Testing H3 Text-to-Video on 'Mr. Robot': Stunning Atmosphere But Inconsistent Likeness

Positive_Writing_883 · reddit · 2026-08-08

A Reddit user tested the H3 text-to-video model (at 768p) to recreate a scene from the live-action show Mr. Robot.

Key Takeaways:

Atmosphere & Lighting: The cold fluorescent lighting and psychological thriller vibe were surprisingly well-captured.
Continuity & Likeness: The main character Elliot's likeness, voice, and framing were convincing and cleaner than expected.
The Miss: H3 failed to accurately reproduce Christian Slater’s likeness, though it got the wardrobe and general look right.

The post includes a highly detailed, 4-shot prompt covering camera movements, character actions, dialogue, and soundscape design, serving as a great reference for multimodal generation workflows.

Original post →

Testing H3 Text-to-Video on 'Mr. Robot': Stunning Atmosphere But Inconsistent Likeness

More from Multimodal