Apple modified the audio of its WWDC 2026 keynote video to prevent nearby devices from waking up whenever “Siri” was mentioned. Observers on X identified the change after analyzing spectrogram screenshots of the presentation audio. The technique attempts to defeat wake-word detection by removing the specific acoustic profiles that trigger Apple hardware.
Key details
- Frequency bands: Audio engineers removed signals in the 3kHz, 4kHz, 5kHz, and 6kHz ranges during mentions of the Siri name.
- Hardware targets: The goal was to silence HomePods, iPhones, iPads, and Macs located in viewers’ homes.
- Mixed results: Multiple users reported that their devices still activated during the stream despite these surgical audio cuts.
- Precedent: Amazon implemented a similar “notching” strategy in 2017 for Alexa television advertisements to avoid triggering Echo speakers.
Why it matters
This technical maneuver highlights the friction between voice assistant marketing and the reality of always-on hardware. By attempting to “mute” its own software during a major presentation, Apple acknowledges that its devices cannot yet distinguish between a broadcast and a live user command reliably. The reported failures of this filtering suggest that modern wake-word detection relies on more than just specific frequencies, making it harder for broadcasters or podcasters to talk about these tools without disrupting their audience. For competitors like Amazon or Google, this remains a persistent hurdle for any video content that mentions their specific digital assistants.
Read the full story at MacRumors
Leave a Reply