# Advanced
URL: /docs/advanced-settings
Update channel, retention, paths, machine ID, and the reset button.
***
title: Advanced
description: Update channel, retention, paths, machine ID, and the reset button.
--------------------------------------------------------------------------------
**Settings → Advanced** holds the things you'll touch rarely — once at setup, once when something's wrong.
## Performance
* **Preload Whisper Model** — keep the local speech model loaded in memory so the first push-to-talk is fast. On by default. Turn it off to save RAM if you mostly use the cloud model.
## Behaviour
* **Preserve clipboard contents** — Amical pastes by writing to your clipboard, then restoring whatever was there before. On by default. Turn it off if you'd rather Amical leave its transcript on the clipboard.
* **Debug Mode** — extra logging for bug reports. Off by default.
## Updates
* **Update Channel** — **Stable** (default) or **Beta**. Beta gets new features earlier and bugs to match. The auto-updater follows whichever you're on.
## History
* **History Retention** — how long Amical keeps audio files attached to your transcriptions. **1 day, 7 days, 14 days, 28 days, Never**. Default is **Never**. The transcript text always stays; only audio is pruned when you set a cutoff.
## Privacy
* **Anonymous Telemetry** — see [Telemetry](/docs/telemetry) for what's collected and the side-effects of turning it off.
## Diagnostics
* **Data Location** — where Amical stores your database, models, and audio. Read-only.
* **Log File Location** — where logs live. The **Download** button next to it saves the current log file out for sharing in a bug report.
* **Machine ID** — your install's anonymous ID for telemetry. **Copy** puts it on your clipboard; include it in bug reports.
## Danger zone
* **Reset App** — wipes transcriptions, notes, vocabulary, settings, and downloaded models. Sign-out doesn't happen automatically; sign out separately if you also want to clear your account session.
The confirm dialog lists everything that will be deleted. The action runs a clean restart of the app afterwards.
# AI formatting
URL: /docs/ai-formatting
Turn a raw transcript into punctuation, structure, and an app-appropriate tone.
***
title: AI formatting
description: Turn a raw transcript into punctuation, structure, and an app-appropriate tone.
--------------------------------------------------------------------------------------------
AI formatting runs your raw transcript through a language model that adds punctuation and structure, and shapes the output to fit the app you're dictating into. An email gets a greeting on its own line; a Slack message stays a one-liner; meeting notes come back as bullets.
## Turn it on
Open **Settings → Dictation → Formatting** (marked **Alpha**) and flip the switch.
The toggle is disabled until at least one of these is true:
* You have a local language model synced under **Settings → AI Models → Language** (Ollama or any OpenAI-compatible endpoint), or
* You're signed in *and* using the **Amical Cloud** speech model. The cloud formatter rides on the same connection as cloud speech, so both have to be set up.
The **Formatting model** picker below the toggle lets you choose which model does the rewrite — Amical Cloud, or any local language model you've synced.
## What changes per app
Amical detects the focused app and adjusts the formatting rules. The same dictation comes out differently depending on where it lands.
### Email
Input: `hi john um i wanted to follow up on our meeting the proposal looks good but we need to revise the timeline thanks sarah`
Output:
```
Hi John,
I wanted to follow up on our meeting. The proposal looks good, but we need to revise the timeline.
Thanks,
Sarah
```
Greeting, body, and closing land on their own lines. Tone stays professional. Short replies stay short — no greeting is invented if you didn't say one.
### Chat (Slack, Discord, iMessage)
Input: `hey um quick question do you know if the deploy went through i saw some errors in the logs`
Output:
```
Hey, quick question - do you know if the deploy went through? I saw some errors in the logs.
```
Conversational. Dashes and commas for natural pauses instead of paragraph breaks. Emoji preserved. Short replies don't get padded.
### Notes
Input: `meeting notes attendees john sarah mike discussed the roadmap action items sarah to finalize design by friday mike to review budget`
Output:
```
Meeting Notes
Attendees: John, Sarah, Mike
Discussed the roadmap.
Action items:
- Sarah to finalize design by Friday
- Mike to review budget
```
Headings and bullets when the content implies a list. Action items get pulled out. Scannable, not prose-heavy.
### Amical Notes
Input into a note in Amical itself goes a step further — it's formatted as Markdown that adapts to length: a one-liner stays a paragraph, medium-length content gets bullets, longer content gets headers and sub-sections. Bold for emphasis, code blocks for technical content.
### Everything else
Apps that don't match a known category (search bars, terminal, address fields) get a plain cleanup — punctuation and casing, no structural rewriting.
## How Amical reads the context
Three signals shape what comes out:
* **The focused app** — Slack, Mail, your editor, a browser tab. This is the main signal; it picks the formatting ruleset above.
* **The focused field's role** — on macOS, accessibility permission lets Amical tell a search bar from a long-form composer, so the same dictation gets shaped differently inside the same app.
* **The active language** — auto-detected per recording unless you've pinned one in **Settings → Dictation**.
The signal is the focused app, not the words you say. Dictating "send John the report" into a search bar gives you the literal words — Amical doesn't issue actions on your behalf. Granting accessibility permission on macOS improves accuracy by letting Amical read the focused field's role.
## Notes
* Formatting needs a language model — without one, the toggle is locked and you'll see "Formatting won't run - no language model available."
* Cloud formatting needs cloud speech *and* a signed-in account; switching either off disables it.
* The feature is **Alpha** — output can vary run-to-run, and edge cases (mixed languages, very long dictations) are still being tuned.
* For per-app *customization* of how formatting behaves — your own presets, tone overrides, custom skills per app — see [Personalization](/docs/personalisation).
# Android quick start
URL: /docs/android-quick-start
How to install Amical for Android — public beta on Google Play.
***
title: Android quick start
description: How to install Amical for Android — public beta on Google Play.
----------------------------------------------------------------------------
Amical for Android is in public beta on Google Play.
1. Install [Amical from Google Play](https://play.google.com/store/apps/details?id=ai.amical.app) (also linked from the [download page](/download)).
2. Open the app and sign in with your Amical account.
3. When prompted, grant **Accessibility Service**, **Display over other apps**, and microphone access. These let Amical's floating bubble appear over the focused app and insert your dictation.
