When the demo runs, Éloi speaks in three different languages while the lips and jaw keep pace with the audio, and the mouth shapes read as a promising match for natural human articulation. That is the heart of the project: the system is built to copy the way a person's mouth moves while talking, and the 17 degrees of freedom spread across the lips and jaw are what allow each sound to get its own shape. English words such as 'Achieve,' 'Core,' and 'Miracle' come out with distinct articulation, and Korean and Japanese lines get the same treatment, since the speech synthesis is designed to work across multiple languages.
Construction-wise, the head mixes hard structure with soft surfaces. A perforated metallic cranial plate sits above articulated purple eyeball assemblies fitted with mechanical eyelids and eyebrows, while the lower face and mouth are covered in flexible light purple skin. The whole head rides on a multi-axis motorized neck mechanism that includes internal wiring and metal linkage rods.
What stands out to me is how the flexible mouth area and the rigid plate above it are treated as one expressive system rather than two separate parts. The open question is also worth respecting: the maker plans to show the robot's expressive flexibility next and wonders whether a non-human character can still offer meaningful interaction. That seems like the right thing to be unsure about at this stage.