Three tools, one workflow, and a course lesson that's ready to publish. ChatGPT handles the thinking and writing, Descript handles the recording and editing, Canva handles the visuals. I've seen this combination become the default production stack for course creators on Ruzuku — not because it's the fanciest setup, but because each tool solves a specific bottleneck without adding complexity.
What you’ll walk away with:
- A complete course production workflow using AI at every stage
- Clear handoff points between tools
- A system that gets faster with each lesson you produce
Phase 1: Plan and write with ChatGPT (30-45 minutes)
Start each lesson with a clear learning outcome — what the student will be able to do after this lesson that they couldn't before. Feed this to ChatGPT along with your course context, and ask for a structured lesson draft.
Prompts to try
Lesson outline
"I'm creating a lesson for my course on [topic]. The learning outcome is: students will be able to [specific skill/outcome]. Write a lesson outline with: an opening hook (1-2 sentences), 3-4 key teaching points with examples, a practice exercise, and a summary. Keep the tone conversational — I'll be recording this as a video."
Discussion prompt
"Write 2 discussion prompts for this lesson that will generate real conversation, not just 'I agree' responses. One should connect the lesson content to students' own experience. One should present a scenario where students have to apply what they learned."
Quiz questions
"Create 3 assessment questions for this lesson: 1 multiple-choice testing recall, 1 short-answer testing application, 1 reflection question testing deeper understanding. Include answer guidance for each."
The key principle: ChatGPT gives you a starting framework, not a finished product. Its lesson outlines are structurally sound but generic. Your job is to add the specific examples, the stories from your practice, the nuances that only someone with your experience would know. That's what students are paying for — not information (which is free), but your curated, experienced perspective on that information.
For deeper prompting techniques, see our AI prompting guide for course creators.
Phase 2: Record and edit with Descript (45-60 minutes)
Descript's core innovation is transcript-based editing: you record, it auto-transcribes, and you edit the video by editing the text. Delete a paragraph of filler words from the transcript, and the corresponding video is cut. This is faster than traditional timeline editing for course content, where most of your edits are removing ums, long pauses, and false starts.
The recording workflow
- Open Descript, start a new recording. Choose screen + camera for slide-based lessons, or camera only for talking-head content.
- Talk through your lesson script. Don't read it word-for-word — use it as a guide. Natural delivery with occasional stumbles sounds more authentic than a perfect read.
- Don't worry about mistakes. This is the point — Descript makes fixing them trivial. Restart a sentence, pause to collect your thoughts, cough. You'll cut it all in editing.
The editing workflow
- Run filler word removal. Descript highlights every "um," "uh," "like," "you know" — remove them in one click. This alone can tighten a 20-minute recording by 2-3 minutes. But keep some natural pauses — over-editing makes you sound robotic.
- Enable Studio Sound. AI-powered audio enhancement that reduces background noise, echo, and room reverb. The difference between raw recording and Studio Sound is often dramatic.
- Trim dead space. Scroll through the transcript and delete obvious gaps — the 5 seconds where you checked your notes, the false start at minute 7, the coughing fit at minute 15.
- Generate captions. Auto-transcription accuracy is high enough for most content. Review for proper nouns and technical terms, then export with burned-in or SRT captions.
- Export. MP4 for video lessons, MP3 for audio-only versions students can download.
For detailed walkthroughs of each Descript feature, see our guides on filler word removal, Studio Sound, and AI eye contact correction.
Phase 3: Design with Canva (20-30 minutes)
Canva handles the visual materials: slides you record over, worksheets students download, and thumbnails for your lesson library. The key to speed is templates — design once, reuse for every lesson.
What to create per lesson
- Lesson slides (if recording screen-share) — use a consistent template with your brand colors. One key point per slide. Export as PDF for student download after watching.
- Worksheet or exercise — a one-page PDF tied to the lesson's practice exercise. Fillable fields if students submit digitally, printable if your audience prefers paper.
- Lesson thumbnail (optional) — a branded image for your course platform's lesson library. Consistent design helps students navigate.
If you're creating slides for the first time, start with Canva's presentation templates and modify the colors and fonts to match your brand. For worksheets, see our Canva worksheet guide.
Phase 4: Publish on Ruzuku (15 minutes)
With your video edited, materials designed, and discussion prompts written, the final step is uploading everything to your course platform. On Ruzuku, each lesson is a step in your course: add the video, attach the worksheet PDF, write the discussion prompt, and set the quiz questions. Students see a clean, sequential learning experience — video, text, exercises, and discussion all in one place.
If you're running a cohort, set your drip schedule so lessons release on your timeline. If it's self-paced, students work through at their own speed. Either way, the lesson is live and ready for students.
The human layer: where AI stops and you start
I want to be specific about what this workflow does and doesn't automate, because I've seen creators misunderstand the boundary.
AI automates the mechanics — outlining structure, drafting prose, removing filler words, cleaning audio, generating captions, laying out designs. These are production tasks where AI provides clear time savings.
You provide the substance — your unique frameworks, your client stories, your professional judgment about what matters and what doesn't, your ability to read a student's confusion and adjust your explanation in real time. Across 32,000+ courses on Ruzuku, the highest-priced and highest-completing courses aren't the ones with the best production — they're the ones with the most engaged instructors.
The workflow above gets you from blank page to published lesson in 2-3 hours. But the quality of that lesson depends entirely on the thinking you do before and after the AI tools run. An AI-outlined, AI-drafted, AI-edited lesson that's generically correct is less valuable than a rougher lesson that contains a real insight from someone who's done the work.
What this gets wrong
- ChatGPT's lesson scripts sound like ChatGPT. If you don't heavily edit them, every lesson will have the same rhythm and vocabulary. Students notice. Budget at least as much time for editing the script as generating it.
- Descript's filler removal can be over-aggressive. Removing every "um" makes you sound unnaturally polished. Leave some natural hesitations — they make you sound human.
- Canva templates can look generic. If you use the default template without modifying colors and fonts, your materials look like everyone else's. Spend 30 minutes creating a custom template for your course, then reuse it for every lesson.
Related guides
- The $0 AI Course Creation Stack — the same workflow using only free tiers
- What AI Can and Can't Do for Course Creators — the balanced perspective
- Recording Course Videos with Descript — deep dive on Descript
- Creating Course Slides with Canva — detailed Canva walkthrough
Now bring it to life
You've got the workflow. The hardest part isn't the tools — it's recording the first lesson. Pick your highest-impact lesson (the one that delivers the clearest result for students), run it through this three-tool pipeline, and publish it. Start free on Ruzuku and upload your first lesson today. You can build the rest of the course around it.