Technical reference
System requirements & known limits
What you need to run AdjiCera, what it does without an internet connection, the formats it reads and writes, and the honest edges of what we have tested so far. Ready to install? Head to Download.
Student and Professional editions
macOS and Windows — no install
No connection needed to record or transcribe
1 · Supported platforms
Windows. The AdjiCera Student and Professional desktop apps run on Windows 10 and Windows 11. Both installers run entirely on your PC and need no internet after install.
macOS. There is no separate Mac download. You join or host meeting calls with live transcription directly in your browser, on any Mac — Apple Silicon (M1, M2, M3) or Intel. Nothing to install.
iPhone, iPad & Android. Mobile is on the roadmap as a review-and-study companion — review transcripts and flashcards on the go, synced through your account — while recording stays on your PC or Mac for now. You can join the mobile waiting list from the Download page.
2 · Supported browsers
The browser experience — the web meeting app and AdjiCera Call — is built for current versions of Google Chrome and Apple Safari. A recent, updated browser with camera and microphone access is required for live calls and in-browser transcription.
Exact minimum browser versions, and formal support status for Microsoft Edge and Firefox, are being finalized before launch — email support@adjicera.com if you need confirmation for a specific browser.
3 · Minimum hardware
- Disk space. Room for the installer plus your recordings. The Student installer is roughly 685 MB and the Professional installer roughly 1 GB; recordings and transcripts grow with use, so leave comfortable headroom on the drive.
- Microphone. Any working microphone — built-in or external — is required to record audio.
- Camera. Required only for video on AdjiCera Call. It is not needed to record a lecture, record a meeting, or transcribe audio.
- CPU & memory. AdjiCera transcribes and runs its study tools on your device's CPU — no dedicated GPU or graphics card is required, and none of the local features depend on one. See CPU-based transcription for how that works. A reasonably modern processor and adequate RAM still improve processing speed.
We are still finalizing published minimum CPU and RAM figures from testing across representative low-, mid-, and high-tier Windows laptops (see Section 9). Until those numbers are published, contact support@adjicera.com if you want to check your machine.
4 · Internet & connection guidance
On Windows, the core loop is offline: recording, transcription, speaker labels, summaries, flashcards, quizzes, concept maps, action items, and analytics all run on your device without a connection. Airplane mode is fully supported.
You need a connection only to:
- Activate your trial and keep your subscription state in sync across devices.
- Place or join an AdjiCera Call (real-time audio/video).
- Use the optional Prime AI features, which send the text you choose to a third-party AI provider (see Privacy).
Specific bandwidth recommendations for calls (minimum and recommended upload/download) are being measured across network types before launch and are not published yet.
5 · Calls, participants & the TURN relay
AdjiCera Call connects participants directly, peer-to-peer, whenever the network permits. When a firewall or NAT blocks a direct path, the media — already encrypted end to end — is carried through our connectivity (relay) infrastructure over standard outbound connections. No inbound ports or port-forwarding are required, so calls work behind typical home and office networks.
The maximum meeting size we have formally tested is not yet published. We have verified long single-session recordings (multi-hour lectures and seminars); the tested ceiling for simultaneous call participants is being documented and will be stated here rather than estimated. For a specific group size, contact support@adjicera.com.
6 · Offline capabilities & restrictions
Works offline: recording, on-device transcription, automatic speaker labels, summaries, action items, follow-up email drafts, topic chapters, flashcards, quizzes, concept maps, and analytics.
Requires a connection: trial activation and subscription sync, AdjiCera Call, and optional Prime AI features. If you would rather nothing leave your device, simply do not enable Prime — every core feature runs locally without it.
7 · Recording & export formats
Recordings are captured and transcripts are generated locally, then stored on your device for you to review, export, and share. The transcription models are bundled inside the installer, so there is no separate model download at runtime and no per-minute cloud meter.
The exact list of supported recording and export file formats is being finalized for this page. For confirmation of a specific format you need to import or export today, email support@adjicera.com.
8 · Local model size & processing speed
Because the models ship inside the installer, the download sizes above (roughly 685 MB for Student and roughly 1 GB for Professional) already include what is needed to transcribe on-device. There is nothing extra to fetch before your first recording.
Expected processing speed — how long a finished transcript takes relative to recording length, by hardware tier — is being measured on the reference machines described in Section 9 and will be published here. We are not quoting a speed figure until it is measured.
9 · How we measure our claims
We would rather publish a measured number late than an invented one early. Here is how the claims on this site are being tested so that, when figures are published, they map to a named, repeatable setup:
- Transcription accuracy. Measured as word error rate against human-verified reference transcripts, on both openly available spoken-audio sets and our own real-world recordings, reported per audio condition (clean, noisy, multi-speaker) rather than as a single headline number.
- Latency. Measured as wall-clock delay between spoken audio and the caption appearing, and between pressing Stop and the finished write-up, on each reference hardware tier.
- Processing speed. Measured as transcript turnaround relative to recording length across representative low-, mid-, and high-tier Windows laptops.
- Connection success. Measured from the anonymous connection-diagnostics outcomes described in our Privacy policy — whether a call connected, how long it took, and by which method — with no account identifier, message content, or media attached.
Until each result set is published here, treat the corresponding numbers as pending rather than assuming a figure. When they land, this section will name the test setup alongside every number.
10 · Troubleshooting
Camera or microphone not detected.
Windows SmartScreen on install.
Firewall or NAT blocking a call.
Permissions on repeated failures.
Still stuck? Email support@adjicera.com.
Related
- Private, on-device processing — how nothing leaves your machine.
- Speaker identification — who said what, generated locally.
- Meeting app without a bot and local meeting recorder.
- AdjiCera vs Otter.ai · Privacy · Pricing · For work
Frequently asked questions
What operating systems does AdjiCera support?
Do I need an internet connection?
Does AdjiCera add a bot to my meeting?
Where are my recordings stored?
Do calls need a special network or open ports?
You already recorded the lecture.
Now actually remember it.
For Windows · ~685 MB · One-month full trial · No card up front · Mac in browser
Your recordings stay on your device—no cloud, no third parties, no data tracking. Complete privacy by design.