FAQ

General

What is Ummless?

Ummless is an AI-powered voice-to-text tool for developers. It captures your speech, transcribes it locally with Whisper, and refines it with Claude using customizable presets.

How is Ummless different from regular dictation?

Regular dictation gives you raw text — typos, filler words, and all. Ummless adds an AI refinement layer that transforms your speech into polished, professional text based on your selected presets.

Is Ummless free?

Ummless offers a free tier with limited daily refinements. Pro and Team plans are coming soon with higher limits and additional features.

Privacy & Security

Does my audio leave my device?

No. All speech-to-text processing happens on-device using Whisper. Your audio is never sent to any external server for transcription.

Is my refined text stored?

Your transcripts and refined text are stored in your Convex database for session history. You can delete any session from the dashboard. Anthropic does not use API data for model training.

Can I use Ummless offline?

Speech-to-text works offline since it runs on-device. However, AI refinement requires an internet connection to reach Claude's API.

Platform & Requirements

What platforms does Ummless support?

Ummless currently supports macOS (Apple Silicon and Intel). Windows and Linux support are planned.

What are the system requirements?

macOS 13 (Ventura) or later. 8 GB RAM minimum, 16 GB recommended. An internet connection is required for AI refinement.

Does it work with any microphone?

Yes, Ummless uses your system's default audio input. Any microphone that works with macOS will work with Ummless.

Usage

How do I start recording?

Press Cmd+Shift+Space to open the recording palette, then press Space or click the record button.

Can I use multiple presets at once?

Yes! Preset stacking lets you combine multiple presets. For example, "Professional" + "Concise" gives you formal, brief text.

How do I create a custom preset?

Go to Presets in the dashboard, click "Create Preset", and write the instructions for how Claude should transform your text.

How accurate is the speech recognition?

Whisper provides excellent accuracy for English and many other languages. Accuracy improves in quiet environments with a good microphone.

Technical

What speech model does Ummless use?

Ummless uses OpenAI's Whisper for on-device speech recognition and Anthropic's Claude for AI text refinement.

Can I self-host Ummless?

Not currently. Ummless requires Convex for real-time sync and Anthropic's API for refinement.

Does Ummless support languages other than English?

Whisper supports many languages for transcription. AI refinement quality varies by language but works well for major languages.