fix(denoise-tester): play models in both ears; match the call's DeepFilterNet strength #483
No Reviewers
Labels
Clear labels
a11y
area: appearance
area: auth-session
area: build-ci
area: calls
area: desktop
area: media
area: messaging
area: mobile
area: moderation
area: navigation
area: notifications
area: settings
area: threads
bug
dependencies
docs
duplicate
enhancement
help wanted
invalid
needs-human-review
performance
planning
priority: critical
priority: high
priority: low
priority: medium
qa
question
research
security
tech-debt
ux
wontfix
Accessibility: keyboard, screen reader, contrast, motion
Client area: appearance
Client area: auth-session
Client area: build-ci
Client area: calls
Client area: desktop
Client area: media
Client area: messaging
Client area: mobile
Client area: moderation
Client area: navigation
Client area: notifications
Client area: settings
Client area: threads
Something is not working
Third-party package versions and advisories
README / LOTUS_* docs wrong or missing
This issue or pull request already exists
New feature
Need some help
Something is wrong
Re-render storms, leaks, heavy work on hot paths
Data loss, security hole, or crash on a main path
Broken feature or serious usability problem
Minor issue or polish
Wrong behaviour in an edge case or notable degradation
Manual QA: shipped, needs a human in a real environment
More information is needed
XSS, unsafe URLs, data leaks, auth/session
Code health, dead code, fragile patterns
Usability or visual inconsistency
This won't be fixed
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: LotusGuild/cinny#483
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Fixes two problems found while testing with "Record & compare".
RNNoise and Speex only played in the left ear
Chrome's Opus/WebM recordings often decode as stereo. The models only process channel 0, and the default
channelCountMode: "max"ignoreschannelCount: 1. As a result the node outputs two channels with only the left one filled.The fix adds a
monoDownmix()node before the models, in both playback and the live monitor. The RNNoise, Speex and gate nodes are also pinned to explicit mono. Lotus Call already did this (#24), so real calls weren't affected.Verified in Chromium with a stereo speech clip through RNNoise:
DeepFilterNet sounded choppy at word ends
The tester ran DFN at attenuation limit 80, while calls used 60. Both mute the background outright, so words drop into dead silence. This now uses 25, matching LotusGuild/element-call#65; the measurements are in that PR.
🤖 Generated with Claude Code
https://claude.ai/code/session_017PCb59uY3UKqDqt7B6hzd5