H3 Omni Ref Iteration Review

TASK: Create a 5-second portrait video of the woman from <Picture 1>.
USE: Transfer only the head turns, facial expressions, and speaking-performance timing from <Video 1>.
KEEP: Her face, dark curly hair, white fur coat, white trousers, and the bright windowed setting from <Picture 1>.
EXCLUDE: The man, beard, dark clothing, room, camera look, and audio from <Video 1>. The delivered video is silent.

subject_definitions:
<Subject 1> is the woman from <Picture 1>, including her face, dark curly hair, white fur coat, white trousers, and bright windowed setting.
<Video 1> supplies only head turns, facial expressions, and speaking-performance timing.

summary:
[reference generation] Create a 5-second portrait of <Subject 1> performing the facial and head motion from <Video 1>.

retention_analysis:
<Subject 1>: fully_preserved - retain identity, clothing, and the windowed setting from <Picture 1>.
<Video 1>: attribute_transfer - transfer only performance motion; exclude source identity, clothing, room, camera look, and audio.

detailed_description:
[Shot 1] Keep <Subject 1> in the bright windowed setting from <Picture 1>. Apply only the head movement and facial-expression sequence from <Video 1>.

overall_soundscape:
Silent delivery.

non_diegetic_music:
N/A

TASK: Create a 12-second video of the man from <Picture 1> performing the speaking and mug-holding motion from <Video 1>.
USE: Transfer only head movement, facial timing, hand motion, and mug handling from <Video 1>.
KEEP: The man's face, grey striped shirt, dark trousers, and plain white-wall studio setting from <Picture 1>.
EXCLUDE: The woman, blonde hair, kitchen, source clothing, source camera look, and audio from <Video 1>. The delivered video is silent.

subject_definitions:
<Subject 1> is the man from <Picture 1>, including his face, grey striped shirt, dark trousers, and plain white-wall studio setting.
<Video 1> supplies only speaking, head movement, hand motion, and mug handling.

summary:
[reference generation] Create a 12-second video of <Subject 1> performing the motion from <Video 1>.

retention_analysis:
<Subject 1>: fully_preserved - retain identity, clothing, and white-wall setting from <Picture 1>.
<Video 1>: attribute_transfer - transfer only performance motion; exclude the woman, kitchen, source clothing, camera look, and audio.

detailed_description:
[Shot 1] Keep <Subject 1> against the plain white wall from <Picture 1>. Apply only the speaking and mug-handling performance from <Video 1>.

overall_soundscape:
Silent delivery.

non_diegetic_music:
N/A

TASK: In the 8-second <Video 1>, replace only the visible speaker with the woman from <Picture 1>.
USE: <Picture 1> only for the replacement speaker's identity and clothing.
KEEP: The source room, animal pictures, camera, framing, gestures, expression timing, and full timeline from <Video 1>.
EXCLUDE: The original speaker's visible identity. Do not change any other visual content. <Audio 1> will be reused unchanged after generation.

subject_definitions:
<Subject 1> is the replacement speaker from <Picture 1>.
<Video 1> is the source clip to edit.
<Audio 1> is the synchronized source soundtrack reused after generation.

summary:
[video editing + reference generation] In the 8-second <Video 1>, replace only the visible speaker with <Subject 1>.

retention_analysis:
<Subject 1>: fully_preserved - use the identity and clothing from <Picture 1>.
<Video 1>: partially_preserved - retain room, animal pictures, camera, framing, gestures, expression timing, and full timeline; replace only visible speaker identity.
<Audio 1>: fully_copy - reuse as the complete synchronized soundtrack after generation.

detailed_description:
[Shot 1] Match <Video 1> from its first frame through its last frame. Replace only the speaker's visible identity with <Subject 1>. Do not alter any other visual content.

overall_soundscape:
Reuse <Audio 1> unchanged and synchronized after generation.

non_diegetic_music:
N/A

TASK: In the 8-second <Video 1>, replace only the visible speaker with the woman from <Picture 1>.
USE: <Picture 1> only for the replacement speaker's identity and clothing.
KEEP: The source room, animal pictures, camera, framing, gestures, expression timing, and full timeline from <Video 1>.
EXCLUDE: The original speaker's visible identity. Do not change any other visual content. <Audio 1> will be reused unchanged after generation.

subject_definitions:
<Subject 1> is the replacement speaker from <Picture 1>.
<Video 1> is the source clip to edit.
<Audio 1> is the synchronized source soundtrack reused after generation.

summary:
[video editing + reference generation] In the 8-second <Video 1>, replace only the visible speaker with <Subject 1>.

retention_analysis:
<Subject 1>: fully_preserved - use the identity and clothing from <Picture 1>.
<Video 1>: partially_preserved - retain room, animal pictures, camera, framing, gestures, expression timing, and full timeline; replace only visible speaker identity.
<Audio 1>: fully_copy - reuse as the complete synchronized soundtrack after generation.

detailed_description:
[Shot 1] Match <Video 1> from its first frame through its last frame. Replace only the speaker's visible identity with <Subject 1>. Do not alter any other visual content.

overall_soundscape:
Reuse <Audio 1> unchanged and synchronized after generation.

non_diegetic_music:
N/A

Edit <Video 1> only by converting every frame to high-contrast black and white. Preserve the complete source timeline, people, hutch, backyard, action, framing, and camera exactly.

Edit <Video 1> only by applying a crisp balanced contrast grade. Start from the exact first source frame and preserve the complete mouse movement, hand movement, desk, framing, and timing. <Audio 1> will be reused unchanged after generation.