Report bugs on [Discord](https://amical.ai/community) or [GitHub](https://github.com/amicalhq/amical/issues).
# Billing
URL: /docs/billing
Plans, usage, seats, invoices, and how cloud quota actually works.
***
title: Billing
description: Plans, usage, seats, invoices, and how cloud quota actually works.
-------------------------------------------------------------------------------
Billing and plan management live in your Amical account in the browser. The desktop app links into the right page; the subscription itself is edited in the web portal, not inside the desktop app.
## Where to find it
* **From the desktop app** — sign in, then in the sidebar under your account, click **Usage & Billing**. Amical opens your account in the browser at the plan page.
* **In a browser** — go to **app.amical.ai** and sign in; the Billing tab is in the sidebar under your organization.
Plan details are visible to organization owners and admins. Members see usage but can't switch plans or manage payment.
## What's on the plan page
The page is laid out in three bands:
1. **Current plan** — the plan name, subscription state (active / trialing / past due / canceled), the biller (Dodo for self-serve, custom for Enterprise), and a **Manage billing** button when applicable. A **Contact us** link goes to support.
2. **Usage metrics** — three cards: Words used, Team seats, and Refresh cadence.
3. **Available plans** — Free, Premium, and Enterprise cards side-by-side with a **Monthly / Yearly** toggle that changes pricing and CTAs.
Underneath, there's a per-plan **Inclusions** breakdown and a **Usage** detail panel with the current billing-period dates.
## Plans
| Plan | Best for | What you get | Seats |
| -------------- | --------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------- |
| **Free** | Personal, local-only or light cloud use | Unlimited local dictation, fast cloud models with a weekly cloud word cap (2,000 words/week), custom vocabulary, AI formatting, community support | 1 |
| **Premium** | Power users and small teams | Everything in Free + unlimited cloud words, team management, email support, monthly or yearly billing | Small team limit |
| **Enterprise** | Larger orgs with security needs | Everything in Premium + dedicated hosting, custom models, ultra-low latency, SSO/SAML, centralized billing, bulk-pricing discounts, dedicated support | Custom |
Free is always available; you don't lose anything if you let Premium lapse, you just go back to Free's cloud word cap.
## How cloud words work
The **Words used** metric counts words transcribed by the **Amical Cloud** speech model. Local dictation (a Whisper model you've downloaded) is **unlimited and free** on every plan and never counts.
Quota has three properties worth knowing:
* **Limit** — the plan's word allowance per refresh period. Free is capped (2,000 words/week); Premium is unlimited; Enterprise is unlimited.
* **Refresh** — `weekly` on Free, `monthly` on most paid plans, `daily` for some Enterprise contracts, `never` on a few legacy plans. Whatever your plan says, the **Current period starts/ends** dates on the Usage panel are authoritative.
* **Scope** — `per-user` (each member has their own quota) or `organization` (the whole org shares one pool, "pooling"). Premium ships per-user; Enterprise can be either. The Usage panel labels this as **"Per-user quota"** or **"Shared across organization"**.
When you hit the cloud word limit on Free, cloud transcription stops until the next weekly refresh. Local dictation keeps working — Amical doesn't lock the app, just the cloud speech path.
## Seats
**Team seats** counts active members against your plan's seat limit. Pending invitations also count, so the seat is reserved until the invite is accepted, expired, or cancelled.
When you try to invite past the limit, the invite fails with a "no seats available" error. Free up a seat (cancel a pending invite, remove a member) or upgrade.
See [Team](/docs/team) for how membership and roles work.
## Switching plans
Each plan card has a button:
* **Free → Premium** opens a checkout flow with your chosen billing interval (monthly or yearly). Yearly billing is discounted vs. monthly.
* **Premium → Premium yearly** (or vice versa) also goes through checkout; the proration is handled by the biller.
* **Premium → Enterprise** opens a contact path — Enterprise pricing is bespoke.
* **Downgrade or cancel** is handled inside the billing portal — see below.
Upgrades take effect immediately. Downgrades and cancellations take effect at the end of the current billing period; until then you keep the higher tier.
## Manage billing (invoices, payment method, cancellation)
The **Manage billing** button on the plan page opens the Dodo billing portal in a new tab. That's where you:
* Update your payment method
* View and download invoices
* Change your billing email
* Cancel your subscription
A failed payment puts the subscription into a `past_due` state. The biller automatically retries; if it can't recover, the subscription is canceled and you drop to Free at the end of the grace period. Update your card in the billing portal to recover.
The Manage billing button only appears when your subscription is on Dodo (the default biller for self-serve plans). Enterprise customers on a custom contract should email support to make billing changes.
## Enterprise
Enterprise covers everything in Premium plus:
* **Dedicated hosting** and **custom models** for low-latency, high-throughput use
* **Ultra-low latency** routing for real-time dictation
* **SSO / SAML** for centralised identity
* **Centralized billing** — one bill across all members, no per-seat purchase
* **Bulk-pricing discounts** for larger seat counts
* **Dedicated support** with response-time commitments
Pricing is custom. Click through the Enterprise card on the plan page or email **[support@amical.ai](mailto:support@amical.ai)** to start the conversation. The plan-page header also has a **Contact us** link that goes to the same address.
## Notes
* Plan data is fetched live from the Amical auth backend; if the page says **"Plan unavailable"**, you're either not signed in or you're a member (not an owner/admin) on a plan that hides details. Check with your owner.
* Yearly billing renews once a year on your subscription anniversary; monthly renews each month on the same date.
* The **Refresh** label on the Usage card tells you when your cloud word quota resets — it's separate from the billing renewal date.
* Cancelling doesn't delete your data. Notes, vocabulary, and local transcriptions stay on your machine; sign-in and cloud features just stop.
# Cloud AI models
URL: /docs/cloud-ai-models
Use Amical's hosted models — sign in, no local setup.
***
title: Cloud AI models
description: Use Amical's hosted models — sign in, no local setup.
------------------------------------------------------------------
If you'd rather not download anything or your machine isn't a great fit for local Whisper, use Amical's cloud option. You sign in once, then pick **Amical Cloud** from the same **Default Speech Model** dropdown your local models live in.
## Sign in
Open **Settings → AI Models → Speech**. The **Amical Cloud** row in the model table shows a **Sign In** button instead of a download button. Click it, complete the sign-in flow in your browser, and the cloud model becomes selectable.
