On July 24th, I attended a “Smart Experience Camp – Crowdtesting Session for the Full-Voice In-Car System” event held by XPeng Motors in Beijing, where I used and tested the highly anticipated full-voice in-car system on the XPeng P7.
After nearly 50 media outlets and users rated the system, the full-voice in-car system on the XPeng P7 received the highest score.
During the event, I asked a number of questions to XPeng Motors’ AI product senior expert, Chaos. Here are the summarized answers:
- What does “24K sampling technology” mean? Simply put, it means that the sound quality of the voice package is better and eliminates the “2G call” quality of the past. Additionally, the data size is larger.
- The voice package is downloaded locally by the user through usage. Since data is directly accessed from local storage, it will become faster to use over time. However, the difference is relatively small (a few milliseconds).
- AI voice is an innovative technology project and is in a blank technological area. In terms of iterative upgrades, user feedback accounts for 30%, with most of the work done by engineers innovating new features.
- The release time for G3i and P5 has not been publicly announced.
- Since the 8155 has a stronger computing power, its utilization rate is lower than that of the 820A. However, the CPU usage for voice itself is already very low, even after an update, it only uses up about 1% of the P7’s usage rate.
- XPeng P already has the permanent wake-up function built in, but in order to ensure a good experience and prevent false triggering, it has limited applications.
- For certain physical command operations, the car’s system will respond without a voice command, eliminating the need for voice replies. However, for certain random conversations, the system may choose to filter or block certain responses. (This is why sometimes XPeng will shake its head when you talk to it, but can still execute tasks like turning on the air conditioning).
- This time, the AI voice is fully synthesized, not a custom recording like Baidu maps or Amap’s “celebrity voice package”.- Currently, we do not offer customized voice options for users due to weak demands and subpar recording effects (which cannot compare with the realistic effects of AI or celebrity voiceover customization).
XPeng has already started releasing Smart 2.6.1 beta versions for some P7 users. Don’t forget to check for updates for your own P7. 🍉🍉🍉
Information source: XPeng official
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.