AI Content Generation

LearnBuilder's Creator Agent can generate an entire course structure — outlines, lesson content, and individual blocks — from a short description of your learning objectives. You review and refine; the AI does the heavy lifting.

How it works

AI generation follows a multi-stage pipeline:

1. Describe -> 2. Generate outline -> 3. Generate content -> 4. Generate media

Stage 1: Describe your course

Start from the chat box on your dashboard ("What do you want to teach?"). Rather than a single long form, the conversational wizard proposes a course shape from your topic and confirms each decision with you one at a time:

• Topic & knowledge files — describe the course in a sentence or two; attach PDFs, Word, PowerPoint, EPUB, or images with the paperclip to use as the source of truth
• Audience — auto-detected from your topic when you name one, otherwise picked from an experience level (New / Some experience / Experienced / Mixed)
• Shape — confirmed setting by setting: number of lessons, lesson length (5, 10 or 20 minutes of learner time — this sets how many blocks and words each lesson gets), detail per lesson, tone, whether to use a story or case study, instructional approach, activity mix, image style, and an optional cultural setting (asked before drafting because it shapes the story and cast). On the Free plan the wizard pre-selects the shape that fits your free credits and shows a fit indicator as you change it.
• Goal — what should change in the real world, the desired behaviour, what people do wrong now, what makes it hard in real life, and what kind of change it is. Every lesson the AI writes or rewrites is aimed at closing this gap. You can edit these answers later under Course settings → Learning design → Goal; the lesson settings dialog shows the current course goal so each lesson's outcome is written against the same frame.
  The kind of change picks the design blueprint the AI follows:

  Kind	The gap	What the AI builds
  Know it	People don't know the rule, term or fact	Retrieval practice: fill-in-the-blank for exact values, matching, "state the rule" questions, flashcards, one confusable-pairs lesson
  Do it	People know what to do but don't execute it reliably	Worked examples, step ordering, practice runs with scoring, "what do you do next" questions, an error-recovery lesson
  Decide well	The right action depends on the case	Contrasting cases, branching scenarios, decisions the learner must justify, an ambiguous-cases lesson

  
  Each blueprint also caps the block mix — at most one drag & drop per lesson and a ceiling on multiple-choice questions — and tells each lesson which rhythms earlier lessons already used, so a course doesn't repeat the same skeleton in every lesson.
• Structure from your document — when you attach a PDF or Word file with chapters, turn on Structure to make the outline follow the document chapter by chapter instead of letting the AI decide the split.
• Draft & review — title, description, cover image, objectives, the story (keep it or ask for changes), and a cast of characters (name, gender, cultural origin, role, voice, and portrait) are drafted in the chat for you to adjust before creating

The instructional approach is injected into the prompt as an explicit blueprint — Action-first lessons centre on real on-the-job decisions; Objective-first ties each lesson to a measurable objective; Story-driven sequences the course as a narrative.

A link beneath the chat box opens the original step-by-step form wizard, which exposes the same options as labelled fields if you prefer that to the conversation.

For the full walkthrough of the conversational wizard, see Creating Your First Course.

Stage 2: Generate an outline

The Creator Agent produces a structured outline:

• Course title and description
• Suggested lessons with titles and learning goals
• Recommended block types per lesson

Review the proposed lessons and structure. You can accept it as-is, edit lesson titles or descriptions, or regenerate with different parameters. Click Create Lessons to turn the outline into actual lessons in your course.

Stage 3: Generate lesson content

Once the outline is approved, you can generate content for individual lessons or the entire course:

• The agent creates appropriate blocks (text, video scripts, quizzes, dialogues, smart diagrams)
• Content is scoped to the lesson's learning goal and to the lesson length you chose
• Every generated block carries a one-line purpose — why it is in the lesson. Open it from the ⓘ in the block header; the same dialog offers Change with AI if the block doesn't do its job
• If you've uploaded Knowledge Files, the agent uses them as source material
• Open reflection questions in generated Feedback blocks come with AI feedback enabled, so learners get a constructive, ungraded response to what they write

Stage 4: Generate media

After content is generated, the pipeline continues with media generation:

• Illustrative images and diagrams — AI generates images (Seedream v5 by default) for content cells marked with image prompts, and draws vector diagrams for cells marked as diagrams
• Talking videos — lip-synced character videos are generated from script blocks (requires a main character photo)
• Scene videos — short image-to-video clips (MiniMax H3 Max by default) where a scene calls for motion. They are wordless by design: video models render written text as gibberish and speech as babble, so no slide, screen or sign is ever the subject and nobody talks to camera. Every scene prompt is sent with that guardrail appended, whoever wrote it
• Interactive slideshows — AI builds full interactive simulations from suggestions embedded in the content

Each stage runs automatically in sequence. You can monitor progress and review results as they complete. A quality review card at the top of each generated lesson lists anything the AI found weak (an unanswerable quiz, a drag & drop without a scene, a block that repeats another); Fix with AI repairs the listed items in one go, or close the card with the X.

