This is a nightly release of Vibepollo and contains several bug fixes and may contain new features. Check the commit log for more information.
Windows installer · Linux (Arch x86_64)
Nightly Windows installers are unsigned and include self-signed gamepad components. They may cause false-positive antivirus alerts.
Source: 8b3ea1c4a41c
Since the preceding successfully published nightly Commit log
2.0.1-daily.d20261012.1: compare exact sources.
perf(amf): minimize native output latency
perf(amf): deliver ready output without idle polling
fix(packaging): read prebuilt driver version from INF
perf(amf): default to speed and shorten capture GPU handoffs
Preserve explicit quality choices while aligning backend, settings UI, documentation and translated defaults. Copy fresh cursor frames directly to the shared encoder image before saving the cursor-free desktop cache. Release WGC capture ownership before diagnostics, move encoder border setup outside the capture mutex, and reject abandoned or failed GPU handoffs. Add behavioral ordering and default-consistency regression tests.
fix(amf): verify tile requests and clarify split-frame hints
Merge remote-tracking branch 'origin/master' into codex/amf-low-latency-fast-path
perf(amf): deliver capture-wait completions before conversion
build(windows): propagate FFmpeg BCrypt dependency
perf(windows): opt streaming processes out of power throttling
Windows 11 can give windowless processes EcoQoS scheduling and ignore their timer-resolution requests. Both apply to the host while streaming and to the WGC helper. Request full QoS and honored timer resolution for the host during a stream, restoring system control when streaming stops, and for the helper for its lifetime. Windows 10 falls back to the execution-speed policy alone.
perf(capture): raise capture GPU thread priority with the realtime class
When the realtime GPU scheduling class is granted, request the same absolute GPU thread priority the encoder device already uses, so the capture copy that conversion waits on does not queue behind game work. The NVIDIA+HAGS high-class downgrade keeps relative priority 7.
revert(capture): keep the relative capture GPU thread priority
Absolute GPU thread priority 30 is hard realtime and does not yield to
lower-priority work, so with the GPU fully loaded it could cost the game
frame rate or frame-time consistency, and no capture starvation has
been measured that it would fix. The capture device goes back to
relative priority 7 in every scheduling class.
This reverts commit 0d3cb240105f3507827e389536a35ef1320fd1d2.
Signed-off-by: Ramazan Kara <github@r-kara.de>
Merge remote-tracking branch 'upstream/master' into codex/amf-low-latency-fast-path
Conflicts were additive: master's HEVC GDR GOP setup now runs before the
split-frame hint, the WGC snapshot computes master's composition
timestamp after the early keyed-mutex release, and both sides' includes
and parity tests are kept.
Signed-off-by: Ramazan Kara <github@r-kara.de>
Merge current master while preserving installer policy
Scope AMD latency optimizations and prefer native with attributed initialization rollback
End stream after failed AMD rollback and name initialization failure stages
Preserve initialization exception diagnostics and exclude D3D device loss from rollback
Require legacy support for negotiated decoder reference limits before AMD rollback
Reject ready AMF handoffs after deadline or cancellation and report failing policy cases
Coordinate AMF timeout ownership tests without scheduler sleep assumptions
Retain the latest converted AMF image when ready output skips its submission
Merge pull request #461 from RamazanKara/codex/amf-low-latency-fast-path
perf(amf): minimize native output latency