How to Create an Online Training Course

When was the last time you started a course, and never finished it? (Don’t worry I won’t tell. Unless it’s mandated training, in which case you should definitely finish it.)

Whether it’s a course you saw in your LinkedIn page feed or a certification program you were certain you’d make time for, there’s likely one of three reasons you never finished. 

One, the training wasn’t what you expected. It didn’t align with your experience or skill level. Two, the training was bad. You know what I mean, the content or facilitation couldn’t keep you engaged. Three, you got busy and prioritized other things.  

That’s the reality our employees are facing, too, and what you’ll need to anticipate when designing any training course. So let me help you make a course that you’d actually make time to finish. 

Note: These steps reflect the reality of designing learning content in many organizations. You'll likely recognize them as aligning to the phases of ADDIE. That said, you might be handed a content brief with decisions like target audience and specific resources already determined, in which case, you may skip some of these steps.

Step 1: Define the performance outcome 

Raise your hand if you've had a stakeholder come to you and say, "we need a training on [x]." This is why training fails before it even gets designed. It focuses on content instead of capability.

Effective training courses start with a learning objective and a performance outcome. You can use something like:

This course is for [specific role] who currently [context or gap]. After completing the course, they should be able to [observable action] so that [performance outcome].

No matter how you write it, your learning objective and its performance outcomes should be clearly articulated before you start building. Otherwise you end up with disorganized content that can't be evaluated against business metrics.

Step 2: Design for the gap

Even if you already know who your target audience is (their roles, skill levels, job responsibilities), your job is to do a deeper needs analysis to figure out why there is a gap in performance outcomes today, and how you can close it. 

That means running a content audit alongside your needs analysis: "what does this learner actually need?" and "what content already exists that serves that need?"

The audit questions to start with:

• What do they already know?
• What information already exists, and where?
• Who is a subject matter expert on that audience?

You want to reduce the likelihood that someone opens this course and thinks, this is repetitive. Your job is to reduce extraneous cognitive load by removing unnecessary information. The course should be as focused as possible. Every piece of content should exist because it closes the gap, not because it's related or already written.

Note: AI tools are making it possible to deliver more personalized courses at scale. This capability often depends on two things: an AI-native learning platform, and a robust content library with contextual information. In this scenario, a course could be customized to the learner by taking into account an individual's experience, offering additional context, the ability to skip ahead, or even personalized tutoring.

Step 3: Outline the course

Once you understand your audience and have mapped your existing materials, you can begin to outline your course. And remember, you don't have to cover everything in one course.

A course is made up of modules, each representing a meaningful milestone toward your performance outcome. Within each module, individual lessons should have one job: explain one concept, demonstrate one task, or prepare learners for what comes next. 

Think of it like a car. It's the vehicle that gets an employee from one place to another. If it tries to go to multiple destinations at once, it's a miserable experience for everyone in it.

Keep your lessons focused and your modules sequenced deliberately. Research consistently shows that spaced, sequential learning produces significantly better retention than covering the same material in one sitting, which is really just science confirming what we already knew from university: cramming the night before doesn't work. You're better off having gone to all the classes.

In practice, employees would rather complete ten 5-minute lessons than sit through one 50-minute course. They're also more likely to actually make the time.

Note: There's no single right length for a training course. That said, I'd encourage you to chunk your content into focused, modular lessons. It makes courses easier to update, easier to localize, and easier for AI to adapt into personalized learning plans.

Step 4: Design within your constraints

Most L&D teams already have three things in their tech stack that support course development: an LMS or LXP, an eLearning authoring tool, and an LLM. So this step is less about selecting new tools and more about understanding what your existing tools can actually deliver. Tools can unlock design ideas or block them.

Your LMS determines how the course is delivered and tracked. Your authoring tool shapes what's actually buildable. And your LLM can accelerate almost every stage of development, from synthesizing existing content and writing learning objectives to generating scripts, activities, and assessments. The catch is that your authoring tool or SCORM requirements may constrain what you can actually deploy, so it's worth understanding those limits before you get deep into development.

When teams come to Synthesia for support with course development, it's often because they need a better way to produce video-based learning that still embeds within their authoring tool or LMS.

Note: If you're evaluating eLearning authoring tools, we have a guide reviewing the top tools.

Step 5: Develop your content

Because I can't design a personalized workflow for every tool combination in this post, what I want to show you is how video content gets built inside a course workflow using Synthesia.

