Uh oh!
There was an error while loading. Please reload this page.
feat(audio): allow sending continuous audio - #4261
Conversation
feaddf3 to
adcbfd4CompareLGTM, nice and simple, and seems fine as a hidden setting. I guess the only downside is that the audio stream is CBR so all those zeroes still consume 512k/1.5Mb/2Mb. |
Bundle ReportBundle size has no change ✅ |
ReenigneArcher
commented
Oct 11, 2025
I believe this a setting that comes from the client itself, right? |
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@## master #4261 +/- ##
==========================================
- Coverage 11.98% 11.97% -0.01%
==========================================
Files 87 87 Lines 17572 17581 +9 Branches 8076 8082 +6 ==========================================
Hits 2106 2106 - Misses 14569 14578 +9
Partials 897 897
Flags with carried forward coverage won't be shown. Click here to find out more.
|
mariotaku
commented
Oct 11, 2025
@ReenigneArcher Yes. Only when the client requests |
e4b43d3 to
e3069d9Compare
Uh oh!
There was an error while loading. Please reload this page.
ns6089
commented
Oct 13, 2025
It's simple because it has an inherent buffer underrun - buffer is only filled when |
majima-everywhere
commented
Apr 9, 2026
How can I make my client request continuous audio? I think my issue is neither related to sunshine nor moonlight, but to my tv itself (sony x85k, SoC MT5895) that adds a buffer to my audio. I can reduce the latency when I tweak the settings but it is still noticeable. My hope is that sending continuous audio to client might sync the audio and video clock somehow but I can be wrong. I don't understand why the audio is in sync with the host on my phone client despite having a horrible +30ms decoding latency, while I encounter audio lag on my tv connected through hdmi to my pc client with near 0 decoding latency. |
* fix: DXGI adapter memory leak (upstream d3af56d, #4340) * fix: KVM cursor disappearing due to Mouse Keys (upstream fd2bfaa, #4407) * feat: continuous audio support (upstream fbcf211, #4261) * fix: clear tray icon when unpairing the last paired client (upstream d6bc76e3d, #4890) * fix(audio): close TOCTOU between has_audio_ctx_ref and get_audio_ctx_ref When the microphone redirect helpers do has_audio_ctx_ref() followed by get_audio_ctx_ref(), a release between the two calls would re-trigger start_audio_control() and attach mic ops to a brand-new isolated audio context. Add shared_t::try_ref() and audio::try_get_audio_ctx_ref() so init/release/write_mic_data acquire an existing reference atomically and never resurrect a stopped context. Reported-by: coderabbitai (PR #551 review) * fix(audio): satisfy mic-redirect virtual interface on Linux/macOS The upstream mic-redirect feature (init_mic_redirect_device, release_mic_redirect_device, write_mic_data, plus is_sink_available on macOS) declared four pure-virtual methods on audio_control_t but never provided overrides in the PulseAudio (Linux) or AVFoundation (macOS) backends. That left pa::server_t and macos_audio_control_t abstract, so std::make_unique<...>() in audio_control() failed to compile on both platforms (only the Windows backend builds today). Add minimal no-op stubs that return -1 / true / void as appropriate so the non-Windows targets build again. Mic redirect-to-host stays unimplemented on those platforms — wire-up can land later without re-touching every cross-platform site. Reported-by: coderabbitai (PR #551 review)
Description
When host has no audio playing, Sunshine won't send audio. For webOS client, when surround sound is used, video will freeze when there is no audio coming in.
This PR allows the host to send empty audio frame in such case, mitigating webOS freezing issue.
Screenshot
Issues Fixed or Closed
Roadmap Issues
Type of Change
Checklist
AI Usage