## What runs in the cloud
* **Speech** — your audio gets streamed to Amical's hosted speech endpoint and transcripts come back.
* **Language (formatting)** — if you turn on **Formatting Alpha** in Settings → Dictation, the formatting pass also goes through the cloud unless you've pointed Settings → AI Models → Language at a local provider.
## When you need cloud specifically
A few features only run against a cloud speech model:
* **Formatting Alpha** — rewrites your transcript through a language model. Settings → Dictation surfaces this requirement when you turn the toggle on.
* **Reporting a transcription** — the flag icon in **Settings → History** sends a transcription to the team for quality work. Only available on cloud transcripts, and only if Anonymous Telemetry is on.
## Notes
* Sign-in is account-based, so your cloud entitlement, custom vocabulary, and preferences follow you across devices.
* The local and cloud options aren't exclusive — keep a local Whisper downloaded as a fallback for offline.
* Pricing, usage limits, and data handling are documented separately on the Amical website.
# Shortcuts
URL: /docs/custom-hotkeys
The five global shortcuts Amical registers and how to rebind them.
***
title: Shortcuts
description: The five global shortcuts Amical registers and how to rebind them.
-------------------------------------------------------------------------------
Amical registers five global shortcuts. They all live under **Settings → Shortcuts**.
| Shortcut | Default | What it does |
| ------------------------- | ---------- | ---------------------------------------------------------------------------------------------------------- |
| **Push to talk** | Fn | Hold to record, release to transcribe and paste. |
| **Hands-free mode** | Fn+Space | Press once to start a recording, again to stop. |
| **Paste last transcript** | Cmd+Ctrl+V | Re-pastes the most recent transcription into the focused app. |
| **New note** | Cmd+Ctrl+N | Opens a fresh note in Amical. |
| **Draft** (Alpha) | Fn+Ctrl | Hold to speak an instruction; review the generated text, then insert it. [Amical Cloud only](/docs/draft). |
## When to use which dictation shortcut
* **Push to talk** is best for short bursts — a quick reply, a search query, a one-liner. Releasing the key always stops the recording, so it can't run away on you.
* **Hands-free mode** is best for longer passages — drafting an email, talking through a thought, code comments. Clicking the [floating widget](/docs/desktop-widget) triggers the same mode without touching the keyboard.
Both modes share the same silence-detection, length cap, and slip-press threshold — see [AI dictation](/docs/dictation) for the exact numbers.
## Rebind one
1. Open **Settings → Shortcuts**.
2. Click the chip for the shortcut you want to change.
3. Press the keys. Amical captures the combination live.
If the combination conflicts with something else (a system shortcut, an already-bound Amical shortcut), Amical reverts the change and shows the conflict in a toast.
## Allow injected keystrokes (Windows)
Amical only responds to **physical** key presses. Some Windows setups send keystrokes another way — they *inject* them: remote desktop sessions, KVM switches, key remappers like PowerToys, and mouse or macro software that fires a key combo from a button. By default Amical ignores injected keys, so a background tool can't set off your shortcuts by accident.
Turn on **Allow injected keystrokes** (under **Settings → Shortcuts**) when you *want* those keys to count — for example, you've bound an Amical shortcut to a mouse-based macro or a macro key, or your keyboard reaches Amical through remote desktop or a KVM and your shortcuts otherwise never fire. Injected keys then count the same as physical ones.
Once injected keys are allowed, a tool sending keys in the background can interfere with your shortcuts:
* **Setting a shortcut** — the recorder can pick up an injected key you didn't press.
* **Triggering a shortcut** — a shortcut can fire on its own, from a tool rather than from you.
* **Dictation** — a recording can stop unexpectedly when an injected key trips your Hands-free or Push-to-talk shortcut.
If your shortcuts start behaving unexpectedly, turn **Allow injected keystrokes** back off, so Amical again reacts only to keys you physically press.
## Notes
* Shortcuts are global — they fire while any app has focus.
* Amical syncs them with a native helper to suppress the default behaviour of those keys, so a shortcut bound to **Fn** won't trigger the macOS dictation popup.
* If a key gets stuck (held when the app didn't see the release), Amical clears stale state every 10 seconds.
# Custom prompts
URL: /docs/custom-prompts
Bring your own formatting prompt — planned, not yet shipped.
***
title: Custom prompts
description: Bring your own formatting prompt — planned, not yet shipped.
-------------------------------------------------------------------------
Custom prompts aren't configurable in Amical yet.
The plan is to let you write the prompt that shapes how your transcript gets rewritten — so you can ask for "no contractions", "Slack-friendly", "always start emails with a greeting on its own line", or anything else specific to how you write.
What ships today is the **Formatting** toggle in Settings → Dictation (Alpha). It's an on/off switch with a fixed prompt; it requires a cloud speech model and a signed-in account, and the model picker lives under Settings → AI Models → Language. There's no place to edit the prompt itself yet.
## Track progress
* [GitHub issues](https://github.com/amicalhq/amical/issues) — share the prompts you'd want
* [Discord](https://amical.ai/community)
# Vocabulary
URL: /docs/custom-vocabulary
Teach Amical the words and replacements that matter to you.
***
title: Vocabulary
description: Teach Amical the words and replacements that matter to you.
------------------------------------------------------------------------
If Amical keeps mishearing a name, an acronym, or a piece of jargon, add it here and the speech model will know to listen for it. You can also map common misspellings to the spelling you actually want.
Open **Settings → Vocabulary** and click **Add Word**.
## Two modes
The dialog has a **Replacement** toggle:
* **Off (single word)** — type the word as you want Amical to spell it. Good for proper nouns ("Kubernetes", "Anthropic"), acronyms, and jargon.
* **On (replacement)** — type the misheard form on the left and the correct form on the right (`acme corp` → `Acme Corp`). Anything Amical transcribes matching the left side gets swapped for the right.
## Edit or delete
Hover any row in the list to show edit and delete buttons. Edits use the same dialog. Delete asks for confirmation.
## Notes
* Entries take effect on your next recording — no restart.
* Vocabulary is per-account, so it follows you across machines once you sign in.
* The list is paged at 100 entries; the search and sort happen client-side within that window.
