ExoSnap

Windows-native screen, application, and region recorder with a native NVENC pipeline. MKV-first recording with on-stop MP4 remux (faststart), multi-track audio routing, webcam PiP overlay, crash recovery, and built-in diagnostics. Dark-mode-first Qt 6 UI.

Status / current release

• Latest release: 0.8.1 — "Diagnostics as a feature + robust MSI packaging".
• Pre-v1 notice: settings, preset, and recording-history schemas may change incompatibly before 1.0.0.
• Platform: Windows 10/11 x64 (Windows 11 is the primary target; Windows 10 is best-effort).
• Hardware encoder: NVIDIA NVENC only. An NVIDIA GPU with supported NVENC capability is required. AMD, Intel, and software encoding are not yet supported.
• Distribution: portable ZIP and MSI. Builds are not yet code-signed (see Code signing below); Windows SmartScreen may warn on first launch.
• Privacy: no analytics and no background telemetry. Recordings and settings stay local. The only network features are opt-in and consent-gated: an update check against public GitHub Releases (off by default for self-built binaries) and a privacy-scrubbed crash report you explicitly choose to send (Stage 0 assisted GitHub issue, or Stage 1 automated upload to Sentry with EU data residency). See PRIVACY.md.

Features (0.8.1 — diagnostics wave + robust MSI packaging)

• Diagnostics as a feature — a first-class diagnostics engine with a typed FixAction model (Auto / Assisted / External) so each detected issue carries the safest concrete remedy.
• Pre-flight readiness gate — surfaces notices and hard blockers before a recording starts, so a misconfigured capture is caught up front rather than failing mid-record.
• Live pipeline monitoring — frame drops, A/V drift, and disk-fill ETA at low cost, with encoder-vs-capture-vs-disk-bound classification and root-cause correlation (e.g. VRR-vs-CFR judder).
• Post-flight report card — frame-drop %, peak A/V drift, and pipeline health after each recording.
• PresentProvider with PresentMon — an opt-in, elevation-gated provider for tearing/game-present observation feeding the judder correlation.
• In-app updates — the Settings Updates card is wired to the consent-gated GitHub-Releases flow.
• Settings polish — reliable hover popovers (info-i and the countdown chevron), corner-anchored compare hints, and a de-nested Hotkeys card in line with the v10 canon.

From the 0.7.0 codec & color wave

• HEVC video (NVENC) in MKV and MP4 (hvc1 sample entry), alongside H.264 and AV1.
• 10-bit output (P010) for HEVC Main10 and AV1 — SDR at higher precision (no HDR transfer curve).
• BT.709 color metadata on every MKV/MP4 output, plus a selectable Y'CbCr color range (Full/Limited). Per-display HDR capability is surfaced in Diagnostics.
• Recording-error dialog — a modal failure report with an opt-in, privacy-scrubbed GitHub issue.
• Theme picker as four live preview cards (two dark, two light); single-click source selection commits and returns straight to the preview.

HEVC, hvc1, and 10-bit encoder paths are functional end-to-end but not yet validated across all NVIDIA GPU generations under live recording. Use H.264 or AV1 if you encounter issues.

From the 0.6.0 Audio v2 wave

• Per-track gain & mute and a brickwall limiter (on by default) on the mixed audio bus, so per-source levels can exceed full scale without hard clipping.
• Microphone DSP chain — high-pass filter, noise gate, AGC, and RNNoise neural noise suppression — each stage off by default (capture is byte-identical when disabled), toggled individually.
• Lossless audio: uncompressed PCM and compressed FLAC (both MKV-only), alongside Opus and AAC.
• Channel / sample-format model (ADR 0030): selectable output sample rate (44.1 / 48 / 96 kHz), mono or stereo, and 16/24/32-bit depth for the lossless codecs, plus a FLAC compression level control. Opus stays at 48 kHz; capture is resampled/rematrixed once after the mix bus.

From the 0.5.0 settings & media-capability wave

• TOML preset store with profile export/import: presets are human-readable and shareable; the nested [[audio.sources]] model serializes cleanly. Flat app settings stay on QSettings.
• Canonical video rate control — Constant quality (CQ), Variable bitrate (VBR), Constant bitrate (CBR) — mapped per encoder underneath (NVENC), plus a bitrate control. Never mislabeled as "CRF".
• Audio encoding controls: bitrate plus Opus frame duration and complexity.
• Settings redesign with a Default/Expert split: common controls up front, expert rate-control, bitrate, frame-timing, and audio DSP placeholders behind an Expert toggle, with inline info hints and search.
• Curated themes (two dark, two light) replacing the free accent picker, so theme colors never collide with status colors.
• Automatic split by size (in addition to by time): the first limit reached wins; MP4 split produces one remuxed progressive segment each.
• Preset manage dialog (rename, duplicate, delete, set default).

Carried over from earlier waves

• Local-first crash capture (out-of-process Crashpad) with a next-launch, privacy-scrubbed, consent-gated crash dialog; two-stage delivery (Stage 0 assisted GitHub issue, Stage 1 opt-in Sentry upload with EU data residency under the official-build gate).

• Update checking with Stable and Preview channels and a signed update manifest (ed25519 via Monocypher + SHA-256, downgrade/rollback protection); updates off by default for self-built binaries.

• Capture-excluded on-screen overlays (recording status, diagnostics, countdown) — visible on screen but excluded from the captured frame via WDA_EXCLUDEFROMCAPTURE.

• Tray icon with idle/recording/paused states and an unread notification badge.

• Notification toasts: saved, low storage, unexpected stop, recovery available.

• Close-to-tray (opt-in); opt-in interactive quick-control pill overlay.

