This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
What if your whole marketing team fit on a laptop, and fin...
For further actions, you may consider blocking this person and/or reporting abuse
55 seconds for the whole pipeline feels suspiciously fast, video gen alone usually burns through that. Does the clock include the video ads or is that copy and images only? Asking because I glue my own outreach tools together and the handoffs are exactly where my time goes
Hi Ivan,
Yes, the 55 seconds includes the video. It's the whole kit: copy, studio photos, the deck and the 10-second video.
It's quick because the video isn't AI-generated. I cut the product out of the photo, put it on a backdrop, and add camera moves and text with FFmpeg. The copy and the visuals also get made at the same time, so nothing waits on a handoff.
That was one run on my MacBook Pro (M3 Pro) on the slower Best quality mode. Fast mode is about 18 seconds. It's a single run, not an averaged benchmark, so your results may differ a little.
Thanks for asking! I'll update the post to make it clearer that the video is built from the photo, not AI-generated.
nice product video, is this with higgsfield?
Thanks @muhammadmurtuzahussain! Hyperframes in claude code terminal to generate the product video + Elevenlabs voiceover. github.com/heygen-com/hyperframes
By far, these are the most Gemma models I’ve seen used so far. Insane llm architecture design
thank you
great project
thank you
Thanks advik for reaching out to us and providing the tool to use. Loved it👍
Thanks!
This lines up with what I've seen running small client-side tools myself: the product data genuinely never leaves the machine, and that's a real win. The part I'd add is that local execution moves the trust boundary rather than removing it — the thing you now have to trust is the model files and the dependency tree you pulled down to run them. A tampered model checkpoint or a compromised transitive package runs with the same access to those product photos as the pipeline itself.
None of that contradicts your claim — the photos stay off other people's servers, full stop. It's just that "runs on your own computer" quietly swaps "trust their server" for "trust your supply chain", and that second one is the part most local-first pitches skip. Curious whether you pin model hashes or lock dependency versions for exactly this reason.
cool, ngl u cooked fr :)
hahahah thanks😊 @amyy22!!