# Desktop widget
URL: /docs/desktop-widget
The floating pill that starts a hands-free recording with one click.
***
title: Desktop widget
description: The floating pill that starts a hands-free recording with one click.
---------------------------------------------------------------------------------
The widget is a small floating element that sits on top of every other window. Click it once to start a hands-free recording, click the stop icon to end it.
## States
* **Idle** — a thin pill, 8 px tall. Click-through is disabled until you hover, so you won't trigger it by mistake.
* **Hovered** — expands so you can hit the click target.
* **Recording (hands-free)** — wider, with a 6-bar waveform that pulses with your voice and a red stop icon. Click the stop icon to end the recording.
* **With notes shortcut** — if the notes feature flag is on, hovering reveals a small notebook icon next to the main button. Click it to open the notes window.
## Show or hide it
**Settings → Preferences → Show widget while inactive** controls whether the widget is on screen when you're not recording. On by default. Turn it off if you only want the widget visible during a recording — push-to-talk and the global shortcuts still work either way.
## Notes
* The widget is draggable; macOS and Windows remember its position between launches.
* Clicking it triggers the same code path as your hands-free shortcut, so the 6-minute cap and the 5-second silence auto-stop apply here too.
# AI dictation
URL: /docs/dictation
How Amical turns your voice into text in any app.
***
title: AI dictation
description: How Amical turns your voice into text in any app.
--------------------------------------------------------------
You hold a key, talk, and the words paste into whatever has focus. That's it. Everything below is the knobs around that loop.
## Two ways to dictate
* **Push to talk** — hold the shortcut (default **Fn**), talk, release. Amical pastes when you let go.
* **Hands-free mode** — press the shortcut (default **Fn+Space**) once to start, again to stop. Useful for longer passages or when you don't want to keep a key held.
Both are bound under **Settings → Shortcuts**. See [Shortcuts](/docs/custom-hotkeys) for when to use which.
## Settings that matter
Open **Settings → Dictation** to find:
* **Auto detect language** — on by default. Amical figures out the language each time. Turn it off to lock to one language; the **Languages** picker below the toggle activates when you do.
* **Microphone** — pick the input device. Defaults to the system default. The list shows every audio input macOS or Windows reports.
* **Formatting** (marked **Alpha**) — runs your transcript through a language model to add punctuation and structure, and adapts the output to the app you're typing into. Needs a local language model or a signed-in Amical Cloud session. See [AI formatting](/docs/ai-formatting) for the full picture, including per-app behaviour and the model picker.
## Limits
* Single recording capped at 6 minutes (Amical warns at 5 and stops at 6).
* If Amical hears no audio for 5 seconds, it stops the recording.
* A push-to-talk press shorter than 500 ms is treated as a slip and discarded.
## Pre-loading the model
**Settings → Advanced → Preload Whisper Model** keeps the local model warm in memory. First-press latency drops; idle RAM goes up. On by default.
# Draft
URL: /docs/draft
Speak an instruction, review the generated text, then insert it. Amical Cloud only.
***
title: Draft
description: Speak an instruction, review the generated text, then insert it. Amical Cloud only.
------------------------------------------------------------------------------------------------
Plain dictation pastes your words the moment you stop. Draft is different: you speak an *instruction*, Amical generates the text, and it waits in a review window so you can read it before anything lands in your document.
Today Draft works only on Amical Cloud. Select **Amical Cloud** as your speech model under **Settings → AI Models → Speech** (see [Cloud AI models](/docs/cloud-ai-models)). With an on-device speech model selected, the Draft shortcut only transcribes your words, like normal dictation.
On-device speech models will work with Draft soon, as long as you have a formatting (language) model configured under **Settings → AI Models → Language**.
Draft is marked **Alpha**: it ships and works, but the behaviour may still change.
## How to use
1. Hold the **Draft** shortcut — default **Fn+Ctrl** on macOS, **Ctrl+Win+Alt** on Windows.
2. Speak your instruction, for example "reply that I'll join the call five minutes late."
3. Release the key. Amical generates a draft and opens the review window.
4. Press **Enter** to insert it into whatever app has focus.
Draft can also work on text you've already selected. Select it, hold the Draft shortcut, and say how to change it — Amical rewrites your selection and shows the result in the review window. Press **Enter** to insert it. You don't need to copy anything first.
## The review window
The generated text sits in a floating window above your work. You can select and scroll it, then:
* **Insert** — press **Enter**, or click **Insert**, to paste the draft into the focused app and close the window.
* **Copy** — click **Copy** to put the draft on your clipboard instead. The button reads **Copied** for a moment.
* **Dismiss** — press **Esc**, or click the **✕** in the top corner, to discard the draft without pasting.
The window stays open until you insert or dismiss it — it won't disappear on its own.
## Redo a draft
Not happy with the result? With the review window still open, hold the Draft shortcut again and speak a new instruction. The window shows **Listening…** while you talk and **Drafting…** while it generates, then swaps in the new text. **Enter** is disabled during this step, so you can't insert the old draft by accident.
Draft is push-to-talk only — the [hands-free](/docs/dictation) toggle won't start it.
## Notes
* The shortcut lives under **Settings → Shortcuts → Draft**; rebind it like any other ([Shortcuts](/docs/custom-hotkeys)).
* Draft writes from your spoken instruction; plain [dictation](/docs/dictation) types your words as-is. Reach for Draft when you want Amical to compose something, not transcribe what you said.
# FAQ
URL: /docs/faq
The questions that keep coming up.
***
title: FAQ
description: The questions that keep coming up.
-----------------------------------------------
## Does Amical work offline?
Yes, once you've downloaded a local Whisper model. **Settings → AI Models → Speech** is where you pick and download one. Cloud models obviously need a connection.
## Where does my data live?
Locally. Transcripts, notes, vocabulary, and audio files all sit in Amical's app-data directory — see **Settings → Advanced → Data Location** for the exact path on your machine. Cloud features (cloud speech, cloud formatting, sign-in) send the audio for that request and nothing else.
## Can I use my own LLM provider?
For language models (the **Formatting Alpha** pass), yes — **Settings → AI Models → Language** lets you connect Ollama or any OpenAI-compatible local endpoint.
For the speech model itself, you pick from the local Whisper variants Amical ships, or use the cloud model. Bringing your own speech provider isn't supported today.
