v0.3.1

What's Changed

• Release: rename the published APK from the generic Gradle name by @guyiome in #19
• Run Android Lint (lintDebug) in CI by @guyiome in #23
• Add CodeQL static analysis workflow (java-kotlin) by @guyiome in #22
• Add security policy and issue/PR templates by @guyiome in #24
• Update compose-bom and appcompat (Dependabot #21, #20) by @guyiome in #25
• Fix camera preview/mesh overlay freeze after in-app language change by @guyiome in #26
• Replace the placeholder logo with the final artwork by @clem_dreaw by @guyiome in #30
• Update Gradle wrapper, AGP and actions/setup-java (Dependabot #27, #28, #29) by @guyiome in #31
• Prepare v0.3.1: bump version to 0.3.1 (versionCode 4) by @guyiome in #32

Full Changelog: v0.3.0...v0.3.1

v0.3.0

v0.3.0 — First stable release

AndroidMoCap turns an Android phone into a standalone facial motion-capture tracker for VTubing:
front camera → 52 ARKit blendshapes + gaze/head pose (MediaPipe), streamed live over the local
network to Blender/Unity, VBridger, or directly to VTube Studio. No cloud, no third-party app
dependency.

This is the first non-beta release. v0.2.0-beta.1 already proved the build/signing/distribution
pipeline; this release adds a couple of real features on top and promotes the app out of beta.

Highlights

• Automatic best-pipeline selection based on device capabilities (GPU/CPU, RAM, cores, ARCore
  support) — nothing to configure, with automatic GPU → CPU fallback if the GPU delegate fails.
• Triple network output: VMC/OSC (Blender, Unity), iFacialMocap/UDP (VBridger), and direct
  VTube Studio integration via its own Plugin API — including an optional VBridger-style
  translation to VTube Studio's default parameters (FaceAngleX, MouthSmile...), so a streamer
  already rigged for VBridger doesn't need to run it.
• Mirror mode (on by default) so blendshapes and head orientation match what you see in the
  live preview.
• Per-blendshape weight adjustment — new since the beta — fine-tune individual blendshapes
  (e.g. an over- or under-reactive one) without touching the model.
• In-app language selector (English/French), independent of the system language.
• Semi-automatic update check — new since the beta — the app now checks GitHub Releases and
  links directly to the latest one; no silent install, this is the release that makes it start
  actually detecting updates (a prerelease-only "latest" was invisible to it before).
• Power-save mode, on-demand neutral-pose calibration, low-battery alert, exportable logs for
  bug reports.

Known limitations

• Experimental tongue-out detection (off by default, in Settings > Experimental features): a
  jaw-open gate combined with an embedding classifier compared against your own short calibration
  recording — works well in testing but hasn't yet cleared the reliability bar for a default-on
  feature; enable it at your own pace, feedback welcome.
• On the ARCore-backed tracking tier, a harmless native MediaPipe log warning may appear
  continuously in logcat; tracking itself is unaffected.
• No dedicated depth sensor on Android (unlike iPhone/TrueDepth) — an inherent hardware ceiling on
  tracking accuracy that no software fix removes.

Installation

Download app-release.apk below and install directly (Android will warn about "unknown sources" —
expected for an APK distributed outside a store). See the README for
PC-side setup per target (Blender/Unity, VTube Studio, VBridger).

Feedback

Bug reports and feedback welcome via Discord guy_iome or GitHub Issues.

v0.2.0-beta.1

What's Changed

• Bump org.jetbrains.kotlin.plugin.compose from 2.3.20 to 2.4.10 by @dependabot[bot] in #6
• CI: switch to enhanced Gradle cache now that the repo is public by @guyiome in #13

New Contributors

• @dependabot[bot] made their first contribution in #6
• @guyiome made their first contribution in #13

Full Changelog: https://github.com/guyiome/AndroidMoCap/commits/v0.2.0-beta.1