← Builds

Build · Internal

Courseware Creator

Turns source material into structured, narrated lessons.

Challenge

The Courseware Generator intake: paste source content, set audience and tone, pick a narrator voice and lesson type

Turning what you know into a lesson someone can actually learn from is instructional design. A good micro-lesson has a shape: it gains attention, states an objective, presents the idea, gives the learner a chance to practice, and checks for understanding. It carries its meaning in narration, defines terms before it uses them, and closes on active recall. Building that, scripting it, recording clean narration, and producing a video to match is a small production effort every single time.

I have spent years building lessons like that by hand. I wanted a way to take source material, usually my own, and produce a complete micro-lesson from it, grounded in the same instructional-design principles, narrated and assessed, while skipping the production line each one normally demands.

Build

The analyzed structure: learning objectives and a multi-module curriculum with per-module timing and content

You paste source content or upload a file, set the audience level, tone, narrator voice, lesson type, and target length, and the pipeline takes it from there. It reads the material and builds the outline on Gagné's Nine Events of Instruction, sequencing each lesson from gaining attention through presenting content, eliciting practice, and assessing. Bloom's Taxonomy sets the cognitive level the objectives and questions aim for, so a beginner concept lesson and an advanced judgment lesson are pitched where they belong. From that outline it writes the narration scripts, then renders them to audio in the voice you picked.

The video customization page: theme presets, custom colors, branding image slots, a captioned preview, and the generated chaptered video

The last stage turns the narrated lesson into video. You style it with theme presets, custom colors, and your own title, logo, and bumper screens, then preview the captioned result before committing. There are two ways to render: a fast pass through FFmpeg when you just need the lesson out, and a more produced version through Remotion when it needs to look finished.

Outcome

The finished lesson: a complete narrated audio lesson broken into twelve timed modules, ready to play or download

What comes out is a finished lesson: a complete narrated audio track split into timed modules, a full transcript, a self-assessment quiz drawn from the material, and a chaptered video, all bundled into one downloadable package. A paragraph of notes goes in, and a lesson a person can sit with comes out, on whatever device, with recall questions waiting at the end. It is the production team I never had, for the kind of teaching I have spent years doing by hand.

The Stack

Same pattern as the rest of the platform: Claude as pair-programmer for the build, Qwen on local hardware doing the recurring work, structuring the lesson, writing the scripts, drafting the quiz, and synthesizing throughout. Narration is generated by a local text-to-speech voice. Nothing about the source material leaves my hardware.