If you've ever used a traditional video tool, especially ones bundled with eLearning software, you know how laborious it is. I used to have a dedicated PC laptop just for video editing. AI video works differently.

Generate a first draft

Most L&D teams start one of two ways: they upload existing materials like a slide deck or PDF into our AI video generator to get a first draft in minutes, or they start from a template.

If you're new to designing learning experiences for video, templates are a lower-friction entry point. Ours are built by instructional designers, so you can focus on the content rather than the layout. Enterprise clients receive customized and branded templates.

Choose your avatar

Once you have a draft, decide whether you want to use an avatar. You can customize everything from the language and accent to clothing and physical environment. Enterprise clients can create studio-quality custom avatars of themselves.

Note: Some customers prefer to use the avatars for narration only, as having an avatar onscreen adds to cognitive load.

Customize your scenes

From there, decide what other components to incorporate: screen recordings, motion graphics, b-roll, or interactivity such as knowledge assessments and branching scenarios. Preview before generating and check that pronunciation, pacing, and flow all make sense.

Note: some generated assets may appear as placeholders or low-fidelity previews until you generate the final video.

Localize your video 

When you've generated your video, you can localize it into over 160 languages with a few clicks. If you’re localizing videos, be sure to include native speakers when you conduct any pilots. 

If you want to see how an organization uses Synthesia for onboarding and customer support training, check out this case study featuring Sky Italia.

Step 6: Gather feedback and iterate

At this point, whether you have a complete draft of a course or discrete components like a training video, you want to start gathering feedback from SMEs and a pilot group of learners.

For SMEs, ask them to review whether the content is accurate, up-to-date, and complete. For the pilot group, ask them where they get confused or stuck and where they disengage. If possible, I recommend having a group go through portions of the course in real time so you can see where they get stuck.

This is your opportunity to collect feedback and incorporate it before a broader launch. You can often find fixes that, once published, are harder to correct.

Step 7: Publish and evaluate your course

When you're ready to publish, consider three things:

• Where will this course live, and how will people find it?
• Who will have access, and what happens when it's shared?
• How do you update the course without creating confusion or outdated versions?

The decisions you make here will shape what you can measure and how reliably you can track it, especially if you're choosing between an LMS or a format like xAPI.

Note: Synthesia videos can be incorporated into courses as a SCORM package, embedded via a published link or HTML embed, or exported as an mp4 (though mp4 won't support interactivity). They can also be directly integrated into some LMS/LXP platforms and localized into 160+ languages with a few clicks.

Finally, go back to your learning objective and performance outcome to close the evaluation loop. The only question that matters at this stage is whether the course led to the behavioral change you defined at the start. I go deeper on how to measure that here.

When courses aren't enough

In a 1993 article, a group of psychologists made the case for deliberate practice as the foundation of expert performance. What they found is something you're likely familiar with: to build skills, you need structured, feedback-rich practice environments. You need to practice something, repeatedly, to get good at it. 

Courses aren't optimized for that type of skill development. Which is why LinkedIn's 2025 Workplace Learning Report states that "L&D professionals must shift away from the idea that courses are the only solution," and Josh Bersin describes the shift as moving "beyond courses and catalogs to build dynamic enablement engines."

Courses transfer information and build understanding. They are not the optimal format for behavior change. That's why L&D teams are moving away from the rigidity of courses toward a more fluid notion of learning — always on, available in the flow of work, and increasingly shaped by AI.

That doesn't mean you should stop building courses, or delete your content library. What do you think is going to provide the grounding for all those personalized learning paths AI generates? Instead, I want you to think of course creation, and content creation more broadly, as part of your role as knowledge architect.

Here's the shift that matters: in an AI-driven learning ecosystem, the only lever you may actually control is the quality of your inputs. Content needs to be designed to stand alone with enough context, clarity, and focus that it can be pulled apart into discrete lessons, stitched together with other content, surfaced as a response to an internal LLM query, or assembled into a personalized learning path without losing its meaning. 

A lesson that only makes sense inside a specific course, in a specific sequence, won't survive that kind of learning environment; a lesson designed around a single, well-defined outcome will.

It's going to take time and tooling to build the ecosystem being described. In the meantime, your job is to prioritize culling content that is outdated or irrelevant, and to focus on building an agile content system that can be quickly updated and repurposed. That all starts with effective learning design.