Tam combines recording, live transcription, speaker identification and generated notes on iPhone and Mac. The official description focuses on moving from a conversation to a summary, decisions and action items, with questions answered from earlier recordings. Its audience is someone who needs to revisit what was actually said rather than rely on memory alone.
This First look uses the App Store listing. We have not recorded a meeting or evaluated recognition. The privacy boundary is explicit: notes are stored on the device, but speech recognition and note writing use external services through a relay. Local storage does not mean the whole processing route is local.
Based on official store information. We have not tested this app hands-on.
What it does
The basic workflow is to start recording, watch a transcript develop and receive notes when recording stops. The developer says action items and decisions are checked against the conversation. That is a quality claim to verify against the source, not proof that every generated task belongs to the correct person.
Language detection can follow multilingual conversation, including switching during a sentence according to the description. Notes, captions and answers can be presented in the user's language. The source names several supported languages, but we have not assessed accuracy for a particular pair, accent or specialist vocabulary.
Speaker identification separates voices during a recording. Naming a voice can let Tam recognize that person in later recordings, with voiceprints described as staying on the device. Voice recognition and ordinary transcription have different data paths, so keep the on-device voiceprint claim separate from the external speech-processing route.
On Mac, Tam can record both sides of several named calling applications and browser calls, with an offer to record when a call starts. No bot joins the meeting according to the developer. On iPhone, Talk provides spoken interpretation between two people. Those are different modes, not a promise that every feature operates identically on every device.
Answers to questions about saved conversations include a reference to the relevant moment. Notes and transcripts can also be shared as text. A practical evaluation is a non-sensitive short conversation whose decisions are already clear, then comparing the transcript, generated notes and cited moment.
Before you install
The developer names Soniox and Anthropic's Claude as outside services used through Tam's relay. It says nothing is retained there or used for model training, while voiceprints remain local. These are provider and developer statements, not the result of our own audit. Review the current policy before recording confidential work.
Recording permission is separate from the absence of a meeting bot. Agree the activity with participants and follow the policy that applies to the conversation. An app being able to capture both sides does not establish permission to do so in every setting.
Names, numbers and responsibilities deserve direct review. A fluent summary can omit a qualification, and a speaker label can be wrong. Compare important action items with the actual moment rather than treating the generated list as an agreed record without checking.
The listing gives health-related examples of asking what a clinician said. Tam should not become an independent medical authority: verify consequential instructions with the appropriate professional and original context. This is an accuracy boundary, not a claim that the app diagnoses or recommends treatment.
We have not tested interruptions, long-session reliability, speaker recognition or account-free recovery. On-device notes still require an appropriate retention and backup plan, and sharing text creates a new audience for the conversation.
Price and access
The linked US download is free, with fifteen free minutes every month and no card required according to the description. Tam Pro offers unlimited recording at $6.99 monthly or $49.99 annually, with automatic renewal through the Apple Account. Confirm current regional terms and device requirements before subscribing. Unlimited recording is an access allowance, not a guarantee of unlimited accuracy or reliable recovery.
Our take
The strongest proposition is connecting a generated note to the part of the conversation that supports it. That can make review more concrete than a summary alone, provided the user checks recognition and speaker attribution. The honest external-processing disclosure is also central to deciding what belongs in the app.
Start with a short, non-sensitive example and compare every claimed decision with what was said. Test the device and mode you actually need, including the free monthly allowance, before depending on a long meeting or multilingual exchange.
We have not verified accuracy, privacy implementation or recording stability. Our take is a source-based introduction to a conversation-recall workflow, with participant permission and evidence review remaining part of the user's responsibility.
A question answered from past notes should be checked against its cited moment. The answer may be concise while the original sentence contains an exception or condition. Convenient retrieval supports verification rather than removing the need for it.
For a speaker test, choose a harmless exchange with clearly different turns. Confirm the labels before asking the app to attribute commitments. We have not measured how overlapping voices or similar voices affect recognition.
Sources
App icon from the official store listing.
