One Creator Can Run the Whole Course Video Library

The short answer: One person can run an entire course video library. Golpo scripts the lesson, plans the scenes, designs the whiteboard, narrates it, syncs every picture to the voice, and returns a polished video in minutes. Upload a selfie and the instructor is on screen. Use your own audio or an authorized voice clone and the whole catalog still sounds like you.

Why this is practical now: A course used to wait on a scriptwriter, an illustrator, a voice artist, an animator, and an editor. Golpo does that production in one pass, so a creator can publish a new visual lesson whenever the curriculum is ready, then keep going until the library covers the course.

The catalog can still feel personally taught. The same face can introduce every module. The same voice can carry every explanation. Frame-by-frame editing fixes a label without starting over. The Golpo API, the Codex skill, the Claude skill, and the Golpo MCP server can generate the next lesson from Cursor, Claude Code, VS Code, or Claude Desktop while you stay on the teaching.

Use the full Golpo production system

Golpo is built to be the world's best AI whiteboard explainer video maker. Give it a topic, a prompt, a script, an audio recording, or a source document. It scripts, plans the scenes, designs the visuals, narrates, times and synchronizes every frame, and returns a polished video. One person can run that system across a whole course.

Across the library, keep one production system but choose the presentation mode lesson by lesson. Use pure whiteboard explanation for a worked idea, the instructor's voice for continuity, and a selfie-based presenter when the creator should introduce or close the lesson.

• Start with an idea or source. Golpo can draft the script and plan the scenes from a topic, a document, or your outline. Script Mode can also speak the exact words you already trust.
• Keep the creator present. With AI Presenter, a course creator, founder, professor, or YouTuber can upload a selfie and appear beside the whiteboard. One person can lead a large lesson library without recording every visual sequence on camera.
• Use the right voice. Own Narration turns uploaded or live-recorded audio into the visual lesson. An authorized Voice Clone gives recurring lessons a recognizable voice at scale. Consent and pronunciation review remain required.
• Repair instead of restarting. The frame-by-frame editor can replace text, images, references, or an individual visual when the QA pass finds a local problem.
• Automate at scale. The Golpo API can be called from Codex, Claude Code, or through the Golpo MCP server in Cursor, VS Code with GitHub Copilot, Claude Code, Claude Desktop, and Visual Studio. The course platform can generate, poll, review, and publish without a person moving files between tools.

What Golpo produces, and what the course platform controls

A solo creator who wants a lesson without a camera already has a path in course videos without recording. Turning a pile of existing files into a catalog is the job of building a library from course content. Putting a recognizable presenter beside the picture is the Avatar lesson library. This page owns the house rule those guides can point at: the lesson is finished for a learner when its check agrees, and not before.

Spoken words and picture directions stay apart. The script-to-video guide is the control. If a stage direction lands in the script, Golpo may read it aloud. Numbers that must appear on screen belong in the narration, and the pictures have to match those numbers. Video Instructions and Voice Instructions are where the look and the delivery are locked for a series.

The house rule

A lesson ships when four records exist and agree.

1. Source. A curriculum note, a standard, a paper, a keyed item, a passing test, or a rate table with an effective date. The URL or the file version is stored with the lesson.
2. Draft. The platform’s creator agent writes a one-minute script from that source. It may also write the practice item. It does not publish.
3. Independent check. A different agent, or a person who did not write the draft, solves the problem again or reads the source again. Agreement is a number, a quotation, or a test result, not a feeling that the draft “looks right.”
4. Watch-back. After Golpo returns the file, someone listens for the numbers and looks at the labels. A mismatched digit, a diagram that shows a different step, or a failed generation returns the attempt to an instructor. The platform does not quietly generate a replacement and hope.

The QA agent can reject a draft, accept a draft, or mark it uncertain. Uncertain is not a soft yes. Uncertain waits. The same three outcomes apply to the pictures. A video the learner can press play on is the narrow case where script check and watch-back both passed, and an instructor has not pulled it back.

Why the library is built around practice

Dunlosky, Rawson, Marsh, Nathan, and Willingham reviewed ten learning techniques in Psychological Science in the Public Interest (2013). They rated practice testing and distributed practice as high utility. Elaborative interrogation, self-explanation, and interleaved practice were moderate. Highlighting, rereading, and the keyword mnemonic were low. Summarization and using imagery for text learning were also in the low group. Reconfirm that table in the paper before any script that cites it is locked.

