Hi HN,
I’ve been experimenting with automating product video creation using generative models.
The idea was simple:
Can a single product image be transformed into multiple ad-style video variations automatically?
I built a structured pipeline that:
Takes one product image
Generates multiple visual scenes (different backgrounds, angles, contexts)
Converts scenes into short ad clips
Applies different hooks and formats (UGC-style, cinematic, problem-solution)
Under the hood it’s essentially:
Prompt-engineered image models
Video generation models
Scene templating logic
Variation-based rendering system
The goal isn’t perfect cinema — it’s speed and iteration for small sellers who need many creatives to test.
Still early. I’m mostly curious:
Does this feel technically interesting?
Or does this just become noise in a world full of AI tools?
Would love honest feedback.
https://promomotions.com
