July 8th Office Hour Recap: Model Training, Budget-Friendly Workflow, Product Adverts

Hey friends just wrapped up todays Office Hour Call. You can check out the full video here: https://youtu.be/UCITY9cD4fE

And here's the TLDR of what we talked about:

Character Model Training vs. Reference Images

Omar shared experiences with model training, noting inconsistencies in character details and clothing
Brandon advocated for using reference images in tools like Midjourney's Omni Reference instead of model training
Omni Reference demo showed strong character consistency with a single image input
Model training for character consistency is often unnecessary; using reference images in tools like Midjourney or Flux is more efficient

Budget-Friendly Workflow for AI Video Creation

Recommended stack: Midjourney ($10/mo) + Google Flow/VO3 ($20/mo) = $30/month total
This setup allows for creating 1-2 minute-long concept videos, including iterations
Google Flow (VO3) provides 1,000 credits, roughly equating to 50 eight-second video clips

Product Advertisements

Combining AI-generated characters with product images can be a challenge; green screen techniques and post-processing may be needed to get the product images looking 10/10 consistent

2 comments