A human point
of view.

HITL character expression guide

Version 1.1.0. 42 individually regenerated transparent PNGs: 21 per host.

The original sheets remain unchanged. These derivatives were regenerated at the user's request. The files are packaged for review and reuse; likeness remains subject to host review.

Expression selection

Within illustrated Oscar reflective opinion/explainer beats, use oscar.front.thinking by default, per his 2026-10-28 preference. For podcast video production, the camera/B-roll direction approved on 2026-10-30 lives in FRAME.md, under Podcast video production. The portrait complements real host footage. It is an illustrative pose, not a measured emotion label; keep another expression when the brief or scene calls for it.

Angle and placement

• Use front views for direct address and either side of a two-host composition.
• Three-quarter-right portraits face viewer-right. Place them left of the explanation they look at; choose a front portrait on the right.
• Use a fixed square box and object-fit: contain. Preserve the complete scalp, ears, glasses, beard, facepalm hand, and shoulders.
• Place the character beside the focal idea, outside the source/caption area. Reset unrelated reactions to a listening pose; retain Oscar’s thinking pose through a reflective opinion/explainer beat.
• Change whole portraits with a short anticipation and settle. Avoid distorting individual facial features or crossfading two fully visible faces.

Use in Hyperframes

Choose an asset in assets/characters/catalog.json, then put its file path directly in a static image element. The catalog is read during authoring, not fetched during a render.

<img id="oscar-reaction"
     src="assets/characters/expressions/oscar/front/skeptical.png"
     alt="Oscar, skeptical illustrated reaction"
     style="width:420px; height:420px; object-fit:contain" />

For an editable swap, bind the image source to a declared string variable with data-var-src. Give each timed wrapper a stable ID and animate its child. Predeclare all visible image assets for deterministic rendering.

A change of expression is a keyed image replacement: old portrait fades out, then the new portrait settles in. Reset unrelated reactions to listening, while keeping the preferred thinking pose through Oscar’s reflective beats. It does not create speech or mouth movement.

Rig and source scope

These are complete head-and-shoulders images. The draft mouth and rig-component atlases have not been aligned or validated, and are excluded from automatic lip sync. The 36 original full-body poses are included as references with source draft status; three Matt files retain explicit cleanup flags.

The canonical mapping uses the visible artwork: Oscar's extra front expressions come from oscar_expression_sheet_v1.png (the sheet containing facepalm); the similarly named front 3x2 sheet is an alternate six-expression set, not six new emotions. Matt's extra front expressions come from matt_expressions_front_3x2_v1.png.

Existing full-body poses

The pack also carries 36 original pose references (18 per host), copied unchanged. They retain source-draft status. Matt's marker and computer poses are flagged for edge/detail cleanup and are not selected as clean default cutouts.

Asset index