## Why didn't my push-to-talk start?
Three usual causes:
1. The press was shorter than 500 ms — Amical treats that as a slip and discards.
2. Microphone or accessibility permission got revoked. Re-check **System Settings → Privacy & Security**, or relaunch Amical to step back through onboarding.
3. Another app is holding the microphone exclusively (some video conferencing apps do this).
## Can I dictate in a language other than English?
Yes. **Settings → Dictation → Auto detect language** is on by default; Amical will pick the language each recording. To pin one language, turn auto-detect off and choose from the picker.
## Does Amical work on Linux?
Not yet. macOS and Windows ship today. The README's feature roadmap tracks platform plans.
## How do I report a bad transcription?
In **Settings → History**, the **flag icon** on a row sends the audio, transcript, and your feedback to the team. Only available on cloud transcriptions, and only when Anonymous Telemetry is on. The fastest path to a better speech model is using this.
## How do I update Amical?
**Settings → About**. It shows your current version, and **Check for Updates** next to it does the rest. Amical also checks on its own in the background, so often an update is already waiting — the sidebar shows **Update Ready** and the button says **Restart to Install**. Click it and Amical reopens on the new version.
The line under the button tells you what's happening: *Up to date*, *Downloading update…*, or *Update ready — restart to install*.
If About shows a version older than **1.7.7**, that release doesn't have the Check for Updates button yet. Download the latest from [amical.ai/download](https://amical.ai/download) and install it over your current copy — your transcripts, notes, and vocabulary stay put.
## Amical keeps popping up and stealing focus on Windows
Make sure you're on **1.9.3** or newer. Open **Settings → About** to check.
**If it's 1.7.7 or newer,** update from right there: click **Check for Updates**, and when the button changes to **Restart to Install**, click it. Once you're back on 1.9.3 or newer, you're done — if you're already on 1.9.3+ and still seeing this, [file an issue](https://github.com/amicalhq/amical/issues).
**If it's older than 1.7.7** (no Check for Updates button), start Amical through its launcher once:
1. Fully quit Amical — right-click the Amical icon in the system tray and choose **Quit**.
2. Press **Win + R**.
3. Paste `%LocalAppData%\Amical\Amical.exe`.
4. Press **Enter**.
Wait a few seconds for Amical to open, then go back to **Settings → About**. It should now show **1.9.3** or newer — 1.10.0, 1.11.x, and so on. If it still shows the old version, reinstall from [amical.ai/download](https://amical.ai/download).
## How do I upgrade or check my plan?
In the desktop sidebar, sign in and click **Usage & Billing** — it opens your plan in a browser, where you can see your current usage, switch plans, and manage payment. Or go to **app.amical.ai** directly. See [Billing](/docs/billing) for the full breakdown.
## Where do bug reports go?
[GitHub issues](https://github.com/amicalhq/amical/issues). Include your Machine ID (Settings → Advanced) and attach the log file (Download button under Log File Location).
## How do I uninstall cleanly?
* **Settings → Advanced → Reset App** wipes your local data.
* Then drag Amical out of /Applications (macOS) or use Settings → Apps (Windows).
* If you signed in, sign out first if you want to clear the session token.
# What is Amical
URL: /docs
Open-source AI dictation for macOS and Windows that runs on your machine.
***
title: What is Amical
description: Open-source AI dictation for macOS and Windows that runs on your machine.
--------------------------------------------------------------------------------------
Amical is a desktop dictation app. You hold a key (or click the floating widget), talk, and the words land in whatever app you have focused — your editor, a chat window, an email draft. The transcription runs locally with Whisper, or in the cloud if you sign in. Notes you take inside Amical stay with you.
It's open source under MIT and works offline once a local model is set up.
Install, grant permissions, take your first dictation.
Push-to-talk, hands-free, auto-detect language, formatting.
Rebind push-to-talk, hands-free toggle, paste-last-transcript, new note.
Whisper for speech, Ollama for language.
The floating pill: click to dictate hands-free.
The project lives at [github.com/amicalhq/amical](https://github.com/amicalhq/amical). Help shape it on [Discord](https://amical.ai/community).
# iOS quick start
URL: /docs/ios-quick-start
How to install Amical on your iPhone or iPad.
***
title: iOS quick start
description: How to install Amical on your iPhone or iPad.
----------------------------------------------------------
1. Open [amical.ai/beta](/beta) on your iPhone or iPad and tap the **App Store** badge.
2. Install Amical and open it.
3. Sign in with your Amical account.
For crashes or specific issues, drop into the [Discord](https://amical.ai/community).
# Labs
URL: /docs/labs
Opt into experimental features before they become defaults.
***
title: Labs
description: Opt into experimental features before they become defaults.
------------------------------------------------------------------------
Labs is where Amical keeps features that are still experimental. Each one ships off — turn on what you want to try, and expect it to change as it's refined.
Open **Settings → Labs**. Every experiment has a toggle and a short description; some carry a badge that flags a requirement, such as **Cloud only** — it runs on Amical Cloud dictation and leaves local dictation unchanged. Your change saves the moment you flip the switch.
Experiments come and go. As one is refined it graduates into Amical as a default and leaves Labs, and new experiments take its place — so what you see in the app may differ from the screenshot above.
# Local AI models
URL: /docs/local-ai-models
Run speech and language models on your own hardware.
***
title: Local AI models
description: Run speech and language models on your own hardware.
-----------------------------------------------------------------
Amical can run entirely on your machine. No data leaves your computer, no account required.
Open **Settings → AI Models** to find two tabs: **Speech** and **Language**.
## Speech (Whisper)
The **Speech** tab lists every Whisper variant Amical supports, plus the **Amical Cloud** option. Each row shows speed (yellow zaps, 1–5) and accuracy (green dots, 1–5) ratings, plus the download size. Click **Download** on the model you want, watch the progress bar, then pick it from the **Default Speech Model** dropdown above the table.
Variants ship today: **Whisper Tiny** (\~78 MB) → **Whisper Base** (\~148 MB) → **Whisper Small** (\~488 MB) → **Whisper Medium** (\~1.5 GB) → **Whisper Large v3** (\~3.1 GB) → **Whisper Large v3 Turbo** (\~1.5 GB). Pick what fits your hardware.