• Refined region-selection overlay and capture-frame dock button.

• MKV-first recording; MP4 delivered via libavformat remux-on-stop (progressive, faststart); MKV and WebM are also supported as direct output formats.

• Crash recovery: a recovery manifest written before each session, plus a startup recovery UI that lets you finish, recover, or discard each interrupted recording on the next launch.

• Low-disk guard: warning at a configurable soft threshold, hard stop at a lower threshold; for MP4 sessions the effective threshold accounts for the transient MKV and output MP4 coexisting during remux.

• FAT32 output-volume detection: a Diagnostics Notice about the 4 GiB per-file limit (recording is not blocked; short clips on FAT32 work correctly).

• Container/codec compatibility registry: invalid combinations are blocked before recording starts.

• Recording split with per-segment background MP4 remux for split MP4 sessions.

• Global hotkeys, single-frame capture, webcam PiP with mirror, live A/V drift and size overlay.

See KNOWN_LIMITATIONS.md for the precise current support boundary.

Product defaults

Setting	Default
Theme	Dark mode
Container	MKV
Video codec	AV1 (NVENC)
Audio codec	Opus
Frame rate	CFR 60 fps
Audio source order	APP, SYS, MIC — all enabled, all separate tracks
Navigation pages	Record, Video, Audio, Output, Webcam, Hotkeys, Diagnostics, Logs, Advanced

Install / run

See README-PORTABLE.md for the portable-release quick start.

Runtime prerequisite: the Microsoft Visual C++ 2022 x64 Redistributable is required. It is normally already present on up-to-date Windows systems. If the app fails to start with a missing-DLL error, install it from https://aka.ms/vs/17/release/vc_redist.x64.exe.

The portable build is not code-signed. Windows SmartScreen may warn on first launch — this is expected.

Building from source

Prerequisites

• Visual Studio 2022 (Desktop development with C++ workload) — provides cl.exe, MSBuild, and the Windows SDK.
• CMake 3.25 or newer.
• Git.

Fast local development loop

Use minimal validation during development, then run the complete gate once before merge.

Implementation loop:

cmake --preset windows-x64-debug
cmake --build --preset windows-x64-debug-exosnap
cmake --build --preset windows-x64-debug-presentation-state-tests
cmake --build --preset windows-x64-debug --target <focused_test_target>
ctest --preset windows-x64-debug -R "<focused_test_regex>" --output-on-failure
scripts\check-format.ps1
git diff --check

Final gate:

scripts\check-format.ps1
git diff --check
cmake --build --preset windows-x64-debug
ctest --preset windows-x64-debug --output-on-failure
scripts\check-quality.ps1 -StaticOnly
cmake --build --preset windows-x64-release-exosnap

scripts\check-quality.ps1 keeps its default configure/build/test behavior for humans and hooks. Use -StaticOnly after a full build and CTest have already run, so final validation does not rebuild solely to run standalone static checks.

Ninja (optional, faster builds)

Install Ninja once:

winget install --id Ninja-build.Ninja

Ninja builds must be launched from a VS Developer PowerShell or Command Prompt so that cl.exe is on the PATH. Available Ninja configure presets: windows-x64-ninja-debug / windows-x64-ninja-release.

# From a VS Developer PowerShell/Command Prompt:
cmake --preset windows-x64-ninja-debug
cmake --build --preset windows-x64-ninja-debug-exosnap

sccache (optional, compiler cache)

Install sccache once:

winget install --id Mozilla.sccache

Enable at configure time: -DEXOSNAP_USE_SCCACHE=ON. Works with both Visual Studio and Ninja generators.

cmake --preset windows-x64-ninja-release -DEXOSNAP_USE_SCCACHE=ON
cmake --build build/windows-x64-ninja-release --target exosnap

Known limitation: sccache can encounter PDB contention errors with MSVC Debug builds when third-party libraries enforce /Zi. Prefer sccache with Release builds or use /Z7 to embed debug info in object files.

Coding standards

• C++ code is formatted with clang-format.
• Static analysis baseline uses clang-tidy.
• Repository-owned targets compile with strict warnings.
• Run scripts\pre-commit.ps1 before committing. It runs formatting checks, optional clang-tidy, and configure/build/test.

Repo layout

app/          application source (UI, models, recording pipeline)
libs/         internal libraries (recorder_core, capability, etc.)
docs/
  roadmap.md  version roadmap and architecture guardrails
  decisions/  architecture decision records (ADRs)
packaging/    WinGet, Chocolatey, and Scoop packaging manifests
scripts/      build, quality, and release scripts
tests/        test targets
cmake/        CMake helper modules (VendorFFmpeg, etc.)
third_party/  vendored third-party sources (libmatroska, libebml, etc.)
tools/        developer tooling

Third-party / licensing

• ExoSnap is licensed under GPL-3.0-or-later; see LICENSE.
• ExoSnap bundles FFmpeg as LGPL-2.1-or-later shared DLLs (dynamic linking). Third-party component licenses are listed in THIRD_PARTY_NOTICES.md.
• The FFmpeg binaries are produced by the companion repository Exoridus/exosnap-ffmpeg-build.

Code signing

ExoSnap participates in the SignPath Foundation free code-signing program for open-source projects, with code-signing infrastructure provided by SignPath.io. Once the certificate is issued, Windows release binaries (portable ZIP and MSI) will be signed; pre-certificate builds are unsigned, so Windows SmartScreen may warn on first launch.

Contributing / CI

GitHub Actions runs configure, build, and test on Windows for both windows-x64-debug and windows-x64-release, plus a lint pass. PRs are the review gate.