The free first course

On the Free plan your first course is meant to fit comfortably inside the €10 of free credits every new account starts with, so generation runs in a capped profile: short lessons by default, at most three AI images per lesson and three per interactive slideshow, one talking video (up to 25 words, in the first lesson) and one AI dialogue (in the last lesson) so you see both features. Publishing the course earns €3 more, and the first learner who starts it another €3. Paid plans and topped-up accounts generate without these caps.

Tip: Knowledge Files dramatically improve generation quality. Upload PDFs, Word docs, or PowerPoints related to your topic before generating. See Knowledge Files.

Block type presets

Before generating, configure block type presets in course settings to control the mix of content the AI creates:

The AI enforces content variety regardless of preset — it will never generate more than 2 text blocks in a row.

Talking video generation

When the Creator Agent generates video blocks, it can create talking-head videos — lip-synced videos of a character speaking a script.

Requirements:

• A main character photo must be set in course settings (or provided per video block)
• A voice must be selected for the character

How it works:

1. The AI writes a script based on the lesson content
2. ElevenLabs text-to-speech generates the audio
3. MiniMax H3 animates the character portrait in a single lip-synced shot (scripts longer than its 15-second ceiling are cut into segments and stitched; VEED lipsync is the fallback). A cinematic multi-shot mode is available per video: it alternates a wide shot, a close-up and a three-quarter angle so no two neighbouring shots share a framing
4. Dialogue video — two course characters talking to each other, cut shot by shot. Pick the two speakers, describe what the conversation should show, let the AI write the lines (or type them), and generate. Each line becomes one shot of the person speaking in a shared setting, alternating wide and close-up framings per character. The lesson generator can propose one when the course has two cast members with portraits; on the Free plan dialogue videos are not generated. Reach it from a video block's Dialogue button.
5. The spoken words are verified against the script, and VTT subtitles are synced to the finished video

Scene variation: Each talking video can include a scene variation prompt that changes the character's background, clothing, or setting. This prevents visual repetition when a course has multiple talking videos.

Interactive slideshow generation

The Creator Agent can auto-generate Interactive Slides blocks — full simulations with slides, elements, triggers, actions, and scoring.

During content generation, the AI may suggest interactive simulations based on the lesson topic. These suggestions are then built out automatically:

1. The AI designs the simulation structure (slides, elements, layout)
2. Triggers and actions are wired up for interactivity
3. Scoring is configured with score, maxScore, and passingScore
4. The result is a fully functional interactive slideshow ready for review

You can also generate interactive slideshows manually using the Generate with AI wizard on any Interactive Slides block.

Task quality control

Every generated slideshow is reviewed before you see it, because its logic lives in actions you can't see on the canvas. The review asks three things about each graded task, in this order:

1. Is there a decision? A learner who skipped the lesson must be able to get the task wrong in a way that reveals a misconception. Two boundaries dragged into the only two slots, or a slider set to the number printed beside it, are clicks, not tasks.
2. Does anything give the answer away? A drag item that names its zone, a hint visible from the start, a correct option styled differently.
3. Can it be solved from what is shown, and is it wired? The accepted answer must follow from the slide's own text; every check must award score and move on.
4. Is the task visible? A task slide opens with the scenario, the inputs and the check button. Rules, rubrics and story go into a briefing overlay the learner closes with "Got it" and can reopen. Slides that open with more than about 900 characters of text, or that shrink text below 12 px to fit, are rebuilt. Text sizes are snapped to one scale (title 20, body 14, labels 12) so every slide shares a hierarchy. The same snap is available in the slideshow editor as Harmonise text sizes.
5. Can the learner read the verdict? For AI-graded writing tasks, the grading prompt must reference real inputs, one check button must grade a combined answer once, and the feedback box must be tall enough for the explanation, not just the score.

Slides that fail are rebuilt automatically. Anything the repair can't fix appears at the top of the lesson editor as a Task design warning pointing at the block, with a one-click AI fix. The review costs a few cents per slideshow.

Generating individual blocks

You can trigger generation at the block level inside any lesson:

1. Click Add Block
2. Select a block type
3. Click the Generate with AI option
4. Describe what you want, then confirm

Refining generated content

AI-generated content is editable — treat it as a first draft:

• Click any block to edit text, questions, or settings
• Regenerate a single block without affecting others
• Adjust tone, length, or focus as needed

AI costs

AI generation is charged in euros at fixed standard prices, drawn from your plan's monthly budget (or pay-as-you-go on the Free plan). You see the exact price before you confirm.

Note: The exact price is always shown on the button before you generate. See AI costs & transparency for the full price list and how the cost ledger works.