Tho PRO

azhan77168's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

posted an update 3 days ago

view post

Post

102

I built an AI video tool for a very specific moment: “I love The Smiths.” 🎧

Upload two portraits, assign the Listener and Speaker roles, and turn them into an elevator meet-cute inspired by (500) Days of Summer.

No prompts to write. Generate in your browser and download the MP4.

https://ilovethesmithsai.online/

Independent fan-made project · Consenting adults only · Paid credits, no subscription

Curious to hear what you think of the results!

See translation

Reply

posted an update 7 days ago

view post

Post

81

I’m building MemeMorph, a photo-to-video project exploring the Bicep Heart AI effect: turning a portrait into a short sequence where a person flexes one arm and uses the opposite hand to complete a heart shape.

What interests me is the gap between generating plausible movement and generating a specific, readable gesture. For this effect, a convincing result needs more than a good final frame—the face, hands, and arm positions should remain coherent throughout the sequence.

I’m exploring how to evaluate these clips across identity preservation, gesture accuracy, and temporal consistency.

Explore the Bicep Heart AI studio preview:
https://mememorph.app/

For people working on image-to-video: which failure do you notice first—identity drift, distorted hands, or an unclear final pose?

See translation

Reply