Record without an account
Use the phone microphone or a supported stock-protocol Omi BLE wearable. Start and stop capture explicitly; a persistent Android notification keeps the recorder visible.
Independent Android companion / 0.6.26
Turn your Omi wearable into a recorder that transcribes English on your phone.
Local speech models. Encrypted recordings. No Omi account or cloud transcription service. Start with your phone microphone; add a compatible wearable when ready.
0.6.26-pilot · 271 MB · ARM64 · Android 8+
CPU-specific pilot. Read the device requirements before installing.
Phone mic or Omi wearable
Audio encrypted on-device.
Bundled English speech models.
Sync only after private-PC pairing.
01 No account to start
02 No model download
03 No cloud ASR fallback
What you can do
The phone handles speech recognition. A PC is optional for recording and transcription, and required for the separate sync and assistant workflow.
Use the phone microphone or a supported stock-protocol Omi BLE wearable. Start and stop capture explicitly; a persistent Android notification keeps the recorder visible.
Whisper small.en produces final English transcripts on the phone. Vosk provides provisional live text. The models ship inside the APK; there is no cloud transcription fallback.
Transcript bodies and normal-capture audio are encrypted on the phone. Record timestamps, type and sync-state metadata are not encrypted. Unsynced records wait locally.
Optional certificate-pinned HTTPS sync goes only to an explicitly paired private-IP receiver. The receiver, pairing credentials and Hermes backend are not included in this APK release.
Map supported button presses to task controls or bookmarks, read device battery status, and adjust LED brightness where supported. No firmware flashing required; hardware features vary.
Start here
No wearable or PC needed for the first recording.
Android 8+ and ARM64. This build’s native Whisper library targets ARMv8.2-A with FP16 and dot-product support. ARM64 alone is not enough to establish compatibility.
Download on the phone, open the file, and allow installation from that browser or file manager if Android asks. First-use model unpacking needs additional free storage.
Open Omi Private while unlocked. Tap Start phone microphone, allow microphone and notifications, speak, then pause. Use Show recent local transcripts to inspect the result.
Tap Stop capture. For Omi, disconnect it from other recording apps, enable Bluetooth, tap Find Omi wearable, grant the requested Bluetooth permission and select it.
This release
0.6.26 changes the reply/capture coordination. This feature needs a separately configured private receiver and assistant backend.
In Tasks → Enable spoken replies, opt in after each app restart. Replies use an installed offline Android voice. Old or interrupted replies do not automatically replay.
Tested ≠ universally supported
Existing release evidence records an in-place update and a successful launch on the development phone. All 22 selected host runners eventually passed across the initial run and focused reruns—not one clean full-suite run.
Still pending: real concurrent speech/capture acceptance, locked-screen routing, physical range-loss reconnection and all-day reliability. Other phone models have not been certified.
Good to know
An independent pilot app, not the official Omi client. The release shares a compiled Android build, not the private application source.
English recognition runs on your phone. Local transcript bodies and normal-capture audio are encrypted; record metadata is not. After explicit pairing, queued data can go to your private PC. This does not promise an independently audited security boundary, or change what other installed apps do.
No. The APK is ARM64-only, Android 8+, and its native inference library is built for ARMv8.2-A FP16 and dot-product CPU features. Budget or older ARM64 devices can be incompatible. No iOS build is included. Check device compatibility before relying on it.
No wearable is required for phone-microphone recording. No Omi account or cloud speech subscription is used by this app. Wearable support initially targets stock Omi Opus codecs 20/21; unknown firmware and hardware are not universally supported.
Only if you already have a compatible private receiver and backend configured. Neither is bundled here. Spoken task markers create proposals requiring confirmation in the configured workflow. Explicit task FINISH controls can submit immediately; speaker matching is experimental and is not authentication.
Not automatically. Explicit stored-audio recovery supports only inspected legacy firmware with working storage and pre-existing recordings. Unknown variants and consumer ring recovery are blocked. Audio never recorded cannot be recreated.
Nothing. Omi Private has its own package. It does not uninstall, reconfigure or stop another app. Disconnect the wearable from competing recording apps before connecting it here. Any other app’s existing cloud behaviour remains unchanged.
Share this page or the GitHub release so recipients see compatibility notes and checksums. No recordings, pairing files, credentials, signing keys or private PC configuration are included. Third-party licence notices accompany the binary.