Summary
Hasan Aboul Hasan documents 15 experiments and 16 different tools tested over more than $200 in spending to find the most realistic way to create an AI clone of himself. After ruling out tools like Google’s Omni (strong visuals but no custom voice on lower tiers) and various lip-sync-over-video approaches that drifted out of sync, he narrows the field down to two winners: Hen and Seedance 2.5.
The video breaks down exactly when to use each: Hen for long talking-head videos at around $4 a minute, built from a single high-quality reference recording with even lighting and visible teeth, and Seedance 2.5 for placing yourself in real scenes at about $14 a minute with a 3-second clip limit. He shares a detailed five-step recipe for Seedance, covering face verification through ModelArc’s API, using multiple reference photos in different expressions, attaching a real audio recording to control acting and lip movement, and using a short reference video to accurately rebuild a real room.
A recurring problem across every tool tested is accent replication, which Hasan solves by recording his own voiceover and injecting it into the generated video rather than relying on voice cloning. This is a practical, tool-by-tool guide for creators or developers experimenting with digital avatars and AI-generated video clones.
📺 Source: Hasan Aboul Hasan · Published September 19, 2026
🏷️ Format: Comparison
