Intent to Ship: Web Speech API contextual biasing

[Chromestatus]

Contact emails

y...@google.com, ev...@google.com

Explainer

https://github.com/WebAudio/web-speech-api/blob/main/explainers/contextual-biasing.md

Specification

https://webaudio.github.io/web-speech-api/#speechreco-phraselist

Design docs

https://docs.google.com/document/d/1AW9DxKYNfnt_wruxmd-_M1islS-MgsUVpaNBen-XvHk

Summary

This feature enables websites to support contextual biasing for speech recognition by adding a recognition phrase list to the Web Speech API. Developers can provide a list of phrases as well as updating them to apply a bias to the speech recognition models in favor of those phrases. This helps improve accuracy and relevance for domain-specific and personalized speech recognition.

Blink component

Blink>Speech

Search tags

speech, recognition, local, offline, on-device, contextual, biasing

TAG review

None

TAG review status

Pending

Risks

Interoperability and Compatibility

None

Gecko: Positive (https://github.com/mozilla/standards-positions/issues/1255)

WebKit: No signal (https://github.com/WebKit/standards-positions/issues/517)

Web developers: No signals

Other signals:

WebView application risks

Does this intent deprecate or change behavior of existing APIs, such that it has potentially high risk for Android WebView-based applications?

None

Debuggability

None

Will this feature be supported on all six Blink platforms (Windows, Mac, Linux, ChromeOS, Android, and Android WebView)?

No

Initially supported on Windows, Mac, and Linux with ChromeOS support to follow.

Is this feature fully tested by web-platform-tests?

Yes

https://wpt.fyi/results/speech-api?label=experimental&label=master&aligned

Flag name on about://flags

Finch feature name

WebSpeechRecognitionContext,OnDeviceWebSpeech,OnDeviceWebSpeechAvailable

Rollout plan

Will ship enabled for all users

Requires code in //chrome?

False

Estimated milestones

Shipping on desktop

140

Anticipated spec changes

Open questions about a feature may be a source of future web compat or interop issues. Please list open issues (e.g. links to known github issues in the project for the feature specification) whose resolution may introduce web compat/interop risk (e.g., changing to naming or structure of the API in a non-backward-compatible way).

None

Link to entry on the Chrome Platform Status

https://chromestatus.com/feature/5225615177023488?gate=4780829605363712

This intent message was generated by Chrome Platform Status.

[Daniel Clark]

> Design docs
> https://docs.google.com/document/d/1AW9DxKYNfnt_wruxmd-_M1islS-MgsUVpaNBen-XvHk

I don’t have access to this document. Could it be made available to non-Googlers?

> TAG review

> None

Can you say why no TAG review?

> Gecko: Positive (https://github.com/mozilla/standards-positions/issues/1255)

I don’t see Gecko taking any position on that thread.

> WebKit: No signal (https://github.com/WebKit/standards-positions/issues/517)

Thanks for engaging with Marcos’ feedback on the WebKit thread. It looks like most of his feedback has now been addressed, so you might consider sending a ping giving that update, with the context that you’re now requesting this to ship in Chromium.

> Is this feature fully tested by web-platform-tests?

> Yes

> https://wpt.fyi/results/speech-api?label=experimental&label=master&aligned

I couldn’t find much test coverage at this link for this addition -- just the one line in https://github.com/web-platform-tests/wpt/blob/master/speech-api/SpeechRecognition-basics.https.html that confirms the phrases​​ array exists and is initially empty. I did see there was an end-to-end manual test at https://github.com/web-platform-tests/wpt/blob/master/speech-api/SpeechRecognition-phrases-manual.https.html which is great. Maybe there’s more that can be added as automated coverage though? For example an automated WPT could check that SpeechRecognitionPhrase​s can be constructed, that they can be assigned to phrases​, that boost​ values outside of the 0.0-10.0 range throw errors, etc.

Thanks,

Dan

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/686eb087.2b0a0220.3a1521.0f8f.GAE%40google.com.

[Evan Liu]

Hi Dan,

Thank you for the thoughtful feedback. Addressing your points:

Design Doc Access: The original document contains some internal information that can't be shared. I've created a public version of the design document that you can access here: https://docs.google.com/document/d/1CBH4r6rxSryY28hOhBQC-DZgjwWbO-FRYg_bbJIfBFo

TAG Review: Given the relatively small scope of this API addition, I initially didn't think a TAG review was necessary. However, I'm happy to request one if you think it would be beneficial.

Gecko Position: We've been working closely with Mozilla on this design (pad...@mozilla.com is a collaborator on the spec change). I can ask him to add a formal position to the thread to make it official.

WebKit Feedback & WPT Coverage: Thanks for the suggestion. I've just updated the WebKit thread to reflect the latest changes.

Regarding test coverage, the end-to-end tests are currently manual because the test environment has limitations in accessing the necessary on-device models. I'll look into ways to improve our automated test coverage and will add more tests for edge cases and error handling.

Thanks,

Evan

[Dan Clark]

(re-sending from the correct account)

Thanks for these updates!

