Omi AI - Smart Meeting Notes

Opening the app and the first recording attempt

Downloading Omi AI from the Play Store takes less than a minute. The app size is reasonable, and the installation process is standard. Opening it for the first time presents a clean interface with a prominent record button. Tapping it starts capturing audio immediately. The app asks for microphone permission, which is expected. What is not expected is what happens next. The recording indicator shows active status, but the transcription does not appear in real time. A spinner appears, and the app informs you that processing will happen after the recording stops. This is the first hint that the offline recording claim has a catch. The app can record without an internet connection, but it cannot transcribe without one. The audio file sits locally until a connection is restored, at which point it uploads for processing. This creates a dependency that is not clearly communicated in the marketing material.

The interface itself is functional but cluttered. The main screen shows a list of past conversations, each with a title, date, and duration. Tapping one opens a detailed view with a transcript, summary, and action items. The transcript is searchable, which is useful for finding specific mentions. The summary is generated automatically and is usually accurate. The action items are extracted from the conversation and presented as a checklist. These features work well when the recording is successful. The problem is that success is not guaranteed. Multiple users have reported that the app stops recording without warning, or that the transcript is missing entire sections of the conversation. One reviewer noted that they "miss conversations I thought were recorded but are not." This is a critical flaw for an app that promises to capture everything.

How offline recording and cloud transcription create a failure point

The offline recording feature is one of Omi AI's most touted capabilities. The app can record audio without any network connection, which is useful for meetings in basements, subway rides, or any location with spotty coverage. The audio is stored locally on the device. When a connection is restored, the app uploads the file to the cloud for transcription. This sounds straightforward, but the reality is more complicated. The transcription process is not automatic. The user must manually open the app and ensure the recording has been uploaded. If the app is closed or the device is restarted, the upload may not happen. This is exactly what happened to one user who recorded a meeting, walked to an elevator, and lost all insights because the app failed to sync. The core promise of "set it and forget it" is broken by this manual sync requirement.

Compare this to Otter AI, which offers real-time transcription and automatic syncing. Otter processes audio on the cloud as it is being recorded, so there is no gap between recording and transcription. Omi AI's offline-first approach is more privacy-focused, but it sacrifices convenience. The trade-off is significant for users who need immediate access to meeting notes. The delay between recording and transcription can be hours or even days if the user forgets to sync. This is not a theoretical concern. Multiple GitHub issues and user reviews describe lost recordings and failed uploads. The app's reliability is undermined by this design choice. The user is left wondering whether their recording will actually be transcribed or if it will disappear into the void.

The brain interface feature adds another layer of complexity. The Omi wearable device, which costs $89, includes an electrode that can detect brain signals. The founder, Nik Shevchenko, claims the device can "read your mind" by recognizing mental focus. In practice, this feature is experimental. It only understands one channel of brain activity and requires the device to be taped to the user's temple. The Verge described the setup as "medical tape" and "sci-fi optimism." The brain interface does not work reliably and is not a core part of the app's functionality. It is a novelty that distracts from the app's primary purpose: recording and transcribing conversations. Users who buy the hardware expecting a mind-reading assistant will be disappointed.

The real-world experience: disconnections, bugs, and unreliable performance

The most common complaint about Omi AI is the constant disconnection. The wearable device frequently loses its Bluetooth connection to the phone. When this happens, the app stops recording. The user may not notice until they check the app and find that the recording was interrupted. One reviewer described seeing "Your Omi Disconnected" notifications "A LOT" and noted that the device disconnects when you "walk away from the device or do anything to even slightly inconvenience it." This is not a minor issue. It is a fundamental flaw that makes the device unreliable for capturing long conversations or meetings. The app's core value proposition is to capture everything, but it fails to do so consistently.

The app itself has its own set of bugs. The UI is described as "random and unorganized" by one user, who also noted that there is "ABSOLUTELY TOO MUCH AI!!!" The app's features are extensive, but they are not well-organized. The Brain Map, which visualizes how memories connect, is confusing and does not provide clear value. The Daily Recap is a nice idea but often includes irrelevant information. The action items are extracted from conversations, but they are not always accurate. One user complained that the app adds tasks "randomly" and that the flags or tags are "worthless." The app's complexity works against it. Users want a simple, reliable recording tool, not a second brain that requires constant management.

Customer support is another pain point. Multiple users have reported that refund requests are ignored and that support emails go unanswered. One review on PissedConsumer gave the app a 1.0 star rating and described a negative experience focused on refunds and poor customer service. The developer's response to negative reviews on the Play Store is polite but generic. The same template is used repeatedly, promising to pass feedback to the development team. There is no evidence that this feedback results in meaningful changes. The app's development tempo has slowed, with recent updates only including bug fixes and performance improvements. The changelog is not published on the Play Store, so users have no way of knowing what has actually been fixed.

The transcription quality is a mixed bag. When the app works, the transcripts are accurate. The app uses OpenAI Whisper and Deepgram for transcription, which are industry-leading services. The summaries are also generally good, capturing the key points of a conversation. However, the accuracy drops significantly for non-English languages. One GitHub issue reported that despite setting Portuguese as the main language, the transcription frequently displayed "random or 'weird' characters/words from other languages that were not spoken." This corruption occurred consistently in nearly every conversation recorded using the L. Pendant. Users who need multilingual support will find the app unreliable.

The cost of the ecosystem and what the app actually requires

The app is free to download, but the full experience requires a hardware purchase. The Omi pendant costs $89. The monthly subscription for unlimited transcription is $16 per month. This adds up to a significant investment for a device that does not work reliably. The hardware is open source, which is a positive for developers, but it is not polished for everyday users. The setup process is technical and requires configuration. The app's marketplace offers thousands of tools, but most are niche and require technical knowledge to use. The average user will not benefit from these features. They will be frustrated by the complexity and the constant disconnections.

The app's privacy features are a strong selling point. The data is encrypted in transit with TLS and at rest with AES-256. The app does not share data with third parties and does not collect any data, according to the Play Store's data safety section. Users can run the app locally without any cloud dependency, which is a significant privacy advantage. The app is also SOC 2 and HIPAA compliant, making it suitable for healthcare and enterprise use. These features are rare in the AI transcription space and set Omi AI apart from competitors like Otter AI and Fireflies.ai. However, the privacy benefits are undermined by the reliability issues. A secure app that does not work is not useful.

The app requires Android 10.0 or later. The APK size is approximately 95 MB, with additional storage required for offline recordings. The app requests permission to access the microphone and location. The location permission is used to improve transcription accuracy and is not required for basic functionality. The app also requires Bluetooth permission for connecting to the wearable device. The permissions are standard for a recording app and do not raise significant privacy concerns.