Already-downloaded models show a delete button instead of a download button. Deleting frees disk; you can re-download later.
## Language (formatting & rewriting)
The **Language** tab pulls from local providers. Two are supported today:
* **Ollama** — install [Ollama](https://ollama.ai) on your machine, pull a model (`ollama pull llama3.2`, etc.), and Amical syncs the list.
* **OpenAI-compatible** — point Amical at any OpenAI-compatible local server (LM Studio, vLLM, llama.cpp servers).
Pick a default from the **Default Language Model** dropdown at the top. The selected model is what the **Formatting Alpha** toggle in Settings → Dictation will use.
## Notes
* Whisper models on macOS run on Apple Silicon's Metal backend. Intel Macs work, just slower.
* The **Preload Whisper Model** toggle in **Settings → Advanced** keeps the model warm in memory for fast first-press latency.
# MCP integrations
URL: /docs/mcp-integrations
Connect Amical to MCP-compatible tools — planned, not yet shipped.
***
title: MCP integrations
description: Connect Amical to MCP-compatible tools — planned, not yet shipped.
-------------------------------------------------------------------------------
MCP support isn't in Amical yet. This page describes the intent, not current behaviour.
The plan is to let Amical talk to any [Model Context Protocol](https://modelcontextprotocol.io) server you connect — your editor, your task tracker, your notes app — so a voice command can read, search, or change things in those tools.
Today, Amical can already talk to language model providers (OpenRouter, Ollama, OpenAI-compatible) under **Settings → AI Models**. MCP would extend that to *tools and data*, not just models.
## Track progress
* [GitHub issues](https://github.com/amicalhq/amical/issues) — propose specific MCP servers you'd use
* [Discord](https://amical.ai/community) — design discussion happens here
# Languages
URL: /docs/multi-language-support
Dictate in many languages, run the app's UI in a few.
***
title: Languages
description: Dictate in many languages, run the app's UI in a few.
------------------------------------------------------------------
Amical handles two language settings independently: the language Amical *transcribes*, and the language the *app interface* is in.
## Dictation language
**Settings → Dictation → Auto detect language** is on by default. Each recording, Amical figures out the language from the audio.
If you only ever dictate in one language and want to skip detection, turn auto-detect off. The **Languages** picker below the toggle lights up — choose your language there. The list is the full set the speech model supports.
## App interface language
**Settings → Preferences → Language** sets the language for menus, settings labels, dialogs, and toasts. Options today:
* **System** — follow your OS locale
* **English**
* **Español**
* **日本語**
* **繁體中文**
Changing this prompts you to restart Amical so the new strings load. You can do it now or wait until next launch.
## Notes
* The two settings are independent. You can run the UI in English and dictate in Spanish, or any other combination.
* New languages are added with the app — file a request on [Discord](https://amical.ai/community) if yours is missing from either list.
# Notes
URL: /docs/note-taking
Voice-first notes that live inside Amical.
***
title: Notes
description: Voice-first notes that live inside Amical.
-------------------------------------------------------
Amical ships a notes panel where every note has a title, an optional emoji, and a body you can dictate or type into. Notes are sorted by most recently edited.
## Create a note
Three ways:
* Press the **New note** shortcut (default **Cmd+Ctrl+N**, bound under Settings → Shortcuts).
* Open **Notes** from the sidebar and click into the list.
* Click the notebook icon on the floating widget when it's visible.
## Auto-dictate on new note
**Settings → Preferences → Auto-dictate on new note** starts a hands-free recording the moment a new note opens. Off by default. Turn it on if you'd rather speak first and edit later.
## Edit a note
Inside a note:
* Click the emoji button to pick one, or leave it blank.
* Click the title to rename.
* The body is a rich-text editor — type or dictate into it like any other text field.
* The three-dot menu deletes the note (with a confirm dialog).
## Notes window
Notes also have a separate floating window when the notes feature is enabled. It opens from the widget's notebook icon and stays on top of other apps so you can dictate while referencing something else on screen.
# Personalization
URL: /docs/personalisation
Per-app skills that tune how Amical formats what you say.
***
title: Personalization
description: Per-app skills that tune how Amical formats what you say.
----------------------------------------------------------------------
Personalization is in beta. Defaults work out of the box; custom prompts are still locked behind a tooltip ("Coming soon") in this build.
**Settings → Personalization** is where you tell Amical *how* it should format your dictation in specific apps and websites. The defaults already cover the common ones — chat, email, notes — and you can layer custom skills on top for anything else.
## How it's organised
The page has two sections:
* **Defaults** — built-in skills (one per app type) that apply automatically based on the focused app. You can change their polishing level, tone, and which apps/sites they cover, but you can't delete them or rename them.
* **Custom** — skills you add yourself. Each one targets a specific list of apps and sites and overrides the defaults there.
Each skill is a row you can expand to edit.
## What's in a skill
Every skill has the same set of knobs:
* **Name** — a label. Custom skills only.
* **Mode** — **Preset** today; **Custom** prompt is shown but disabled with a tooltip ("Coming soon").
* **Preset** *(when mode is Preset)* — one of **Default**, **Personal chat**, **Work chat**, **Email**. Picks the formatting ruleset.
* **Polishing** — **None**, **Low**, **Normal**, **High**. How aggressively Amical cleans up filler words, run-ons, and stutters. Higher polish drifts further from your exact words.
* **Tone** — **Casual** or **Formal**. Affects word choice and punctuation in the rewrite.
* **Apps** — the desktop apps this skill applies to. The **Add app** picker includes Slack, Linear, Cursor, Notion, iMessage, Apple Mail, Outlook, Spark, WhatsApp, Discord, Superhuman. You can also type a bundle id, AUMID, or exe name to cover anything not in the picker.
* **Websites** — the web hostnames this skill applies to. The **Add website** picker includes Gmail (web), Outlook (web), Slack (web), Linear, Notion, WhatsApp (web), Discord, X, GitHub. You can paste any hostname to add it; Amical normalises it (lowercase, no protocol, no path).
Default skills come pre-seeded with sensible Apps and Websites lists. If you've customised either list and want to go back, each list has a **Reset to defaults** link.
## How a skill is chosen at dictation time
When you start dictating, Amical looks at the focused app or the active browser tab's hostname and picks the first matching skill — a custom skill if one matches, otherwise the matching default. If nothing matches, the **Default** skill runs as a catch-all.
