Can we customize the microphone timeout for web speech api #137

Activity

1. nicovalencia commented on May 10, 2025

   

   nicovalencia

   on May 10, 2025

   Last edited by nicovalencia

   More actions

   I'm seeing Chrome throw a no-speech error if nothing is detected in 10 seconds. I don't see anything in the spec to set this timeout though, and if you want to keep it open for longer and try to restart the recognition instance on a no-speech catch, it errors because it's already been started.

   I'm also seeing on iOS, it stays open and will not throw a no-speech error.

   The use case here is to have an open mic for talking with an assistant. I would like it to stay open, even when an assistant is responding so the user can interrupt or respond to subsequent questions, and if it's automatically erroring after 10s, this cannot work.

   Maybe there is another approach I'm missing?

   Edit: I noticed this only happens the first time the recognition instance is started without detecting anything. The workaround I have right now is to catch the no-speech error, teardown the existing recognition instance and any handlers, then build a new one and start it. For some reason, the second one doesn't seem to be throwing any errors if you leave it open (just like how Safari handles it).

   👍React with 👍1Reacted by Aleix Suau