The design rule this library takes from that paper is small. A lesson that only restates a page is the low-utility pattern. A lesson that makes the learner retrieve, then brings the same idea back later, is the high-utility pattern. The on-demand spoke is where a single video pauses for a problem. The hub’s job is to refuse a library made only of re-reads.

That paper does not say a one-minute video will raise a score, a completion rate, or revenue. This guide does not estimate those either. It says which techniques the authors rated, and it uses the two high-utility ones as a production rule: put a check inside the lesson, and bring the idea back on a later day.

The fifteen jobs under this rule

Each guide below owns one job. Where an older guide already owns a neighboring job, the link is a handoff, not a rewrite.

• SaaS accounting lessons — one journal entry, with the assumption that makes the revenue number true.
• Explain This — Algebra Practice Studio approves a script, then sends it to Golpo API v2 and stores the job ID.
• Lesson plus practice — the answer arrives after a pause, not during the explanation.
• The learner’s wrong step — start from the line they wrote and show where it breaks.
• Test-prep questions — an approved question or screenshot, a verified worked solution, reasoning for every choice, and no unsupported score promise.
• Beside the written lesson — the paragraph stays canonical; the video draws one case.
• Exercise walkthroughs — a blank page, legal steps, and a check at the end.
• Syllabus libraries — standard, objective, item, and answer stored on the same row.
• University overviews — the time budget and the first idea, for students who already have a seat.
• Professor-signed concepts — a named reviewer, a date, and a block on release.
• Programming plus tests — the spoken result is the result a unit test just printed.
• Language editions — the same item ID and the same numerals, after a glossary is signed.
• Textbook companions — indexed to the section and the answer key, with an original parallel item.
• Training updates — the file a learner watched is tied to the rate’s effective date.
• Recurring questions — a repeated question becomes a mini-lesson only after a duplicate check.

A week that can produce two checked lessons

Monday is source day. One lesson is the learning-technique rule above. The other is a house arithmetic check: fifteen percent of 80. Ten percent of 80 is 8. Five percent of 80 is 4. Eight plus four is 12. The creator agent may draft both scripts. It may not mark either one ready.

Tuesday is the independent check. For the paper, the QA agent has to find the utility ratings in the text, not in the draft. For the percent, it recomputes 0.15 × 80 and also the split 8 + 4. If either path disagrees, the draft stops. The platform records the disagreement. It does not ask Golpo to “fix the math” inside a prompt.

Wednesday is render day, and only for a script the check accepted. The request uses Script Mode so the spoken words are the approved words. Thursday is watch-back: every spoken number against the script, every label against the allowed list, and a listen for a claim the source does not make. Friday is the instructor queue, which should be short if Tuesday was real. A queue full of first-time fact repairs means the QA agent is being used as a spellchecker.

The two lessons below are that week. They are Script Mode renders. The article holds the citation and the recomputation. The videos speak the lesson.

Own Narration: OFF · Script Mode: ON · Duration: 1 minute · Orientation: Horizontal (16:9) · Style: Professional Clean · Color: ON · Language: English · Voice: Female 1 · Pace: Normal · Music: None

A finished course can still ask people to study in a way that fades. In 2013, John Dunlosky, Katherine Rawson, Elizabeth Marsh, Mitchell Nathan, and Daniel Willingham reviewed study techniques for Psychological Science in the Public Interest. They rated practice testing high utility. They rated distributed practice high utility. Practice testing means the learner has to bring the idea back, not only look at it again. Distributed practice means the same idea returns on later days, instead of staying inside one sitting. Elaborative interrogation, self-explanation, and interleaving were rated moderate. Highlighting, rereading, and keyword mnemonics were rated low. This course uses those two high ratings as a design rule. Every lesson ends with something to retrieve, and the idea comes back later in the term. A clear picture is not the finish line. The rating describes the technique. A review can make a weak lesson less likely. It cannot guarantee what a learner will remember.

A course shelf. Two techniques, practice testing and distributed practice, are marked high. Three techniques are marked moderate. Three are marked low. End on a lesson that ends in retrieval and returns on a later day. Allowed labels: 2013; Practice testing: high; Distributed practice: high; Moderate: elaborative interrogation, self-explanation, interleaving; Low: highlighting, rereading, keyword mnemonics. Do not add a sample stamp, a demo label, or any claim that the lesson is fictional. Show only the labels listed. If a number appears, it must match the narration exactly. No extra dates, prices, percentages, logos, or quotations.

