MeshCore Open TAK 9.5.0-tak (build 1161)

MeshCore Open TAK 9.5.0-tak (build 1161) · 2026-10-21 19:33 UTC · commit 856e5a2

Follows MeshCore Open 9.5.0 with the TAK fork's features added.

Platform	Download
Android (arm64)	meshcore-open-tak.apk, signed with the permanent release key (installs over the previous build)
Windows (x64)	meshcore-open-tak-windows-setup.exe (installer), or the .zip to run without installing
macOS 14+	meshcore-open-tak-macos.dmg: drag to Applications
Linux (x86_64)	meshcore-open-tak-linux-x86_64.AppImage (chmod +x and run), or the .tar.gz

The desktop builds are unsigned: on Windows choose More info → Run anyway; on macOS right-click the app → Open the first time.
Once installed, the app offers later builds itself (Settings → App updates).

Recent changes:

• Bundle the ESP32 partition catalog and GitHub root certificates
• LoRa OTA: test fake takes an explicit route; keep live dropdown values
• Firmware updates from Mike Carper's MeshCore Open fork
• Watch: auto-zoom to the newest watched report
• MediaMesh: Find MediaMesh, which preset is in use here
• Watch: Bloodhound on a report opens the map
• Walkie-talkie: drop a needless null assertion
• Walkie-talkie: push-to-talk on a channel over MediaMesh

AI sketch HD model pack (ft2, 512 px)

AEIC-SE ft2 model pack for ncnn at 512 px, used by MeshCore Open TAK's AI sketch HD photo option. Exported by .github/workflows/aeic-ncnn-models.yml from mikecarper/AEIC@b00795c. Not an app release.

Mimi voice model v1 (ONNX)

Mimi neural audio codec by Kyutai (https://huggingface.co/kyutai/mimi),
licensed CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/).
Converted to ONNX for MeshCore Open VOX: fixed 20 s input, 8 codebooks
(~1.1 kb/s), weights stored as fp16 and cast to fp32 at load time.