top of page
Writer's pictureNicole Papaioannou Lugara

AI Avatars for L&D: How to and How NOT To Use Them

Recently, we had a chance to chat about human-looking AI avatars with the awesome team at VMaker. They're the latest L&D-focused vendor to release AI avatars, competing against brands like Synthesia and HeyGen.



As options for these products become more available (and even more affordable), the questions for Learning and Development pros are:


  • Should we use AI avatars?

  • And if so, when?

  • And on the reverse side, when shouldn't we use them?


Here at Your Instructional Designer, we've been using AI avatars for just over a year to build content for clients and our own learning products. These are the use cases and guidelines we've established.


The Right Use Case

Reminder: we're talking specifically about photorealistic human-looking AI avatars in this post. There are others kinds of avatars that leverage AI, such as ReadyPlayerMe avatars, Vyond puppets, and the puppets you can create from vector images with Adobe Animate.


  1. Use It to Create Characters: AI avatars can be used to create characters for your videos. This includes video hosts and narrators, characters for branching scenarios, and so much more.

  2. Use It When You Need to Explain More Than Influence: If you're just going to be creating "click next" training anyway and all you're doing is offering explanation, an AI avatar can add a bit of flair to bring the content to life. If you need something more personable and influential, use a real human.

  3. Use It When You want to Introduce New Content or Provide a Brief Overview: Use avatars to give a consistent voice to introductory videos, where personal connection isn’t as critical.

  4. Use It as a Guide in Procedural Training: For training focused on specific steps (e.g., software tutorials), avatars can narrate the process, allowing you to produce content quickly with minimal personalization.

  5. Use It When Things Change a Lot: Where filming a human actor can be costly and time intensive (filming, wardrobe, makeup, film editing, audio editing, etc), text-to-speech style AI avatar content is easily updated. The ease of updates make it's a great solution for content that needs to be updated frequently. It also makes it a great solution for building prototypes and examples before committing to more serious investments of time and resources.

  6. Use It When You Need Consistency More Than Emotion: For large-scale training programs where a consistent look and feel are needed across multiple topics, avatars offer a reliable format.

  7. Use It When the Content Requires Significant Localization: If the content needs to be reproduced in multiple languages, an AI avatar tool with multiple language options can greatly cut the cost. We still recommend a human review the scripts for accurate translation, but it can speed up and reduce the cost of the process of localization immensely.

  8. Use It as an Add-On When You Wouldn't Have Hired a Human: If there's some extra you could do to support learning that can be made possible by using an AI avatar, do it. For example, we used AI technology to create audio downloads of eBook-like content. With humans, this would have been too expensive to include. With AI avatars, we can add this accessibility feature at minimal additional cost.

  9. Use It When You Have Good Content: A crappy script will be crappy, whether a human or AI is delivering it. Good video content starts with good writing.


Here are some examples from our work:



What To Consider

Each of the factors below can influence how effective AI avatars are in meeting your instructional goals and user experience goals. Consider them early on to help ensure you're using AI avatars as a strategic addition rather than just a novelty.


  1. Audience Preferences

    • Does your audience value a human touch? For learners who prefer engaging with real people, an avatar might feel too impersonal.

    • Will avatars resonate well with different demographics or cultural backgrounds?

  2. Content Sensitivity

    • Is the subject matter personal, sensitive, or emotionally charged? Avatars often struggle to convey empathy effectively in such cases.

    • Does the content require trust-building or rapport, which might be better suited to a real person?

  3. Engagement Level Needed

    • Will learners benefit from seeing facial expressions and body language that feel genuinely responsive?

    • Is there a need for live, two-way interaction that an avatar cannot provide?

  4. Update Frequency

    • How frequently will you need to update this content? Avatars are excellent for frequently changing information due to easy text-to-video editing.

    • Does your budget or timeline benefit from a scalable solution where only text changes need to be made?

  5. Budget and Resource Allocation

    • Do you have the resources for quality avatars with realistic visuals and accurate lip-syncing?

    • How much cost/time savings will avatars actually provide versus traditional video production?

  6. Technical Limitations and Support

    • Are there tech barriers, like bandwidth or platform compatibility, that might impact avatar performance on various devices?

    • Do you have access to tech support to troubleshoot if avatar rendering or playback issues occur?

  7. Brand Alignment

    • Does an AI-generated avatar align with your organization’s values and brand personality? For brands that emphasize personal connection, a human trainer might be a better fit.

    • Does the avatar style, voice, and demeanor reflect your brand’s identity accurately?

  8. Realism and Quality of Avatars

    • Are the avatars realistic enough to avoid distraction or disengagement? Low-quality avatars may detract from the learning experience.

    • Do the available avatars pronounce technical terms correctly, or will you need extensive customization?

  9. Language and Accessibility Considerations

    • Will avatars support the necessary languages, dialects, and accents required for your diverse audience?

    • Are accessibility features like captions and transcripts easy to implement for avatar-based content?

  10. Learner Feedback and Iteration

    • Can you test avatars with a small group of learners to get feedback before a full-scale launch?

    • Do you have a plan to iterate based on learner feedback to improve avatar effectiveness in your courses?


Dos and Do Nots

Do:

  • Use Avatars for Efficiency: When speed and consistency are priorities, avatars can streamline production.

  • Customize Appropriately: Tailor avatars to reflect the demographic and professional context of your audience. Choose avatars that align with your organization’s brand and ethos.

  • Balance with Human Touch: Supplement avatars with real people in Q&A sessions, discussions, or case studies to foster connection.

  • Test Avatars with Your Audience: Gather feedback on how the avatars are perceived to ensure they resonate positively and support engagement.


Don’ts:

  • Don’t Use for High-Emotion Topics: Avoid avatars where human warmth and empathy are essential. For example, training on empathy-based skills or DEI-related topics can benefit from real people to reflect the emotional nuances involved.

  • Don’t Skip Quality Assurance: Ensure avatars pronounce complex industry-specific terms accurately and check scripts for clarity.

  • Don’t Overuse: Over-relying on avatars can make training feel robotic or impersonal; balance with real instructors for a more engaging experience.

  • Don’t Ignore Cultural Nuances: AI avatars should reflect diverse backgrounds carefully and respectfully to avoid stereotyping or cultural insensitivity.


FAQ

Are you using other tools to create the video, or are you using only the video editors attached to the AI avatar tools?


Yes, post production enhancements can be a great way to really make your video content shine and avoid the passive experience of a too-long talking head video. We often use elements designed with Canva and Camtasia in conjunction with our AI avatar content.


Why don't we often see videos with AI avatars that more than 10 minutes long?


Two big reasons:


1) Many of these AI avatar generator tools set limits on how long your videos can be-- either with credit, token, or scene limits.


2) Talking head content gets dull, especially with an avatar that has less expression and variety in their modes of communication. It's great in snippets, but most of us will quickly get bored listening to someone drone on for 10+-- human or not.


Can I connect the avatar to a GPT to create an interactive chatbot-like tool?


Yes, several of the AI avatar tools we mentioned at the beginning of this post will allow for this kind of customization if you pay extra. It is not a standard feature.


That said, cost is most likely to hold the solo tinkerer back. Aside from the cost of tool upgrades, you will also likely need to pay for an enterprise grade LLM and/or per token cost for inputs and outputs. If you have an active chatbot or one that takes a lot work to get accurate responses from, you can easily rack up several thousands of dollars in bills.


For an example of what this might look like, check out this digital twin experiment from @LearnWithMyra (Myra Roldan).


You can download the slides from the VMaker hosted webinar here.


 

Want to work with us to lead your team into the future?


Commentaires


bottom of page