edu-math-tutorial

Math explanation video · Domain knowledge

This skill guides how to organize a math problem into a high-quality step-by-step explanation video. It does not provide a rendering script: each explanation scene is one on the main timeline creation.type=motion_clip Element, and the narration is creation.type=audio Element, and the final video is composed by Creator's deterministic rendering pipeline.

1. Problem breakdown method

First complete the solution analysis, then plan the video. Produce three layers of facts in order (write them into the delegation task brief; no need to save them to disk):

1. Problem statement: the complete problem text, with all mathematical expressions represented in LaTeX; if there is a figure, first describe it in words.
2. Solution chain: choose the most conventional solution aimed at students, and break the process into 2–5 atomic steps, each doing only one thing. The standard chain for a linear equation in one variable is: move terms → combine like terms → divide to make the coefficient 1; the standard chain for a geometry problem is: read and annotate the figure → cite the theorem → substitute and solve.
3. Common error points: annotate at most one error that students are most likely to make at each step (such as "when moving a term, change its sign"), and remind about it synchronously in the narration and on screen.

2. Video structure and explanation pacing

Fixed five-part structure, one scene per part (motion_clip Element):

Section	Content	Duration baseline
Opening	Complete problem statement + one-sentence goal ("our goal is to find x")	6–10s
Step N	One segment per atomic step: previous step result → annotation of this step's operation → result of this step	8–14s each
Closing	Final answer + one sentence of checking or summary	5–8s

Pacing rules:

• Estimate narration duration at about 4.5 characters per second; scene duration = corresponding narration duration + 1–2s silent reading buffer.
• Each section's narration is 1–3 sentences, with a single sentence no longer than 25 characters; in a step segment, state the basis for the operation first, then report the result.
• Adjacent scenes must connect in content: the beginning of the next scene shows the conclusion of the previous scene (the "previous step" column).

3. Narration copy guidelines

• Use spoken Chinese throughout, with fixed ways of reading mathematical expressions:5x-7=13 is read as "5x minus 7 equals 13"; a fraction is read as "20 over 5"; division is read as "20 divided by 5". Do not read the LaTeX symbols themselves.
• Operation sentence pattern: "Move the constant term negative 7 to the right side of the equals sign. When moving a term, change its sign, so the right side becomes 13 plus 7."
• Every sentence must be able to stand alone as a subtitle: do not use filler words such as "then" or "next we will".
• Write the narration text into the corresponding audio Element's creation.script, so the wording can later be changed and re-synthesized.
• Synchronize subtitles sentence by sentence with the narration: each narration sentence corresponds to one subtitle card (see Section 5), and the subtitle text must match that narration sentence word for word.

4. Layout and visual style (key points for the motion_clip prompt)

For each scene, creation.prompt use natural language to clearly describe the visual intent, including the following elements; the actual motion effects document is provided by design_motion_overlays Generation (do not hand-write creation.motion）：

• Theme: Default "Aurora Scholar" — light lavender/indigo gradient background, soft light spots; all scenes within the same video share one theme.
• Teaching panel: Solid white opaque rounded card carries formulas and steps (semi-transparent frosted glass is prohibited), the card has layered soft shadows; step numbers use purple circular badges.
• Formula typesetting: Mathematical expressions are typeset in KaTeX style (serif italic variables, upright numbers and operators), the problem statement/result formula is the visual subject, centered and enlarged; operation descriptions use small capsule labels placed between the two formulas (e.g., "move -7 to the right, change the sign to +7").
• Step card layout: The "previous step" formula is on top (light gray bottom bar), the operation label is centered, and the "result" is below (light green bottom bar highlighted).
• All on-screen text in Chinese: Step badges, operation labels, section titles ("Previous Step", "Result", etc.) are all in Chinese, English annotations are prohibited; each card segment's content must strictly match the step described by that segment's narration (when writing the prompt, append the original narration text of this segment as an alignment anchor).
• Subtitles: Handled by an independent subtitle Overlay Element (see Section 5), do not draw it into the motion_clip document; the teaching panel reserves about 15% height of blank space at the bottom to ensure subtitles do not overlap with the panel.
• Motion effects: Use a gentle upward float fade-in on entry, highlight item by item where formulas change; do not use flashy transitions.

5. Land as Creator native structure (execution flow)

1. Complete the breakdown and segment planning according to Sections 1–2, estimate each segment's narration text and duration.
2. Delegate ai_editing_director to timeline:<timelineId>, the task description states:
   • Create one creation.type=motion_clip main-track Element for each segment (spans are arranged sequentially according to the duration baseline in Section 2,creation.prompt write the full visual content according to the key points in Section 4, including that segment's specific formula LaTeX, Chinese operation annotation text, and the original narration text of this segment);
   • Create a creation.type=audio narration Element for each segment (creation.script uses the copy specified in Section 3, spans align with the corresponding scenes);
   • Create a creation.type=overlay subtitle Element for each sentence of narration:text fill in the original narration text of that sentence (verbatim, no rewriting or omission), arrange the span according to the time position of that sentence within the segment (allocate the segment duration by character count proportion),location use the bottom-centered capsule default values x=0.50, y=0.88, width=0.80, height=0.14; subtitles cannot be omitted, every sentence of narration must have corresponding subtitles covering its time period;
   • After writing, call design_motion_overlays to generate each motion_clip's motion document and subtitle card style, you must pass both captionStyle="uniform" and sceneStyle="edu_steps": narration subtitles use the same fixed template throughout the video, only the text changes; teaching scenes use a deterministic derivation card template (step badge/previous step/derivation row/result highlight fixed skeleton, with built-in subtitle blank space at the bottom), the model only fills in content copy and is forced to Chinese, the layout is absolutely consistent throughout the video. For this purpose, each motion_clip's prompt only needs to clearly write what this segment is about (step name, previous step formula, this step's derivation, result), no need to describe the visual style.
3. Final composition uses Creator deterministic rendering (triggered by the user in the UI or by the existing compose workflow), video is not rendered inside any sandbox.
4. Self-check: verify scene by scene that formula derivations are correct, on-screen annotations are all in Chinese and consistent with this segment's narration, every narration sentence has subtitles and does not overlap the panel, subtitle styles are consistent throughout, and scene durations cover the corresponding narration.