3x Faster Than Typing
Speak 150+ words per minute vs 40 wpm typing. Finish tasks in a third of the time.
Voice Dictation for ML Engineers
Document models, track experiments, ship to production faster
VS Code
Example dictation
so the authentication flow works like this colon new paragraph first the client sends credentials to the auth service second the auth service validates against the database and returns a JWT token third the client stores the token and includes it in subsequent requests new paragraph note colon we're using r s 2 5 6 for token signing
Read this example aloud, or say something of your own.
Ottex outputs
The authentication flow works like this:
1. The client sends credentials to the auth service
2. The auth service validates against the database and returns a JWT token
3. The client stores the token and includes it in subsequent requests
Note: We're using RS256 for token signing
Profile instructions
How it works
Your voice. Right where you work.
Hold Fn
Place your cursor in any app, then hold the Fn key.
Speak naturally
Say what you want to write. Ottex follows your saved writing rules.
Release. It’s written.
Your text appears right where your cursor is.
Why Ottex works for ml engineers
Concrete benefits with real-world impact
Better Context, Better Results
Speaking naturally gives AI agents more context than typing. Describe your intent, explain edge cases, mention file names—your voice carries nuance that typing shortcuts.
Technical Term Accuracy
Cloud AI models excel at technical jargon, code terminology, and specialized vocabulary.
Custom Dictionary
Add names, technical terms, and jargon. Ottex transcribes them accurately every time.
Inline AI Commands
Transform text as you speak: 'Ottex start fix grammar Ottex end'. Edit, improve, translate—all by voice.
Precision Typing
Say 'new paragraph', 'open quote', 'em dash', 'ellipsis'. Full formatting control by voice.
Works with AI Chatbots
Dictate prompts to ChatGPT, Claude, Perplexity—anywhere you type. Better prompts, better answers.
Voice Corrections
Say 'scratch that' to delete mistakes, 'actually' to rephrase. Fix errors without stopping your flow or touching the keyboard.
Global Shortcuts
Start recording with Option+D, use Push-to-Talk, or cancel with Esc. All shortcuts are fully configurable.
Works in Any App
System-wide on macOS—Ottex works wherever you can type. No app-specific integrations needed.
Filler Word Removal
Automatically removes filler words like 'um', 'uh', 'like', 'so' from your transcription. Get clean, professional text without verbal hesitations.
App-Specific Modes
Ottex detects which app you're in and adapts. Coding mode in Cursor, casual tone in Slack, professional in Gmail—automatic context switching.
Context Awareness
Optionally let AI see your screen for smarter responses. Reply to emails with full context, answer messages knowing the conversation history.
Smart Snippets
Define custom text replacements and shortcuts. Say 'work signature' to expand your full email signature, or 'my address' for your contact info.
What you can do
Write Documentation
Dictate technical documentation, READMEs, or API docs. Speaking lets you explain context, edge cases, and examples naturally—better docs with less effort.
Works with
VS Code
Example dictation
so the authentication flow works like this colon new paragraph first the client sends credentials to the auth service second the auth service validates against the database and returns a JWT token third the client stores the token and includes it in subsequent requests new paragraph note colon we're using r s 2 5 6 for token signing
Read this example aloud, or say something of your own.
Ottex outputs
The authentication flow works like this:
1. The client sends credentials to the auth service
2. The auth service validates against the database and returns a JWT token
3. The client stores the token and includes it in subsequent requests
Note: We're using RS256 for token signing
Profile instructions
Prompt ChatGPT & Claude
Dictate prompts to ChatGPT, Claude, Perplexity, or any AI chatbot. Speaking naturally lets you give more context—better context means better results.
ChatGPT
Example dictation
can you help me understand um why the sky is blue like I know it has something to do with light but I'm kinda confused about why it's not violet since violet has an even shorter wavelength question mark explain it simply
Read this example aloud, or say something of your own.
Ottex outputs
Can you help me understand why the sky is blue? I know it has something to do with light but I'm confused about why it's not violet since violet has an even shorter wavelength. Explain it simply.
Profile instructions
Write Research Notes
Dictate research notes, literature summaries, or analysis. Capture your thinking process without losing momentum to typing.
Works with
Notion
Example dictation
H1 Navigation study
H2 Main finding
Participants found the new navigation easy to use but struggled to find export.
Quote I expected export to be next to share end quote.
H2 Next step
Test placing export next to share in the next prototype.
Read this example aloud, or say something of your own.
Ottex outputs
# Navigation study
## Main finding
Participants found the new navigation easy to use but struggled to find export.
> I expected export to be next to share.
## Next step
Test placing export next to share in the next prototype.
Profile instructions
Capture Ideas Quickly
Speak ideas the moment they hit you. Use Ottex anywhere to capture thoughts, then refine them later.
