Imagine listening to a physics explanation. The audio is clear and the reasoning is coherent, yet you are still staring at a thin line in the diagram: does it represent the direction of a force, or the motion of an object?
The explanation has moved on, but you need to stay on this page a little longer.
For AI companion learning, being able to speak the content is only a starting point. The pace of the explanation, how the voice is presented, and whether answers stay anchored to the page in front of the learner all affect whether they can keep participating. EvoClass made ongoing adjustments around these questions in August–September 2026; this article reviews the relevant updates through September 20.
Different roles need distinguishable voices
Choosing a companion character should not stop at a change in appearance. Voice is also part of sustained conversation.
An early-August update gave the spoken explanations of different companion characters corresponding accents, so that choosing a character now also brings a difference in voice. Volume control added afterward lets learners adjust output for headphones, speakers, and their surroundings.
These changes offer room to choose; they do not mean one kind of voice is inherently better for everyone. Whether something is clear and comfortable still needs to be judged against your own listening environment.
Giving the pace back to the person learning
The same piece of content may call for different speeds on a first encounter and on a second review.
Speech-rate control, added in September, lets learners adjust the pace of explanations in the companion interface. When encountering a new concept, they can slow down; when reviewing familiar material, they can choose a speed that suits them better.
The point of this capability is not to compress a lesson as short as possible. Speed should serve understanding: hear one condition clearly, then understand a diagram, then move to the next step. The actual listening experience may differ between voices, and adjusting speech rate does not mean learning outcomes will necessarily improve.
When asking about "here," the AI needs to know where you are looking
There is another kind of mismatch, unrelated to speed.
A learner asks, "Why is this step done this way?" while looking at the current page; if the answering system does not have the same page information, it easily produces an explanation that is too broad.
The September 20 AI Q&A update added information about the current courseware page and standardized page-number expressions across the related interfaces. Voice control in AI companion learning and page context in AI Q&A are two adjacent but distinct capabilities: the former adjusts how something is explained, the latter helps clarify what is being discussed.
Having page context does not mean the AI can reliably understand every image detail, or give the correct answer every time. Learners can still point to specific terms, steps, or locations in a diagram, and check explanations against the original courseware.
Good companion learning needs room for pauses
A learning process worth continuing should allow a person to stop: turn the volume down, slow the explanation, return to the same page, and ask a more specific question again.
This round of updates adds more concrete support for these actions. It does not remove the learner's own judgment; rather, it makes that judgment easier to exercise in the middle of a discussion.
You can enter a course you have access to from the course catalog and look for the available companion learning and Q&A entry points in class. Pick a page you genuinely have not understood, adjust to a comfortable pace, and then ask about one specific step—it is easier to verify than asking the AI to explain everything at once.