anbeime/skill / skills/anything2explainer/anything2explainer
anything2explainer
Unverified给一个主题,产出一条黑底 MG 风格(幕底可选星点或点阵波)、有配音字幕章节进度条的科普讲解视频(中文或英文;Remotion 代码动画;时长由用户定,常用 3–5 分钟)。内含可编译模板、图元库、配音/分镜/渲染工具、风格与动效规范、多 agent 分工协议与 QC 判据,以及一条完整样片(《RAG 与知识库》)作为质量标尺。Turn any topic into a narrated motion-graphics explainer video in Chinese or English, on a black canvas with a star-field or dot-field backdrop, TTS voiceover, subtitles and a chapter progress bar, every frame drawn in code with Remotion. Use when the user asks for an explainer, educational or science-communication video about a topic, or wants an article or document turned into a video.
reference/narration-guidance.md
View on GitHub# Writing narration for a short technical explainer — general guidance
Distilled from reading reference narrations of very different styles (an 8-minute conceptual explainer, a 10-minute
glossary-style animated guide, a 100-second rapid overview). These are **high-level principles**,
deliberately not the specific devices any one of those scripts uses. They apply to any 2–10 minute narrated explainer whose
pictures are drawn to the words (motion graphics), in any language.
## 0. The one test
Read a sentence aloud and ask: *does the listener now know, picture, or feel something they didn't a sentence ago?*
If not, the sentence is filler. A five-minute script has room for roughly 45–50 sentences; every one must pass.
## 1. Start inside a situation, not inside a definition
Put the listener in a concrete scene, problem or puzzle before you name the concept. The definition, when it comes,
should feel like the answer to a question the listener already has. A script that opens with "X is defined as…" or
"In this video we will…" has spent its most valuable seconds on nothing.
## 2. One spine
The whole script answers one question or follows one journey (a process from start to end, a tension that resolves,
a mystery that gets solved). Every section is a visible step along that spine, and the listener can always say where
they are and why this step follows the last. The failure mode is the **glossary**: a sequence of concepts each given
the same treatment, in an order that could be shuffled without anyone noticing. If the sections could be reordered,
there is no spine.
## 3. Motivation before mechanism
For every idea, in this order: *what goes wrong without it → how it works → what that buys you*. Never explain a
mechanism the listener does not yet want explained. "Why" sentences are the ones that make a technical script feel
intelligent rather than encyclopedic; they are also the ones most often cut for length. Cut mechanism detail first.
## 4. Concreteness at the sentence level
Every abstract noun gets an immediate, specific instance the listener can picture: a question someone actually asks,
a thing that actually happens, a named object. Numbers are never recited; they are **converted into felt scale**
(compared with something from ordinary human experience) or they are cut. A number the listener cannot feel is noise.
## 5. Analogies are tools, not decoration
Use an analogy only when it lets the listener make a real inference they could not make without it, keep it
consistent while it is in use, and drop it the moment it stops predicting correctly. One good analogy carried through
several beats beats six clever ones used once. Never attach an analogy to every section by template; the listener
hears the template, not the content.
## 6. Voice: a person who finds this interesting, talking to one smart friend
Plain words, contractions, occasional judgment ("this is the part that matters", "that sounds obvious, it isn't").
Honesty about limits and uncertainty ("roughly", "nobody fully knows why") builds more trust than confidence.
No textbook or corporate register, no hedging filler, no announcing what you are about to say. The narrator is
allowed an opinion about what is important.
## 7. Vary the rhythm
Long explanatory sentences followed by a short punch. A question, then a pause, then the reveal. Monotony of
sentence shape is what makes a script *feel* boring even when its content is fine. Read the whole thing aloud and mark
any stretch of three sentences with the same length and the same grammatical shape; rewrite one of them.
## 8. Jargon is a label for something already understood
Explain the idea in plain words first, then attach the technical name ("…this is what people call X"). A term
introduced this way costs nothing; a term introduced first and explained second costs the listener a sentence of
anxiety. Ration terms: every name you introduce must be used again later, otherwise it did not need a name.
## 9. Stakes early, payoff at the end
Within the first minute the listener should know what changes if this is done well or badly (money, time, what
becomes possible). The ending returns to the opening scene or question with the new understanding and reframes it;
it is a resolution, never a summary list and never a farewell.
## 10. Density and cutting
Prefer fewer ideas fully understood over a complete inventory. Cut: restatements, transitions that only announce
("next, let's look at…"), lists longer than three items unless the list itself is the point, and any sentence whose
only job is to be complete. When two adjacent sentences say the same thing at different abstraction levels, keep the
concrete one.
## 11. Say what the picture cannot
In a motion-graphics explainer, the narration and the animation split the work. Narration should carry reasons,
consequences, comparisons and judgments; the picture carries structure, sequence and quantity. Never narrate the
diagram ("here we can see…", "looking at the chart…"). If a sentence only describes what is on screen, cut it and let
the screen do it.
## 12. Precision earns trust
Names, dates, real numbers with sources, real product names. Mark approximations as approximations. Never round a
fact into a slogan. An explainer that is exciting but slightly wrong is worth less than one that is calm and right;
the goal is to be both.
## 13. Hand off at the seams
A section boundary is where a listener is most easily lost: the previous idea is finished, the next one has not started,
and a title card is doing nothing but announcing. Make the join carry weight. The last stretch of a section should leave a
question, a cost or a gap that the next section exists to close; the first sentence of the next section names what was just
achieved and then opens that question - "the manual is built; now the real exam starts". Nobody should have to ask why this
section follows the last, and no section should begin from a standing start ("now let's look at X").
This is not the announcing transition §10 tells you to cut. "Next we'll cover chunking" only reports the running order;
"the manual is only as good as the way it was built" creates the need for the next section. Announce nothing, motivate everything.
## Quick pre-flight for a finished draft
- Opening sentence: is anyone in a situation yet? (§1)
- Could the sections be reordered without loss? (§2) If yes, find the spine.
- Count "why" sentences vs "how" sentences; there should be at least one why per how-cluster. (§3)
- Every number: converted to felt scale or cut. (§4)
- Every analogy: does it let the listener infer something? Is it used more than once? (§5)
- Any sentence that announces, summarizes, or describes the picture? Cut. (§6, §10, §11)
- Read aloud; mark three-in-a-row same-shape sentences. (§7)
- Every technical term: plain idea first, name second, and used again later. (§8)
- Every section boundary: does the previous section leave a question and the next one open by answering it, instead of starting cold? (§13)
- Ending: does it reframe the opening? (§9)
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 470k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |