Keyboard dictation
Hold Fn to speak, release it to insert the recognized text at the cursor in any app, or press Esc to cancel. The app supports more than 60 languages and can switch languages mid-sentence.
Szept
Szept is a macOS app that combines keyboard dictation, AI text cleanup, screenshots, and screen recording for workflows involving AI agents and other Mac applications. It is intended for users who want to speak or show something on screen and paste the result into the app they are using.
Szept
Szept is a lightweight macOS application for dictation, screen capture, recording, and sending that material to AI-assisted workflows. Its central interaction is keyboard-driven: hold Fn to speak, capture a screen with a shortcut, then add instructions and paste the text or media into the app currently in use, including examples such as Claude Code, Cursor, and chat applications.
The app is designed for macOS 14 and later and combines speech recognition through the user's Soniox account with optional AI cleanup. It keeps screenshots and recordings on the Mac, provides searchable history for dictations and media, and includes source code with a direct purchase so users can adapt the project or have an AI coding agent work with it. A direct purchase is listed at $9.99 for the first 100 buyers, with the site stating that the price then becomes $25 and that there is no subscription.
The privacy policy says Szept has no account, sign-in, analytics, crash reporting, or telemetry in the app. Audio is sent to Soniox while the dictation key is held; optional cleanup or translation sends selected text to the provider configured by the user.
Features
Context-aware AI cleanup
Optional cleanup removes ums, repeats, and false starts, adjusts punctuation, and formats text according to the destination, such as email, chat, or code. It can also use a saved email sign-off.
Screenshot tools
Use Command-Shift-A to capture a window or selected region, then annotate, blur, pin on top, extract text, read a QR code, translate, or make a scrollshot.
Screen recording
Use Option-Shift-R to record a region or the whole screen with system audio and microphone input. Camera bubble, click effects, trimming, and MP4 or GIF export are supported.
History and media library
Dictations, screenshots, and recordings are organized in searchable history and a media timeline. Search works across languages, and entries can be copied and pasted again.
Local settings and permissions
The app stores dictionary terms and language hints locally, keeps API keys in the macOS Keychain, and uses macOS permissions for microphone, Accessibility, screen capture, and optional camera access.
Use cases
Explain a software issue to an AI coding agent
Speak a description of a bug or layout issue, capture the relevant screen region, and hand both the instruction and file path to an agent such as Claude Code or Cursor.
Draft messages without typing
Dictate messages into email, chat, or other apps and use destination-aware cleanup to remove false starts and apply the appropriate formatting before pasting.
Create a quick screen demonstration
Capture a product walkthrough or troubleshooting sequence with system audio, microphone input, optional camera bubble, and click effects, then save it as MP4 or GIF.
Collect visual information
Capture information from an existing screen by extracting text, reading a QR code, translating selected content, or making a scrollshot for later reference.
Find and reuse previous work
Reuse earlier dictated text or locate past captures through multilingual search in the local history and media timeline.
Pros and Cons
Pros
- Combines dictation, screenshots, screen recording, and handoff to AI applications in one keyboard-driven workflow.
- Supports more than 60 languages, including switching languages mid-sentence, with a personal dictionary for names and jargon.
- Provides searchable local history for dictations and a separate media timeline for screenshots and recordings.
- Direct purchase includes the app, source code, agent-oriented project context, and one year of updates.
- Screenshots and recordings are processed on the Mac and the app has no built-in account, telemetry, or crash reporting according to its privacy policy.
Cons
- Speech recognition requires a Soniox account and API key, and audio is streamed to Soniox rather than processed by a local speech model.
- Szept requires macOS 14 or later and several permissions for its dictation and capture workflows.
- Optional AI cleanup sends transcript text and destination context to the selected AI provider, so users need to review provider terms and settings.
FAQ
What platforms does Szept support?
Szept is a Mac app for macOS 14 and later. The site describes a signed, notarized DMG for direct purchasers; a Mac App Store version is also referenced in the privacy policy as Voice Dictate.
How does dictation work?
Hold Fn and speak. When you release the key, Szept inserts the recognized text at the cursor in the current app. You can press Esc to discard the dictation.
What can Szept do with screenshots and recordings?
Press Command-Shift-A, then select a window or drag a region. Szept can annotate, blur, pin, copy text from, translate, or read a QR code from the capture. Screen recordings can be started with Option-Shift-R and saved as MP4 or GIF.
What data leaves the Mac?
Speech is streamed to Soniox using the Soniox account and API key configured by the user. Optional AI cleanup sends the finished transcript and destination context to the selected AI provider; it is off until an API key is stored. Screenshots and recordings are processed on the Mac and only go where the user sends them.
What is included with a direct purchase?
A direct purchase includes the app, a private GitHub repository containing the Swift project and related materials, and one year of updates. The site states that the source license allows covered users to use, modify, build, and keep changes private, with personal and team terms defined in the license.
Quick Facts
- Category
- Productivity and developer tool
- Platform
- macOS 14 or later
- Core workflow
- Dictate or capture, add instructions, and paste into the current app
- Speech recognition
- Soniox account and API key; 60+ languages
- Direct purchase
- $9.99 founding price for the first 100 buyers, then $25; no subscription
- Source code
- Included with direct purchase under the Szept Source License 1.0
Szept Alternativen
Tavus
Tavus is an AI video platform for building real-time, face-to-face agents, digital twins, and AI companions. It combines APIs, custom replicas, and multilingual conversational workflows for developers and teams.
CAKIE
CAKIE is a Mac app for recording your screen and voice for an AI agent. It adds a transcript and screenshot highlights, then lets you copy a prompt and paste or attach the context in another AI app.
SpeakoFlow
SpeakoFlow is a free, open-source desktop voice-to-text app for Windows, macOS, and Linux. It supports dictation into any app, voice-driven writing with Flow, cleanup, translation, and a screen-aware assistant.
Sanota
Sanota is an app that turns spoken memories, reflections, and interviews into clear written stories. It supports personal storytelling, family history, and shared memories, with guided prompts and subscription pricing.
Carbon Voice
Carbon Voice is an asynchronous voice messaging app for teams and individuals, with transcripts, AI catch-up, and cross-device access. It helps people and agents communicate without needing a live call.
QuickQuill
QuickQuill is a macOS dictation and transcription app that runs locally on the device. It helps users record meetings, transcribe audio, generate summaries, and export notes without using a cloud service.