H3 Omni Ref Iteration Review

Create an 8-second plain studio portrait of the woman from <Picture 1>. Transfer only the head motion, facial expressions, and speaking performance timing from <Video 1>. Preserve her identity and clothing. Do not copy the source performer, room, camera, or audio.

subject_definitions:
<Subject 1> is the target woman from <Picture 1>.
<Video 1> supplies only head motion, facial expressions, and speaking-performance timing.

summary:
[reference generation] Create a 8-second portrait of <Subject 1> performing the facial and head motion from <Video 1>.

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

detailed_description:
[Shot 1] One continuous medium portrait on a plain neutral studio background. <Subject 1> follows the head movement, facial-expression sequence, and speaking cadence from <Video 1>. Do not copy the source performer or setting.

overall_soundscape:
Quiet neutral room tone. Do not use audio from <Video 1>.

non_diegetic_music:
N/A

Create a 12-second neutral indoor video of the man from <Picture 1>. Transfer only the speaking, head movement, and mug-holding performance from <Video 1>. Preserve the man's identity and clothing. Do not copy the source performer, kitchen, camera, visual style, or audio.

subject_definitions:
<Subject 1> is the target man from <Picture 1>.
<Video 1> supplies only direct-to-camera speaking, head movement, and mug-holding performance.

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

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

detailed_description:
[Shot 1] One continuous medium shot in a simple neutral room. <Subject 1> looks toward the camera and follows the speaking, head movement, and mug-handling timing from <Video 1>. Do not copy the source performer or kitchen.

overall_soundscape:
Quiet neutral room tone. Do not use audio from <Video 1>.

non_diegetic_music:
N/A

Edit <Video 1> only by applying a high-contrast black-and-white grade. Keep every subject, action, object, background, camera frame, and timing unchanged.

subject_definitions:
<Video 1> is the source clip to edit.

summary:
[video editing] The target is a 10-second edited version of <Video 1>. Apply only a high-contrast black-and-white grade.

retention_analysis:
<Video 1>: partially_preserved - preserve all subjects, objects, action, background, framing, and timing; change only color and contrast.

detailed_description:
[Shot 1] Match <Video 1> frame for frame. Remove all color and apply strong but readable monochrome contrast. Do not add, remove, replace, relight, recamera, or retime anything.

overall_soundscape:
N/A

non_diegetic_music:
N/A

Edit <Video 1> only by applying a crisp, balanced contrast grade with clean highlights and readable shadow detail. Preserve the mouse, hand movement, framing, and timing exactly. Reuse <Audio 1> as the synchronized source soundtrack.

subject_definitions:
<Video 1> is the source clip to edit.
<Audio 1> is the synchronized source soundtrack.

summary:
[video editing + audio reuse] The target is a 5-second edited version of <Video 1>. Apply only a crisp, balanced contrast grade and reuse <Audio 1>.

retention_analysis:
<Video 1>: partially_preserved - preserve the mouse, hand movement, desk, framing, and timing; change only the color grade.
<Audio 1>: fully_copy - reuse as the complete synchronized soundtrack.

detailed_description:
[Shot 1] Match <Video 1> frame for frame. Add clean highlights and readable shadow contrast without crushing blacks, changing the lighting setup, or replacing any content.

overall_soundscape:
Reuse <Audio 1> unchanged and synchronized.

non_diegetic_music:
N/A

In <Video 1>, replace only the visible speaker with the woman from <Picture 1>. Preserve the source gestures, framing, room, and timing. Reuse <Audio 1> as the synchronized source dialogue. Do not change anything else.

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 dialogue and room sound.

summary:
[video editing + reference generation + audio reuse] The target is a 8-second edited version of <Video 1>. Replace only the visible speaker with <Subject 1>.

retention_analysis:
<Subject 1>: fully_preserved - use the identity from <Picture 1>.
<Video 1>: partially_preserved - preserve gestures, framing, room, timing, and all non-identity content.
<Audio 1>: fully_copy - reuse as the complete synchronized soundtrack.

detailed_description:
[Shot 1] Match <Video 1> frame for frame. Replace only the speaker's visible identity with <Subject 1> while retaining the source pose, gestures, expression timing, background, and camera. Do not alter any other content.

overall_soundscape:
Reuse <Audio 1> unchanged and synchronized.

non_diegetic_music:
N/A