07:51:15 <RRSAgent> RRSAgent has joined #web-speech-api
07:51:19 <RRSAgent> logging to https://www.w3.org/2025/11/12-web-speech-api-irc
07:51:19 <breakout-bot> RRSAgent, do not leave
07:51:20 <breakout-bot> RRSAgent, this meeting spans midnight
07:51:20 <breakout-bot> RRSAgent, make logs public
07:51:22 <breakout-bot> Meeting: Web Speech API: New Features & Future Directions
07:51:22 <breakout-bot> Chair: evanbliu, Paul Adenot
07:51:22 <breakout-bot> Agenda: https://github.com/w3c/tpac2025-breakouts/issues/5
07:51:22 <Zakim> Zakim has joined #web-speech-api
07:51:23 <breakout-bot> Zakim, clear agenda
07:51:23 <Zakim> agenda cleared
07:51:23 <breakout-bot> Zakim, agenda+ Pick a scribe
07:51:23 <Zakim> agendum 1 added
07:51:23 <breakout-bot> Zakim, agenda+ Reminders: code of conduct, health policies, recorded session policy
07:51:24 <Zakim> agendum 2 added
07:51:24 <breakout-bot> Zakim, agenda+ Goal of this session
07:51:24 <Zakim> agendum 3 added
07:51:24 <breakout-bot> Zakim, agenda+ Discussion
07:51:25 <Zakim> agendum 4 added
07:51:26 <breakout-bot> Zakim, agenda+ Next steps / where discussion continues
07:51:26 <Zakim> agendum 5 added
07:51:26 <breakout-bot> Zakim, agenda+ Adjourn / Use IRC command: Zakim, end meeting
07:51:27 <Zakim> agendum 6 added
07:51:27 <breakout-bot> breakout-bot has left #web-speech-api
08:20:45 <shiestyle> shiestyle has joined #web-speech-api
08:25:33 <mjwilson> mjwilson has joined #web-speech-api
08:26:50 <kbx> kbx has joined #web-speech-api
08:27:18 <kbx> Present+ Kenji_Baheux
08:28:43 <smaug> smaug has joined #web-speech-api
08:29:04 <ningxin> ningxin has joined #web-speech-api
08:29:55 <Kitahara> Kitahara has joined #web-speech-api
08:30:24 <hagio> hagio has joined #web-speech-api
08:31:30 <cwilso> cwilso has joined #web-speech-api
08:31:37 <cwilso> present+
08:32:12 <JRJurman7> JRJurman7 has joined #web-speech-api
08:32:27 <JRJurman7> Room is muted
08:32:33 <shiestyle> present+
08:35:20 <m-alkalbani> m-alkalbani has joined #web-speech-api
08:37:03 <tako> tako has joined #web-speech-api
08:38:12 <anssik> anssik has joined #web-speech-api
08:38:26 <anssik> q?
08:39:44 <kbx> q+
08:42:00 <anssik> q+
08:42:18 <AramZS> AramZS has joined #web-speech-api
08:42:19 <anssik> q?
08:42:35 <AramZS> present+
08:42:39 <cwilso> ack kbx
08:43:21 <msw> msw has joined #web-speech-api
08:43:28 <Hadrien> Hadrien has joined #web-speech-api
08:43:39 <Hadrien> present+
08:43:44 <handellm> handellm has joined #web-speech-api
08:43:51 <handellm> q?
08:43:51 <AramZS> q+
08:44:00 <cwilso> ack ans
08:44:22 <cwilso> q+
08:45:43 <kbx> AramZS: are there limits on biasing?
08:45:54 <kbx> Evan: there is a limit but it's very high.
08:46:16 <handellm> q+
08:46:17 <kbx> Paul: hash table lookup. complexity in the noise compared to the actual speech reco.
08:46:18 <Hadrien> q+
08:46:21 <cwilso> ack ar
08:46:43 <kbx> cwilso: awesome to see this moving. standardization plans?
08:47:07 <kbx> Paul: timeline not perfect as we just had rechartered. came after.
08:47:18 <kbx> Paul: need to wait for next charter. adopted by the wg.
08:47:44 <kbx> cwilso: if you are ready you can rechartered whenever.
08:47:48 <cwilso> ack cw
08:47:54 <ningxin> q+
08:48:07 <kbx> Evan: tackling the other part of webspeech next year (speech gen)
08:48:39 <kbx> Sushan: our model doesn't have a confidence score.
08:48:45 <kbx> Paul: mozilla has one.
08:49:03 <kbx> Evan: was proposed way back then when accuracy wasn't good.
08:49:16 <kbx> Paul: open an issue to make it clear that it's not always there.
08:49:20 <LaurentLM> LaurentLM has joined #web-speech-api
08:49:49 <kbx> Sushan: on quality, word error rate based?
08:50:32 <kbx> Evan: want to let site choose between different quality; still an exploration area
08:51:00 <kbx> Sushan: parallels with structured output for biasing.
08:51:29 <kbx> Evan: cloud speech API has similar biasing mechanism. Exploring LLM based approach in Chrome; Mozilla already using one.
08:51:36 <cwilso> ack next
08:51:58 <kbx> Handellm (Markus; google): when would you select low or high
08:52:32 <kbx> Evan: could be depending on other features being run, balancing usage of resources; could also be the UA doing the best it can given workload.
