fix(audio): bump rusty-opus to 1.x for working in-band FEC #1
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/rusty-opus-1"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
qrtc enables in-band FEC on its Opus encoder (
use_inband_fec = true,packet_loss_perc = 10inAudioSender::new). Withrusty-opus0.9.1 the encoder's LBRR data is malformed, so libopus decodes the stream as loud garbage and the far end hears constant crackling.rusty-opus1.0.0 fixes the in-band FEC encoder, but qrtc's"0.9"requirement cannot resolve to it. This bumps the requirement to"1"and refreshes the lockfile (1.0.2). No source changes.Upstream issue: https://github.com/crabnebula-dev/qrtc/issues/16
Evidence
Same 5 s signal encoded by
rusty-opus(20 ms frames, 48 kHz mono, VoIP, 32 kbps, FEC on) and decoded with the system libopus. Output RMS, input RMS 0.048 (quiet) and 0.196 (normal):normal: 0.218
normal: 0.195
With FEC off both versions decode correctly (0.045 and 0.196), so only the 0.9.1 FEC path is broken. With louder speech-like input the 0.9.1 error was 18-24x too loud.
Real call: a Tauri build of RKD against this branch (FEC left on) was clean on a phone, including loud speech; the same build against
09fb521crackled constantly.cargo test --lib: 34 passed, 0 failed, 1 ignored.Merge danger
[patch]). Upstream is untouched.Not verified / follow-ups:
sbom/qrtc.cdx.jsonstill lists rusty-opus 0.9.1, so upstream's "SBOM matches Cargo.lock" CI job would fail on this commit (needscargo-cyclonedx0.5.9 andnode tests/sbom-check.mjs --update).cargo denywas not run.