Amical

Most dictation tools either nail accuracy at the cost of context, or get the auto-formatting right but trip over technical vocabulary. Amical tries to do both: speech-to-text that adjusts the formatting to whatever app you are in, with a custom-vocabulary trainer for the words your industry uses but the model would not know.

What Amical actually does

The tool runs as a system-wide overlay across Mac, Windows, iPhone, and Android, triggered by a single hotkey from any text field. Two transcription modes run side by side: local processing via on-device Whisper for maximum privacy, or cloud models when speed matters more than air-gap secrecy.

• Speak directly into any text field without switching apps or copying and pasting
• Auto-format your dictation to match the context of whatever app you're in, from emails and Slack messages to social media posts
• Run transcription entirely on your device with local Whisper models for maximum privacy
• Transcribe and dictate across 100+ languages with native-level accuracy, including mixed-language dictation in the same session
• Add your own terminology, jargon, and proper nouns to a custom vocabulary so Amical always gets your industry-specific language right
• Automate actions with AI Workflows—say "message Jane on WhatsApp" and Amical handles it via MCP integrations

In practice, the auto-format step is the differentiator. Dictating into Slack and getting a casual one-liner is one experience; dictating into a doc and getting full sentences with proper punctuation is another. Most dictation tools force you to choose one register and stick with it. Amical reads which app has focus and adjusts the output to match. The MCP integrations push it past pure transcription into thin-agent territory: you can chain dictation with actions in any app the tool has access to, which is the part that actually saves time once you set it up.

The custom-vocabulary trainer is the underrated piece. Feed it your industry jargon, your client names, your product SKUs, and the accuracy floor jumps specifically for the words you care about, not just the ones the base model already knew. The privacy promise (no audio retention, no training on your transcriptions) holds on both the local and cloud paths according to the docs, but verify against the policy yourself if you transcribe client work that comes with its own NDA.

Who it's built for

Amical is for people whose job is mostly typing: writers, engineers, lawyers, doctors, customer-support leads, and the steadily growing pool of remote workers who would rather speak a paragraph than type it. The privacy angle matters specifically to anyone handling sensitive material, since the local-model path means your audio never leaves the device. The cloud path matters when you are moving fast and the words do not need air-gap secrecy. Switching between the two is a hotkey, not a setting menu, which is the small detail that keeps people using it.

The deal

Tier 1 starts at $49 (regular $180) and gets one team member 200,000 cloud-transcription words a month plus unlimited transcription via local Whisper models. Tier 2 at $129 adds a three-seat team and pooling of the cloud word budget. Higher tiers scale by team size and word count, up to 25 seats and 5,000,000 cloud words a month on Tier 5. You can [grab Amical on AppSumo](https://www.grabltd.com/recommends/appsumo-amical/) while the deal is live.