I think it's kind of borderline whether there's enough complexity here to merit a TAG review. My preference is to go ahead and request one in case they do have feedback.

-- Dan

[Dan Clark]

Thanks for these updates!

I think it's kind of borderline whether there's enough complexity here to merit a TAG review. My preference is to go ahead and request one in case they do have feedback.

-- Dan

[Evan Liu]

Sounds good, I've opened a TAG issue for this here: https://github.com/w3ctag/design-reviews/issues/1121

Please let me know if you have any other feedback!

[Sangwhan Moon]

Drive by curiosity question: 

Have you considered a dictionary for SpeechRecognitionPhrase?

If mutating individual SpeechRecognitionPhrase instances in SpeechRecognition.phrase is an intended use case and also has side-effects, I recon this wouldn't be desirable - but curious if that was considered.

Also, the ergonomics of modifying the phrase list seems inconvenient (if it is an intended use case) - e.g. unboost and boost a specific phrase would require a sweep to find the phrase if I am understanding the explainer correctly..

(Likely something the TAG might bring up too)

On Jul 15, 2025, at 5:00, 'Evan Liu' via blink-dev <blin...@chromium.org> wrote:



To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOVsCZ%3DyDEmMn1%3DXxd6E-DA3fS5TuPr1jJZExLTZmgoZ1FUPKg%40mail.gmail.com.

[PhistucK]

Does this basically supersede SpeechGrammar?

☆PhistucK

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/E7442B06-6C6D-44E6-9817-D3F4F27289D6%40chromium.org.

[Evan Liu]

Have you considered a dictionary for SpeechRecognitionPhrase?

Early versions of this design included creating a SpeechRecognitionPhraseList interface as well as a SpeechRecognitionContext to encapsulate the contextual biasing phrases. The latest consensus is to accept a suggestion from dom...@chromium.org to use an ObservableArray<SpeechRecognitionPhrase>.

Does this basically supersede SpeechGrammar?

SpeechGrammar has been deprecated in the Web Speech API spec. While the API implementation in Chrome supports SpeechGrammar, this never actually did anything as the backend speech recognition service didn't support the functionality. See the following issues/PRs for more context around this deprecation:

https://github.com/WebAudio/web-speech-api/issues/57

https://github.com/WebAudio/web-speech-api/pull/163 

https://github.com/WebAudio/web-speech-api/pull/32

[Yoav Weiss (@Shopify)]

How likely are different implementations to implement their reaction to different "boost" values similarly?
If I test my game with a "Zoltan" character (from the explainer's examples) in different implementations with the same "boost" values, am I likely to get similar results?

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscribe@chromium.org.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOVsCZ%3DyDEmMn1%3DXxd6E-DA3fS5TuPr1jJZExLTZmgoZ1FUPKg%40mail.gmail.com.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscribe@chromium.org.

[Evan Liu]

I don't think there are any technical limitations that would prevent other implementations from conforming to how the behavior is specified in the spec. We can add more WPT test cases for different phrases and boost values, but I would caution against constraining the underlying models too much because the speech recognition results may potentially change as models improve over time.

+pad...@mozilla.com is currently working on the Firefox implementation and may have other thoughts on this.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOVsCZ%3DyDEmMn1%3DXxd6E-DA3fS5TuPr1jJZExLTZmgoZ1FUPKg%40mail.gmail.com.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/E7442B06-6C6D-44E6-9817-D3F4F27289D6%40chromium.org.

[Evan Liu]

Hi all,

The TAG review for this has been completed. Please let me know if anyone else has any questions or concerns.

Thanks,

Evan

[Chris Harrelson]

LGTM1

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOVsCZ%3Dk2unxUFvRyuRsR%2B33oLJB8ED%2BbrE95W%2BUpfoRK5JtkA%40mail.gmail.com.

[Yoav Weiss (@Shopify)]

LGTM2

I'm not 100% convinced that the "boost" levels would be fully interoperable, but it may be hard to make them so, so I'm not sure that concern is practical to address. If y'all have ideas to increase its interoperability (by documenting more of what it does in the spec, or something similar), I think it'd be useful.

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscribe@chromium.org.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOVsCZ%3DyDEmMn1%3DXxd6E-DA3fS5TuPr1jJZExLTZmgoZ1FUPKg%40mail.gmail.com.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscribe@chromium.org.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/E7442B06-6C6D-44E6-9817-D3F4F27289D6%40chromium.org.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscribe@chromium.org.

[Dan Clark]

LGTM3, and +1 to Yoav's comment.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOVsCZ%3DyDEmMn1%3DXxd6E-DA3fS5TuPr1jJZExLTZmgoZ1FUPKg%40mail.gmail.com.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/E7442B06-6C6D-44E6-9817-D3F4F27289D6%40chromium.org.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAOVsCZ%3Dk2unxUFvRyuRsR%2B33oLJB8ED%2BbrE95W%2BUpfoRK5JtkA%40mail.gmail.com.