The results are not accurate. Although the system accurately tracks the unmistakable sound of a person's voice and is often easily understood, the synthesizer may generate garbled words. It is still miles better than previous approaches that did not attempt to replicate the vocal tract. Scientists are also testing denser electrodes at the brain interface and more sophisticated machine learning, both of which could improve overall accuracy. This would ideally work with any person, even if they can not train the system before it is put into practice.
This effort may take a while, and at this time there is no fixed roadmap. At least the goal is clear: The researchers want to revive the voices of people with ALS, Parkinson's and other diseases where speech loss is usually irreversible. When this happens, it can dramatically improve communication for patients (who today need to use much slower methods) and help them feel more connected to society.