Calm curriculum lead. Precise with the two high ratings. No promise about scores.

Own Narration: OFF · Script Mode: ON · Duration: 1 minute · Orientation: Horizontal (16:9) · Style: Dry Erase · Color: ON · Language: English · Voice: Male 1 · Pace: Normal · Music: None

Fifteen percent of eighty is the house check before a lesson ships. Ten percent of eighty is eight. Five percent of eighty is four. Eight plus four is twelve. Fifteen percent of eighty is twelve. A line that says fifteen, or a picture that stops at eight, stays on the desk. A creator agent on this course platform drafts the lesson. A separate QA agent recomputes the ten percent, the five percent, and the sum. Those agents are part of the platform. Golpo does not employ them. Golpo draws the script after the figures agree. A passed check makes a wrong number less likely. It cannot guarantee the video is accurate. If the drawn twelve does not match the spoken twelve, the file goes back to the instructor. The lesson is finished when the arithmetic agrees. Twelve is the only figure that passes.

A whiteboard. Ten percent of eighty is eight. Five percent is four. Eight plus four is twelve. A wrong stopping point at eight or fifteen is crossed out. The file returns to an instructor if the drawn twelve disagrees. End on twelve. Allowed labels: 15% of 80; 10% = 8; 5% = 4; 8 + 4 = 12. Do not add a sample stamp, a demo label, or any claim that the lesson is fictional. Show only the labels listed. If a number appears, it must match the narration exactly. No extra dates, prices, percentages, logos, or quotations.

Patient and exact. Pause at eight, four, and twelve. No excitement.

What blocks a lesson

• A number in the script that the recomputation does not reproduce.
• A citation the QA agent cannot find in the named source.
• A picture that shows a step the narration does not say.
• Any description of the creator agent or the QA agent as a Golpo feature.
• A line about score gains, completion, revenue, or “this video is a demo.” The article may say a case is constructed. The video teaches the case as the lesson.
• A generation failure. The attempt returns to the instructor with the saved job ID. The worker does not mint a second job to hide the first one.

Risks the check does not remove

Two agents can share a wrong source. If the curriculum note says fifteen percent of 80 is 15, a careful recomputation from that note will also say 15, and the video will be faithfully wrong. The fix is a source that is the definition or the raw arithmetic, not a summary of a summary. The instructor’s job is to notice when the source itself is the weak part.

A watch-back can miss a label that flashes for a second. That is why the allowed-label list is short and why a mismatched digit is a stop, not a note for later. Speed of drawing is what makes a bad lesson easy to multiply. The house rule is the brake.

Practice testing and spacing are a design choice supported by the 2013 review. They are not a promise that this course will outperform another course. Anyone who wants that claim needs a study this guide does not contain.

The record every spoke has to share

The fifteen jobs look different on the page, and they write the same kind of row underneath. A course that invents a fresh database for each spoke will not be able to answer a simple question in June: which lessons are safe for a learner to open today? The shared record is how the house rule becomes visible.

• Lesson ID and spoke, so an accounting entry is not stored as if it were a test-prep item.
• Source URL or file version, and the date a person opened it.
• Script hash. Any edit, including a one-word edit, changes the hash and clears the prior pass.
• QA result: agree, disagree, or uncertain, plus the recomputation or the quotation the agent used.
• Render job ID, once a generate call has been made. Empty until then. A second job is a new row, not an overwrite that hides the first failure.
• Watch-back result and the name of the person who watched, when a person was required.
• Learner-visible status: blocked, or assigned. Assigned is the only status the player treats as playable.

The instructor queue is a view of rows that are disagree, uncertain, or failed at render or watch-back. It is not an inbox of every draft. If every draft lands in the queue, the QA agent is not being asked to decide. A healthy week in a small course might show two uncertain paraphrases and one bad label. A week that shows forty arithmetic failures means the creator agent is drafting from notes nobody checked, and the fix is earlier, in the source field.

Monday through Friday, described on this page for the technique lesson and the percent lesson, is the same week the spokes run. An accounting team spends Monday on the assumption. A studio spends Monday on the learner’s line. A department spends Monday on the signer’s calendar. Friday is still the queue, not a new spoke. Opening a sixteenth workflow because the queue is long will not shrink the queue. Shortening the allowed-label list and refusing to render uncertain drafts will.

