Installation and setup
Download the installer for your system from the downloads page. On first start the app guides you through three steps: choose a model profile, download the models, grant permissions. After that it is ready to use.
The 14-day trial starts with the first launch, with no account and no credit card, with the full Max scope.
Your first recording
Start a meeting in the tool you already use, or sit down in the meeting room. The app records microphone and system audio, with no bot in the meeting; a screen video is optional.
- Start the recording: in the app, or through the icon in the menu bar (macOS) or the notification area (Windows). If you have switched on meeting detection in the settings, the app offers to start as soon as a call begins.
- During the recording you see the level of both sources. If nothing arrives from the microphone, the app tells you instead of silently recording an empty track.
- After you stop, the transcript, speaker separation and topic minutes are created on your machine. How long it takes depends on the hardware, see below.
The app works out the number of speakers in the call itself. It suggests names when they come up in the conversation. If several people share the same microphone, choose Group (automatic) or the number of people under People on the microphone before recording. You can rename every speaker; the minutes take over the name.
Permissions
macOS
The app needs access to the microphone. For the audio of the other meeting participants, macOS also requires the “Screen Recording” permission, even if no video is saved. The setup assistant requests both permissions directly from the system.
If a permission was denied, grant it later under System Settings → Privacy & Security: Microphone and Screen Recording. Then restart the app.
Windows
Windows has one setting for microphone access by all apps (Settings → Privacy → Microphone). System audio needs no permission of its own; the app captures it directly from the audio device.
If microphone access is missing, the app opens the matching Windows setting.
Models and tiers
After the first download the AI models are stored on the disk; from then on the app runs without a connection. An interrupted download resumes. If an update needs a new model file, the app asks before downloading it; this needs a connection once, and without the file the app cannot be used. The app suggests the tier that suits your memory; you can switch it in the settings.
| Tier | Models | Recommended from | Download |
|---|---|---|---|
| Quality | Whisper large + Gemma 4 12B | 24 GB RAM | 8.6 GB |
| Compact | Whisper large + Gemma 4 E4B | 12 GB RAM | 6.8 GB |
| Light | Whisper small + Gemma 4 E2B | any machine | 3.9 GB |
On machines with at least 16 GB of memory the app also downloads the model for visual evidence (visual minutes); up to 1.0 GB depending on the tier.
Below 12 GB of memory the Quality tier swaps its language model for the compact one, so the machine can load it. It is not a fourth tier, only a safeguard against a model that would never start.
Activating the licence
Online
After purchase you receive a licence key. Paste it under Settings → License and click “Activate online”. The app binds this device to the licence; the server sees the licence key, the device identifier and the operating system, never meeting content.
Each user activates the app on as many devices as their plan allows. Release a device you no longer use in the settings; another one can then activate.
Without internet (Enterprise)
For machines without network access the app exports a request file (Settings → License → Offline activation). Evelan processes the file and delivers a signed response file, which you import on the device. Releasing a device also works offline, through a release file.
Offline activation is part of Enterprise and is set up with your IT.
What the licence server sees is described on the Security page.
Hardware
A Mac with Apple silicon from macOS 14, or a Windows PC with an x64 processor. The tier determines what changes:
- Quality: the large language model for the minutes. On a Mac with 24 GB of memory the minutes of a one-hour meeting take about 4.5 to 8 minutes.
- Compact: the same transcription, a more compact language model for the minutes.
- Light: a smaller transcription model and the smallest language model, for machines with little memory.
The transcription uses the graphics unit (Metal on macOS, Vulkan on Windows) when there is one. The local AI model releases its memory when idle, after 30 minutes by default; you choose the waiting time under Settings, Transcription & AI.
Export
You export any minutes as Markdown, PDF or Word document (DOCX), with letterhead, logo and signature if you have stored them. Recording and screen video can be merged into one MP4 file. PDF and Markdown are in every plan, Word and video from Pro. Export needs no model call and no connection.
To hand minutes over to a DMS, set up a webhook or a folder in the settings; without that setup nothing leaves the machine.
Term and cancellation
Cancel by email to kontakt@evelan.de, effective at the end of the term. With annual payment the licence ends after twelve months, with monthly payment at the end of the month.
After it expires, your recordings, transcripts and minutes stay on your machine and can still be read and exported. New recordings and new analyses need a valid licence.