08:53:03 <kbx> Handellm: nitpicking but could be power efficient or other labels
08:53:12 <cwilso> ack next
08:54:15 <kbx> Hadrien: TTS related; more is needed. very broken in current state, voice selection issues (quality, gender); voices not listed, bad voices; events issues (pausing not working). ebooks field, users need TTS.
08:54:32 <kbx> ... curious to hear your ideas
08:55:06 <kbx> Evan: been neglected thus far; we will be looking at it from Chrome perspective.
08:55:10 <kbx> Paul: same for Mozilla
08:55:33 <kbx> Hadrien: investment in read aloud type features but none of that is exposed.
08:55:54 <cwilso> ack next
08:55:58 <kbx> Evan: yeah, we also had similar situations with speech to text (e.g. live caption), working on it.
08:56:29 <kbx> Ningxin: quality related; if dev select high, download size; resource consumption, etc
08:57:17 <kbx> Evan: high level at the moment; tradeoff between offering more knobs and being stuck with it in the future; potential fingerprinting bits; lots of TBD; some folks are against adding hints; but there has been requests from dev.
08:58:13 <kbx> Ningxin: in WebML WG, how can dev give hints about what they care about for instance power efficiency (VC / Meet / zoom scenario).
08:58:35 <kbx> Evan: could be part of the hints approach. We now supports multiple streams
08:59:22 <kbx> Ningxin: re customization; users with speech handicaps => users' data for those special cases. Thoughts?
08:59:40 <kbx> Evan: there are teams in google but we haven't worked closely with them yet; can make a note.
09:00:20 <kbx> Paul: likely the models will be open, community could contribute; dataset available to all for building models.
09:00:55 <kbx> Tarek: plans to open the API for pointing at a different model?
09:01:07 <kbx> Paul: we are using different implementations and models
09:01:40 <kbx> Anssi: speech installation method to download packs; built-in AI is similar in that regard; are you talking  with this team?
09:01:42 <kbx> Evan: yes
09:02:13 <kbx> Mike: yes, awesome demo; got to polyfill an example with the Prompt API's audio input; showed the prospect of polyfilling this witt other models;
09:02:15 <kbx> q:
09:02:15 <kbx> q+
09:02:31 <kbx> Anssi: contrast with built-in AI API?
09:02:55 <kbx> Evan: main difference: builtin ai allow to monitor progress, but not for web speech.
09:03:09 <kbx> Evan: we are embracing similar patterns where possible (e.g. anti fingerprinting).
09:03:23 <cwilso> ack next
09:03:38 <kbx> Anssi: improves ergo
09:03:56 <kbx> Kenji: trying to align where possible; web speech API has been a thing for a long time so we can't break it.
09:04:09 <kbx> Sushan: mediastream, timestamps?
09:04:23 <kbx> Paul: on the result. Mediastream as a timed source so it works
09:04:40 <kbx> Sushan: ...
09:04:53 <kbx> Paul: all events have timecode
09:05:11 <anssik> present+
09:06:07 <kbx> Handellm: [...]
09:06:08 <kbx> Guido: discussion tomorrow related to the topic. WebRTC Media joint session.
09:06:31 <kbx> Sushan: rate of reco when using mediastream? faster than the event?
09:06:51 <LaurentLM> LaurentLM has joined #web-speech-api
09:06:51 <kbx> Paul: clocked to realtimesource; audio device tied.
09:07:19 <kbx> Paul: with the proposal for burst you could issue the events as fast or as slow as you want
09:07:22 <kbx> ... control the pace
09:07:30 <msw> q+
09:07:57 <kbx> Evan: Speech synthesis already has local processing option
09:08:02 <kbx> Paul: doesn't require heavy resources
09:08:11 <cwilso> ack next
09:08:47 <kbx> msw (mike): like model quality; quantifiable metric on error rate; would that be a reasonable requirement that dev could provide? other domains that they may want to provide like faster than realtime?
09:09:21 <kbx> Sushan: good idea but challenging; Dev may want Language support despite higher error rate.
09:09:41 <kbx> Sushan: maybe raise an issue.
09:09:59 <kbx> Evan: will get an issue to continue the feedback
09:10:29 <kbx> Paul: https://webaudio.github.io/web-speech-api/
09:11:11 <mjwilson> mjwilson has left #web-speech-api
09:11:20 <shiestyle> shiestyle has left #web-speech-api
09:11:36 <hagio> hagio has left #web-speech-api
09:31:54 <LaurentLM> LaurentLM has joined #web-speech-api
11:03:27 <shiestyle> shiestyle has joined #web-speech-api
13:25:25 <tidoust> tidoust has joined #web-speech-api
13:25:30 <tidoust> RRSAgent, draft minutes
13:25:32 <RRSAgent> I have made the request to generate https://www.w3.org/2025/11/12-web-speech-api-minutes.html tidoust
13:26:01 <tidoust> Zakim, bye
13:26:01 <Zakim> leaving.  As of this point the attendees have been Kenji_Baheux, cwilso, shiestyle, AramZS, Hadrien, anssik
13:26:01 <Zakim> Zakim has left #web-speech-api
13:26:06 <tidoust> RRSAgent, bye
13:26:06 <RRSAgent> I see no action items