People will ask for a dashboard of completion, revenue, or score lift. This cluster does not supply those numbers, because it does not have a study that measured them. The dashboard this record supports is operational: how many assigned lessons have a source date inside the course’s review window, how many job IDs failed watch-back, and how many uncertain rows waited more than the course’s own limit. Those counts tell you whether the brake is real. They do not tell you whether the course will sell.

Automate it: review gates and publishing with the Golpo API

For every lesson in the library, the whole path from source to published lesson can run as one pipeline on the course platform. An agent in Codex with the Golpo Codex Skill, in Claude Code with the Golpo Claude Skill, or in Cursor, VS Code with GitHub Copilot, Claude Code, Claude Desktop, or Visual Studio through the Golpo MCP server can call the Golpo API, wait for the render, and hand the file to the platform's review steps. Production stops being the limit on how many checked lessons a course can offer.

1. Attach the reference first. Store the lesson's source: curriculum note, keyed item, paper, passing test, or dated rate table with the lesson before any script is drafted.
2. Review the script against the reference before rendering. If a solution or source exists, the platform's QA agent checks the draft against it: it re-solves, re-adds, or re-reads, then compares. A mismatch or an uncertain result stops the job before Golpo renders anything.
3. Generate with Golpo. The agent sends the approved script in Script Mode, with the course's locked style, voice, Video Instructions, and Voice Instructions.
4. Transcribe what was actually said. Run speech-to-text on the finished video's audio. Learners hear the transcript, not the draft.
5. Run two independent transcript reviews. Make two separate LLM calls, separate from the creator agent and the QA agent. Each gets only the transcript and the reference, with no access to the draft or to the other reviewer's verdict. Each checks every spoken number, term, and claim, then returns pass or fail and quotes the line it objects to. The lesson moves on only when both pass. A failure or a disagreement goes to an instructor.
6. Check the frames. Pull a frame every few seconds and confirm every number and label on screen matches the approved script.
7. Publish. The pipeline posts the video to the course platform or LMS through its API as a draft, then releases it under the course's own approval rule.

FAQ

Can these lessons be generated and published automatically?

Yes, on the course platform's side. Using the Golpo MCP server or the Golpo Claude Skill, a pipeline can check the script against the lesson's source: curriculum note, keyed item, paper, passing test, or dated rate table, generate the video, run two independent transcript reviews, and post the lesson to the platform as a draft. Anything that fails a check goes to an instructor.

Does Golpo include a creator agent or a QA agent?

No. Those agents are proposed components the course platform would own. Golpo receives a script the platform has approved and returns a rendered video. The platform still has to review the pictures and the audio.

Can a QA agent guarantee the lesson is correct?

It can reduce errors by recomputing or by checking a source. It cannot guarantee accuracy. A shared bad source, a missed label, or an uncertain judgment can still pass. Uncertain results go to an instructor.

When is a lesson allowed to reach a learner?

After the independent check agrees with the source or the arithmetic, Golpo has rendered that approved script, and a watch-back of the audio and visuals has passed. A failed generation or a mismatched number goes back to an instructor.

Who writes the script?

The platform’s creator agent drafts from material the course supplies. A person can draft instead. Either way, the words that reach Golpo are the words a separate check has accepted. Shot directions stay in the video instructions.

Why start the library with practice testing and spacing?

The 2013 Dunlosky review rated those two techniques high utility, and it rated highlighting and rereading low. The library uses that as a production rule: each lesson should make the learner retrieve, and the idea should return later. The paper is not a promise of higher scores.

What is the fifteen percent of 80 check for?

It is the house arithmetic. Ten percent is 8, five percent is 4, and 8 + 4 is 12, which is also 0.15 × 80. A lesson that cannot survive that kind of recomputation is not ready to render.

Should every course use all fifteen spokes?

No. Use the spoke that matches the job. An accounting curriculum, a wrong-step diagnosis, and a training-rate update are different promises. The house rule is the part they share.

Where do the camera-free workflow and the file inventory live?

The no-recording workflow is the course-creators guide, and the inventory-and-batch path is the library-from-course-content guide. This page links to them and then stays with the check that decides a lesson can ship.

Related guides

• Course videos without recording — one creator, no camera, from lesson to file.
• Build a library from course content — inventory, taxonomy, and batch conversion.
• Avatar lesson library — a presenter beside the explanation.
• Script to video — keep the approved words intact.
• Explain This and the Golpo API — the spoke where the approved script becomes a tracked job.
• Lesson plus practice — retrieval inside a single minute.

Pick one spoke, lock its source, and ship the first lesson only after the recomputation and the watch-back agree.