Add denoiser support for Google STT plugin

[Panmax]

Summary

• Add denoise_audio and snr_threshold parameters to Google STT plugin to support audio denoising and SNR filtering
• Upgrade google-cloud-speech dependency from >= 2 to >= 2.33 to enable DenoiserConfig support

Description

This PR adds support for Google Cloud Speech-to-Text's denoiser and SNR filtering features, which help improve transcription accuracy in noisy environments.

New parameters:

• denoise_audio (bool): Enables audio denoising to reduce background noise such as music, rain, or street traffic. Note: cannot remove background human voices.
• snr_threshold (float): Controls the minimum loudness of speech required for transcription. This helps filter out non-speech audio or background noise. Recommended values:
  • 10.0 - 100.0 when denoise_audio=True
  • 0.5 - 5.0 when denoise_audio=False

Usage example:

from livekit.plugins.google import STT

stt = STT(
    model="chirp_3",
    location="us",
    denoise_audio=True,
    snr_threshold=20.0,  # medium sensitivity
)

Changes

• livekit-plugins/livekit-plugins-google/pyproject.toml: Updated google-cloud-speech version requirement
• livekit-plugins/livekit-plugins-google/livekit/plugins/google/stt.py:
  • Added denoise_audio and snr_threshold fields to STTOptions
  • Added build_denoiser_config() method to STTOptions
  • Updated STT.__init__() with new parameters
  • Updated _build_recognition_config() to include denoiser config for V2 API
  • Updated SpeechStream._build_streaming_config() to include denoiser config
  • Updated update_options() methods in both STT and SpeechStream classes

References

• Google Cloud Speech-to-Text Chirp 3 Documentation

Summary by CodeRabbit

• New Features

  • Added audio denoising controls for Google Speech-to-Text: enable/disable denoising and set signal-to-noise ratio threshold to reduce background noise during transcription.

• Chores

  • Updated minimum Google Cloud Speech client version requirement to ensure compatibility with the new denoising features.

[coderabbitai]

📝 Walkthrough

Walkthrough

Added denoise_audio and snr_threshold options to Google STT interfaces and options; implemented STTOptions.build_denoiser_config(); updated V2 recognition and streaming config assembly to inject DenoiserConfig when present; bumped google-cloud-speech minimum to 2.33.

Changes

Cohort / File(s)	Summary
Google STT Denoising & Options livekit-plugins/livekit-plugins-google/livekit/plugins/google/stt.py	Added denoise_audio: NotGivenOr[bool] and snr_threshold: NotGivenOr[float] to STTOptions; new build_denoiser_config() returning `cloud_speech_v2.DenoiserConfig
Dependency Update livekit-plugins/livekit-plugins-google/pyproject.toml	Raised google-cloud-speech lower bound to >= 2.33, < 3 to require denoiser-related types/fields.

Sequence Diagram(s)

sequenceDiagram
  participant Client as Client
  participant STT as STT
  participant SpeechStream as SpeechStream
  participant GoogleAPI as GoogleAPI

  Client->>STT: create/update (denoise_audio, snr_threshold)
  STT->>STT: update STTOptions with denoising fields
  STT->>SpeechStream: propagate updated options
  SpeechStream->>STT: request recognition config
  STT->>STT: build_denoiser_config() -> DenoiserConfig?
  STT->>GoogleAPI: StreamingRecognitionConfig (RecognitionConfig + optional DenoiserConfig)
  GoogleAPI-->>SpeechStream: streaming transcripts/events
  SpeechStream-->>Client: deliver transcripts

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• feat (google STT): support profanity filter #4573: Modifies the same STT implementation adding per-request options and updating recognition/streaming config builders.

Suggested reviewers

• Hormold

Poem

🐰
Ears perk up for cleaner sound,
Hopping bytes now swirl around.
Denoise whispers, SNR in tow,
Transcripts clear where circuits go.
Hoppity-hop — the streams glow.

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 25.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main change: adding denoiser support (including denoise_audio and snr_threshold parameters) to the Google STT plugin. It is concise, specific, and clearly reflects the primary objective of the changeset.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

Tip

🧪 Unit Test Generation v2 is now available!

We have significantly improved our unit test generation capabilities.

To enable: Add this to your .coderabbit.yaml configuration:

reviews:
  finishing_touches:
    unit_tests:
      enabled: true

Try it out by using the @coderabbitai generate unit tests command on your code files or under ✨ Finishing Touches on the walkthrough!

Have feedback? Share your thoughts on our Discord thread!

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@livekit-plugins/livekit-plugins-google/livekit/plugins/google/stt.py`:
- Around line 323-343: The diagnostic warns about using a bare dict annotation
for recognition_config_kwargs; update its type to a parameterized Mapping/Dict
with concrete key/value types (e.g., Dict[str, Any] or Mapping[str, object]) so
mypy strict mode doesn't infer Any; locate the variable
recognition_config_kwargs in the function that constructs and returns
cloud_speech_v2.RecognitionConfig (the block that builds
recognition_config_kwargs and returns
RecognitionConfig(**recognition_config_kwargs)) and change the annotation there
(and similarly for the other dict usage around the 555-576 region) to a properly
parameterized type such as Dict[str, object] or Mapping[str, Any].

📜 Review details

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 46d7057 and fd9ff2c.

📒 Files selected for processing (1)

• livekit-plugins/livekit-plugins-google/livekit/plugins/google/stt.py

🧰 Additional context used

📓 Path-based instructions (1)

**/*.py

📄 CodeRabbit inference engine (AGENTS.md)

**/*.py: Format code with ruff
Run ruff linter and auto-fix issues
Run mypy type checker in strict mode
Maintain line length of 100 characters maximum
Ensure Python 3.9+ compatibility
Use Google-style docstrings

Files:

• livekit-plugins/livekit-plugins-google/livekit/plugins/google/stt.py

🧬 Code graph analysis (1)

livekit-plugins/livekit-plugins-google/livekit/plugins/google/stt.py (2)

livekit-agents/livekit/agents/utils/misc.py (1)

• is_given (25-26)

livekit-agents/livekit/agents/stt/stt.py (1)

• model (115-124)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)

• GitHub Check: unit-tests
• GitHub Check: type-check (3.13)
• GitHub Check: type-check (3.9)

🔇 Additional comments (5)

livekit-plugins/livekit-plugins-google/livekit/plugins/google/stt.py (5)

81-84: Good addition of denoise/SNR fields with NOT_GIVEN defaults.
Keeps backward compatibility while exposing the new options.

---

119-134: DenoiserConfig builder cleanly scoped to V2 + optional inputs.
The guardrails look solid and avoid V1 misuse.

---

156-193: Constructor docs and config wiring for denoise/SNR look solid.
Clear parameter behavior and correct propagation into STTOptions.

Also applies to: 247-248

---

435-480: Option updates now propagate denoise/SNR to active streams.
Looks consistent with the rest of the option updates.

---

518-547: SpeechStream option updates correctly carry denoise/SNR and trigger reconnect.
The update path stays coherent with other config changes.

_{✏️ Tip: You can disable this entire section by setting review_details to false in your review settings.}