This means a custom skill for `slack` will override the built-in chat default when you dictate into Slack, but won't affect anything else.
## Adding a custom skill
1. In the **Custom** section, click **Add new** (or **Add custom** if the section is empty). A new row appears with a placeholder name.
2. Expand it and give it a name (e.g., "Work email — formal").
3. Pick a Preset that's closest to what you want.
4. Set Polishing and Tone.
5. Add the apps and websites it should cover.
6. Hit **Save**. Amical confirms with a toast.
To remove a custom skill, expand its row and click **Delete** (only shown on custom skills — defaults can't be deleted). Use **Cancel** to back out of an edit without saving.
## Notes
* Skills only run when [AI formatting](/docs/ai-formatting) is on — they tell the formatter *how* to format, not whether to.
* Custom prompts are a planned addition (the **Custom** mode toggle is already there, locked). When it lands, you'll be able to write your own system prompt for a skill instead of picking a preset.
* This is the personalisation surface for *formatting behaviour*. For [Custom vocabulary](/docs/custom-vocabulary), [Custom hotkeys](/docs/custom-hotkeys), and the [Preferences](/docs/preferences) pane, see their own pages.
# Preferences
URL: /docs/preferences
Launch behaviour, dock visibility, sound, theme, and UI language.
***
title: Preferences
description: Launch behaviour, dock visibility, sound, theme, and UI language.
------------------------------------------------------------------------------
**Settings → Preferences** is where the app-shell behaviours live — startup, where Amical shows up, what makes a sound, what colour the UI is.
## Startup
* **Launch at login** — Amical starts when you log in. On by default.
## Visibility
* **Show widget while inactive** — keep the floating widget visible even when you're not recording. On by default. Turn it off and the widget only appears mid-recording.
* **Show app in dock** *(macOS only)* — when off, Amical lives only in the menu bar. The main window still opens from the menu bar icon → Open Console.
## Sound
* **Mute system audio during recording** — silences other apps' audio while Amical is recording, so you don't pick up a YouTube tab through your mic. On by default.
* **Mute dictation sounds** — Amical plays a soft start/stop chime by default. Turn this on to silence them.
## Notes
* **Auto-dictate on new note** — when you create a new note, immediately start a hands-free recording. Off by default. Pairs well with the **New note** shortcut bound under Settings → Shortcuts.
## Look
* **Language** — the UI language. Choices: **System, English, Español, 日本語, 繁體中文**. Changing prompts you to restart so the strings reload.
* **Theme** — Light, Dark, or System.
# Desktop quick start
URL: /docs/quick-start
Install Amical, grant permissions, and dictate your first sentence.
***
title: Desktop quick start
description: Install Amical, grant permissions, and dictate your first sentence.
--------------------------------------------------------------------------------
## 1. Install
Download the latest release from GitHub.
`brew install --cask amical`
## 2. First-run onboarding
Launching Amical for the first time walks you through five screens:
1. **Welcome** — pick the features you're interested in (Contextual Dictation, Note Taking; Meeting Transcriptions and Voice Commands are listed but marked coming soon).
2. **Permissions** — grant microphone access. On macOS, also grant accessibility access so Amical can paste into other apps. The screen polls every 2 seconds and unlocks Continue once everything is granted.
3. **Discovery source** — tells the team how you found Amical. Pick one of the seven options.
4. **Model selection** — choose **Cloud** (sign in, free, fast, no setup) or **Local** (Whisper on your machine, private, offline). Amical shows a recommendation based on your hardware. You complete a short setup for whichever you pick.
5. **Completion** — confirm your microphone, set your push-to-talk shortcut, and join the Discord if you want.
## 3. Dictate
Hold your push-to-talk key (set during onboarding; default is **Fn** on Mac), say something into any app's text field, release. The text appears.
For longer passes, use the hands-free shortcut (default **Fn+Space**) to start a session, talk freely, and press it again to stop. The floating pill on your desktop also starts hands-free recording — click it once.
## 4. Where to next
* Customize your shortcuts: see [Custom hotkeys](/docs/custom-hotkeys)
* Add words Amical keeps mishearing: see [Custom vocabulary](/docs/custom-vocabulary)
* Pick a different speech model: see [Local AI models](/docs/local-ai-models) or [Cloud AI models](/docs/cloud-ai-models)
# Snippets
URL: /docs/snippets
Trigger phrases that expand to longer text while you dictate.
***
title: Snippets
description: Trigger phrases that expand to longer text while you dictate.
--------------------------------------------------------------------------
Snippets are short trigger phrases you say while dictating that Amical replaces inline with longer text — addresses, signatures, email templates, anything you re-type often.
Dictate normally — "send the package to my home address please" — and the words "my home address" get swapped for whatever you've defined: a street address, a name and number, a multi-line block. The rest of the sentence flows around it untouched.
## Where to find it
**Settings → Snippets** sits alongside [Custom vocabulary](/docs/custom-vocabulary). The editor takes the same shape: a list of entries, a button to add a new one, and a form for the trigger phrase and the expansion it should produce.
Each snippet has:
* A **trigger phrase** — the words you say to invoke it. Make these distinctive enough you won't say them by accident.
* An **expansion** — the text Amical types in place of the trigger. Plain text; dynamic placeholders aren't supported.
## Practical tips
* Make triggers specific. "My home address" beats "address" because you're unlikely to say the full phrase mid-sentence by accident.
* Keep expansions tight when you can — long expansions make accidental triggers expensive to undo.
* Snippets are personal and live on your machine, the same way your vocabulary does.
# Team
URL: /docs/team
Invite teammates, manage roles, and run Amical as an organization.
***
title: Team
description: Invite teammates, manage roles, and run Amical as an organization.
-------------------------------------------------------------------------------
An Amical organization is a group of accounts under one subscription. You invite teammates by email, they accept, and their seats sit on your plan. Team management is included on Premium and Enterprise; the Free plan is per-individual and has no team surface.
## Where to manage your team
Team management lives in your Amical account in the browser. Two ways to get there:
* **From the desktop app** — sign in, then in the sidebar under your account, click **Invite Team**. Amical opens your account in the browser at the Members page.
