini for script generation
  • Fal.ai for image generation and video conversion
  • OpenAI for text generation
  • FFmpeg WebAssembly for in-browser video processing

Key Design Decisions

Campaign Persistence: LocalStorage automatically saves progress, allowing users to resume later.

Reference Images: Users can reference previous scenes and character portraits as style anchors for consistency.

Stage-Based Workflow: Each stage produces reviewable output before advancing, enabling iteration without complete restarts.

Results

Production time for a video ad decreased from a full day to approximately 30 minutes of active work.

https://jovweb.dev/blog/building-ad-forge

Top comments (0)