Communitygithub.com

2471468057-bot/MORL-Subtitle-Match

Reviews theme identification, composition, typography, and full in/out transitions; revises based on current needs, avoiding turning past feedback into global bans.

What is MORL-Subtitle-Match?

This skill provides a comprehensive framework for designing subtitle motion graphics where the subtitle is treated as a short-duration animated piece carrying the actual body text. It prioritizes the user's current themes and requirements—accuracy of the text, readable on-screen presentation, and controlled timing and positioning—over rigid, pre-defined visual constraints like long rectangles, bold strokes, thick frames, or abstract UI. The skill requires planning a thematic event and identifiable composition before fitting the text into it. It supports a wide range of visual styles (simple, abstract, figurative, character-based, or object-based) and forbids elevating past failures or old preferences into permanent prohibitions. The action design follows a causal chain: a trigger moves first, then the structure responds, the text forms, local dynamics play out, and finally the sequence is undone/transformed/recycled. Each piece selects appropriate events rather than defaulting to uniform rotations, sweeps, pixelations, or fades. Technical boundaries like using the host's declared HTML/SVG/CSS static layers and a single MotionSpec clock are respected as hard constraints, but the skill explicitly warns against mistaking these boundaries as aesthetic limitations—free source code is a normal creative path, and components are optional shortcuts, not fallback plans for unsuitable themes. The skill also outlines a project collaboration and effective workflow: new pieces are created freely per current requests, revisions protect only what the user has actually approved, failed drafts are stored as negative examples without freezing future designs, and technical checks (normal proportions, 6-second and 1.5-second tests, sentence changes, forward/reverse seeking, clear start/end) are separated from authorial review and user satisfaction.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/2471468057-bot/MORL-Subtitle-Match/tree/HEAD/assets/design-skills/subtitle-theme-aesthetics

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Theme-First Subtitle Motion Graphics · 2026-09-27

Treat subtitles as short-duration animated pieces that carry the actual body text. The user's theme and current requirements take priority; the subtitle only constrains the text's accuracy, legibility on the actual screen, and controllable timing and positioning. It is not required to be a long rectangle, bold stroke, thick frame, or abstract UI. First conceive a thematic event and an identifiable composition, then adapt the text (short or long) to it. Simple, abstract, figurative, character-based, and object-based approaches are all valid. No single failure or old preference should be elevated to a global ban.

Theme names may reference related symbols, patterns, artifacts, characters, or spatial relationships—no separate design authorization from the user is required. Select a small number of the most recognizable elements and compose them together with the text/supporting structure; do not rely on arbitrary small icons, color swaps, or title claims that something "looks like" a game. If the user explicitly says "no characters / no frame / keep character," that applies only to this task. When no such requirement is given, do not force characters in, nor preemptively forbid them. The existence, thickness, non-rectangular outline, open boundaries, and proportions of the base panel are determined by readability and the theme.

The planning description must briefly cover: the visible thematic basis, the relationship between the body text / supporting structure / emblem, the entrance event, the reading-phase activity, and the exit event. Theme identification check: without the name or description, can the direction be recognized from the silhouette, pattern, or motion? After converting colors to grayscale, does the design still have a unique form? A simple stone slab, green-gold border, or thick white text is only a baseline—do not use a name like "Temple / Ancient" in place of actual design.

Actions follow a causal chain: the trigger moves first → the structure responds → the text forms → local dynamics play out → undo / transform / recycle. Each piece selects appropriate events—do not uniformly default to rotations, sweeps, pixelations, or fades. Opacity can assist visibility; but only full-group opacity changes or slight translations do not constitute a thematic action. Both the entrance and exit should have visible subject transformations; the exit may differ from a reversed entrance. Body text must be clear during the reading segment, while allowed meaningful activity happens in the periphery. For short cues, reduce minor events and itineraries while preserving theme identification; do not accelerate the entire piece and fill it with clutter.

Typography must follow the theme: choose real Chinese fonts, considering stroke temperament, hierarchy, texture, and letter spacing. It is not mandatory to add outlines or embossing to every draft. A design without emphasis words is still a complete piece. Short sentences adaptively tighten; long sentences naturally wrap. Do not alter the body text or fill space with extra fake English slogans.

Use the host's declared HTML / SVG / CSS static layers and the MotionSpec single clock. Safety, text binding, font loading, repeatable seeking, and export consistency are hard technical boundaries. Do not misinterpret these boundaries as aesthetic limits. Free source code is a normal creative path; components are only optional shortcuts. Do not treat unsuitable theme presets as fallback plans for failure.

New works are conceived freely. For revisions of existing works, first clarify the items locked down for this round. Preserve structures and actions the user has approved. Rejected drafts from failures serve only as negative examples—they do not automatically freeze, nor do they automatically overwrite adopted works. Technical passes allow preview / save / adoption; aesthetic suggestions do not trigger unauthorized rework. Perform actual checks: normal proportions, details, animations at normal speed, 6-second and 1.5-second tests, sentence changes, forward/reverse seeking, clear start/end frames. Record technical passes, authorial review, and user satisfaction separately.

Project Collaboration & Effective Workflow

New works are created freely per current requests. Enhancements only protect items actually approved or explicitly locked down by the user; failed drafts are not a satisfaction baseline. Strict material revisions do not change structural actions; explicit reconstruction allows coordinated modifications. Do not conflate the two tasks. Keep old drafts, model first drafts, and local refinements saved separately and identifiable.

For actual generation, load assets/design-skills/abstract-caption.md, material-caption.md, and any matching theme guidance. The design_guidance returns the version and SHA of each file. The project skill stays in sync with the bundled copy. After modifying, test the actual prompts for normal planning, implementation, and revision—do not only edit this file. Safety / font / binding / clock boundaries are governed by the code capability contract; do not repeatedly accumulate old restrictions.

Individual skills in this repo

This repo contains 1 individual skill — each has its own dedicated page.

Related Skills