Works with
Text Editor
Example dictation
oh okay so what if we flip the onboarding completely um instead of asking users to set up their workspace first we let them record something immediately like within 10 seconds of signing up they're already talking and seeing the magic actually no not 10 seconds make it the very first screen and then we save that as their first note so they have an aha moment before they even configure anything
Read this example aloud, or say something of your own.
Ottex outputs
What if we flip the onboarding completely? Instead of asking users to set up their workspace first, we let them record something immediately. Within the very first screen of signing up, they're already talking and seeing the magic. Then we save that as their first note—so they have an aha moment before they even configure anything.
Profile instructions
Reply to Work Messages
Reply to Slack messages 3x faster. Dictate casual updates, detailed explanations, or quick acknowledgments without switching context.
Slack
Example dictation
hey um thanks for the update on the API changes I'll uh review the PR today actually tomorrow morning since I'm still working on the database migration and I'll leave comments if I find anything also can you ping me when the tests are passing
Read this example aloud, or say something of your own.
Ottex outputs
Hey, thanks for the update on the API changes! I'll review the PR tomorrow morning since I'm still working on the database migration. I'll leave comments if I find anything. Also, can you ping me when the tests are passing?
Profile instructions
Create Tickets & Tasks
Dictate Jira tickets, Linear issues, or Asana tasks. Speak the context, acceptance criteria, and details—then move on.
Jira
Example dictation
um users are getting logged out unexpectedly after about 30 minutes of inactivity new paragraph steps to reproduce colon new paragraph one login to the app two leave it idle for 30 minutes three try to do anything new paragraph expected colon session should refresh silently new paragraph actual colon user gets kicked to login screen
Read this example aloud, or say something of your own.
Ottex outputs
Users are getting logged out unexpectedly after about 30 minutes of inactivity.
Steps to reproduce:
1. Login to the app
2. Leave it idle for 30 minutes
3. Try to do anything
Expected: Session should refresh silently.
Actual: User gets kicked to login screen.
Profile instructions
Write Meeting Notes
Dictate meeting notes in real-time or right after. Capture action items, decisions, and key points while they're fresh in your mind.
Works with
Notion
Example dictation
action items colon new paragraph one john to review q 4 budget by friday actually thursday two sarah to schedule client call three uh team to finalize roadmap
Read this example aloud, or say something of your own.
Ottex outputs
Action items:
1. John to review Q4 budget by Thursday
2. Sarah to schedule client call
3. Team to finalize roadmap
Profile instructions
Write Status Updates
Dictate standup updates, weekly reports, or project status. Speak what happened and what's next—done.
Works with
Slack
Example dictation
hey team so quick update um finally got the payment integration working took longer than expected because of some weird edge cases with refunds new paragraph today I'm gonna tackle the email notifications and hopefully wrap that up by end of day new paragraph no blockers just need to stay focused
Read this example aloud, or say something of your own.
Ottex outputs
Hey team, quick update—finally got the payment integration working. Took longer than expected because of some weird edge cases with refunds.
Today I'm gonna tackle the email notifications and hopefully wrap that up by end of day.
No blockers, just need to stay focused.
Profile instructions
Frequently asked questions
Does it handle ML and deep learning terminology?
Yes. Cloud models like Gemini 3 Flash are trained on ML papers, PyTorch docs, TensorFlow guides, and thousands of GitHub repos. They handle 'transformer architecture,' 'gradient checkpointing,' 'mixed precision training,' 'distributed data parallel,' and other technical terms accurately. For project-specific terms (model names, internal pipelines), add them to your custom dictionary.
Can I dictate into Jupyter notebooks in VS Code?
Yes. Ottex works anywhere you can type on macOS, including Jupyter notebooks in VS Code, JupyterLab in the browser, and Google Colab. Dictate markdown cells for experiment notes, code comments, and architecture documentation.
How does it help with debugging training issues?
AI assistants like ChatGPT are invaluable for debugging training problems. But explaining your model architecture, loss behavior, and what you've already tried takes time to type. Speaking lets you explain the full context in 60 seconds: your model, the data, the symptoms, what you've tried. Better context = better debugging suggestions.
Will it slow down my machine during model training?
No. Ottex uses less than 2% CPU because it's native Swift, not Electron. Cloud transcription happens via API. When you're training models on GPU or running distributed jobs, Ottex won't compete for compute resources.
Can I use it for MLOps documentation?
Absolutely. Deployment runbooks, model cards, monitoring playbooks, incident postmortems--all benefit from voice dictation. You can explain complex pipelines naturally, including edge cases and failure modes you'd skip while typing. Work modes let you format differently for Notion docs vs Slack updates.
How much does it cost?
Free forever with local models (Whisper runs on your Mac). Or bring your own API key for cloud models--most ML engineers spend ~$1-3/mo. If you want zero setup, the $9/mo premium tier includes Gemini 3 Flash with 99.1% accuracy.
Start documenting models faster
Download free. Use local models immediately, or add your API key for cloud quality. Start dictating experiment notes and deployment docs in minutes.
Try for freeNo credit card required.