* **In a browser** — go to **app.amical.ai** and sign in. The Members tab is in the sidebar under your organization.
A user can belong to more than one organization. Each browser session has one **active** organization at a time; if you're a member of several, switch active org from the organization picker before managing members.
## Members
The Members tab is the main team page. It shows:
* **Active members** — everyone currently in the organization, with their email, name, role (owner / admin / member), and join date.
* **Pending invitations** — invites that have been sent but not accepted yet. You can resend or cancel each one.
### Inviting someone
You invite by email address. Amical sends a magic link the recipient can click to accept. They sign in (or sign up if they're new), the invitation flips to accepted, and they appear in the active member list. If they don't act on the invite, you can resend it or cancel it to free the seat.
Invitations count against your plan's seat limit even while pending. Sending more invites than you have seats will fail until either someone declines, an invite expires, or you free a seat.
### Roles and what they can do
Amical organizations have three roles:
* **Owner** — full control. Can manage members, change billing, transfer ownership to another member, and delete the organization. Every org has at least one owner.
* **Admin** — can manage members and invitations (invite, change roles below admin, remove members). Can't delete the org or change ownership.
* **Member** — the default role. Uses Amical normally; can't manage other members. Can leave the organization themselves.
Changing someone's role is done on their row in the Members table. Owners can change any role; admins can change members up to admin but can't promote to owner — that's an ownership transfer, which is a separate action.
### Removing someone
Removing a member is also done from their row. It revokes their cloud access immediately — cloud transcription and cloud formatting stop working for them. Their local dictation keeps working (local Whisper doesn't need a sign-in), and their on-device data isn't touched.
Removing the last owner isn't allowed; transfer ownership first, then remove.
## Organization settings
The Organization tab is where the org itself lives:
* **Name and slug** — the display name and URL handle. Both editable by owners and admins.
* **Logo** — optional avatar shown next to the org name.
* **Transfer ownership** — hand the owner role to another member (owners only).
* **Delete organization** — permanent. Cancels any active subscription and removes all members. Owners only.
## Seats and the plan
Each member uses one seat. The plan page shows seat usage:
* **Team seats** — how many seats your members fill vs. the plan's seat limit.
* Adding a member when you're at the limit needs a plan change or a seat addition — see [Billing](/docs/billing).
The Premium plan covers small teams; Enterprise lifts the seat ceiling and adds **centralized billing**, **SSO/SAML**, and bulk-pricing discounts.
## Leaving an organization
A member can leave an org from their Account page. After leaving, their cloud access through that org ends. If they belong to other orgs, those keep working; if not, their cloud features fall back to the Free plan.
## Notes
* Membership is per-account, not per-device. A teammate signs in on each device they want cloud features on; the seat is the account.
* Email changes to a member's address (via their Account page) don't break their membership — the link is by account id.
* Bug or invitation didn't arrive? Send a note to **[support@amical.ai](mailto:support@amical.ai)** with the recipient email and the time you sent it.
# Telemetry
URL: /docs/telemetry
What Amical collects, how to turn it off.
***
title: Telemetry
description: What Amical collects, how to turn it off.
------------------------------------------------------
Amical sends product telemetry — usage events, error reports, and quality metadata — to help the team understand how the app is used and where it breaks. The toggle is in **Settings → Advanced → Anonymous Telemetry**, on by default.
## What gets sent
Telemetry events cover:
* **Lifecycle** — app launches, onboarding screens reached, recording started/stopped, errors.
* **Performance** — transcription latency, model load time, queue length.
* **Quality** — when you press the **Report** flag on a history entry, Amical attaches the audio, the transcript, and the feedback you wrote, so the team can improve the speech model.
Telemetry never includes:
* Your transcripts as content (unless you explicitly hit Report).
* Your notes.
* Audio recordings (unless you explicitly hit Report).
* Custom vocabulary entries.
## Turn it off
**Settings → Advanced → Anonymous Telemetry** is a single switch. Flip it off and Amical stops sending events from that point on.
Two side-effects to know about:
* The **Report** flag in **Settings → History** is disabled while telemetry is off — the action and its data live on the same pipe.
* Crash reports stop too, which means we won't see the bug you're hitting unless you tell us about it.
## Machine ID
Each install gets a stable **Machine ID** (Settings → Advanced → Machine ID). It's how telemetry events from one install group together without naming you. The **Copy** button puts it on your clipboard — useful when filing bug reports so the team can find your events.
## Source
Amical is open source. The telemetry events and what they contain live in the codebase at [github.com/amicalhq/amical](https://github.com/amicalhq/amical). If something here doesn't match what you see, the code is the source of truth.
# History
URL: /docs/transcription-history
Every dictation Amical has done, with audio you can replay.
***
title: History
description: Every dictation Amical has done, with audio you can replay.
------------------------------------------------------------------------
**Settings → History** keeps a record of every transcription. You can search it, copy text, replay the audio, or delete entries.
## What's there
* A **Words Dictated** lifetime counter at the top.
* A search box that filters across the visible page (the latest 100 entries by default).
* Entries grouped **Today / Yesterday / Earlier**.
* Per-row actions on hover: copy text, play audio, download audio, retry transcription, report (cloud only), delete.
## Per-row actions
* **Copy** — puts the transcript text on your clipboard.
* **Play / Pause** — plays the audio Amical kept for that recording. Only available when the audio file is still on disk (see retention below).
* **Download** — saves the audio file out somewhere of your choosing.
* **Retry** — re-runs the transcription with your current speech model. Useful if you've switched models since.
* **Report** (flag icon) — sends the audio + transcript to the team to improve quality. Only on cloud transcriptions, and only if Anonymous Telemetry is on (Settings → Advanced).
* **Delete** — removes the entry and its audio.
## Retention
**Settings → Advanced → History Retention** controls how long Amical keeps audio. Options: **1 day, 7 days, 14 days, 28 days, Never**. Default is **Never** — Amical keeps audio indefinitely until you change this or delete entries manually. After the cutoff (if you set one), the audio file is deleted but the transcript text stays.
There's also a **Delete all** action in the three-dot menu next to the search bar — wipes the entire history at once with a confirm dialog.
## Notes
* The list refreshes every 5 seconds while open; new dictations appear without you having to reload.
* The lifetime word counter includes everything you've ever dictated, even after deletion.