Template — the shape it takes
A brief, not a skeleton.
Say what each part of the piece has to do, roughly how long, and what it must not do. The draft finds its own headings.
- Thought piece
- Build log
- Explainer
- Decision note
In My Words
For solo writers, newsletter authors and founders
In My Words: Your notes.
Your article.
In your words.
Turn notes, transcripts and rough drafts into articles that sound like you.
Bring in what you've already gathered. Pick the shape the piece takes and how it should sound. A model you choose writes the first draft in your voice, and you finish it in a calm editor.
Voice notes, typed upMeeting transcriptsRough draftsMarkdownPlain textCSVJSONA .zip exportHalf an outlineBrain dumps
Import Ideas — ⇧⌘I
Bring in what you have already gathered.
-
1
Drop in your notes
Markdown, text, CSV or JSON files, or a whole .zip export. Give the piece a working title.
-
2
Choose a shape and a voice
A template for the shape it takes, a voice for how it sounds. Add a direction, or let it suggest one from the files.
-
3
Press Go
⌘↩ and the first draft streams into the editor, usually in 15 to 60 seconds. Your notes are kept alongside it.
The first draft — live
Your notes become your article.
Press Go and the draft is set in front of you, line by line, written from your notes and your direction in the shape you picked. Your notes stay with the article as its base knowledge. Stop it whenever you like.
These are real frames from the app, about a third of a second apart. Drag the rule to scrub through.
Two libraries — kept apart on purpose
The shape it takes. How it sounds.
A template is reusable across voices. A voice is reusable across shapes. So a build log, a briefing note or a thought piece can all come out sounding like you, without rewriting either.
Voice — how it sounds
Habits, not adjectives.
A voice holds your writing style, banned phrases, the audio voice that reads it aloud and the look of its pictures. Answer an eight-question interview and press Write the voice, or start from blank.
- Plain first person
- Briefing note
- Explainer
- Field notes
“Opens on a concrete moment rather than a claim.”
“Ends on a door left open, not a summary.”
From the built-in Plain first person voice
The editor — calm, and yours
Then make it properly yours.
A Markdown editor with headings, bold and italic, links, quotes, lists, highlights and comments. Type / for images, galleries and video. Tags, folders, search and a split view with ⌘\.
Change the shape or the voice.
Pick another template or voice and press Apply template & voice. It reshapes the draft and rewrites how it sounds in one pass, and keeps your facts and examples.
The same panel makes the read-aloud audio.
Notes to yourself, kept to yourself.
Comment on a passage with ⌥⌘M, or highlight it with ⇧⌘H. Resolve comments as you go.
Comments are never sent to the model, unless a prompt of your own asks for them.
The jobs you'd do anyway.
Check spelling and grammar reads the whole piece and leaves the voice's habits alone. Suggest image prompts marks where a picture would earn its place.
Write your own prompts with tokens like {{article}} and {{selection}}, and choose what happens when one finishes: show the reply, insert it at the cursor, or replace the article.
Version history
Every draft is kept. Step through Latest, each Draft and your Base knowledge, then press Use this version. Swipe with two fingers, or use ⌥⌘Z and ⌥⇧⌘Z.
- ⇧⌘I
- Import Ideas
- ⌘↩
- Go
- ⇧⌘H
- Highlight
- ⌥⌘M
- Comment
- ⌘K
- Link
- ⌘\
- Split editor
- ⇧⌘E
- Export
- ⌥⌘P
- Published View
Finishing — pictures, sound, publishing
Ready to go out.
Pictures
Add image blocks and press Generate image. Images come from OpenAI's gpt-image-1-mini, drawn in your voice's image style, so the pictures match the writing.
Read aloud
Generate audio saves article.mp3 in the article's folder. Choose an OpenAI voice, such as Fable, Nova or Sage, or one of your ElevenLabs voices, including ones you've cloned.
Publish
Send the piece to Substack as a draft, with its images, videos and voiceover. Write a LinkedIn post and thumbnail, then copy the post and open LinkedIn. Or export Markdown, HTML or plain text.
Your folder, your keys
Plain files. Your own keys.
Your library is Markdown and JSON in a folder you choose. Open it in any editor, search it, back it up however you like.
You bring the models. Use OpenAI or any OpenAI-compatible server, including one on your own Mac, Anthropic Claude for writing, and ElevenLabs for speech. Keys live in the macOS Keychain.
When you ask for a draft, a picture or audio, the text that job needs goes to the provider you picked, under their terms. The privacy policy lists every connection the app makes.
Writing/ ├─ articles/ │ └─ <id>/ one per article │ ├─ article.md │ ├─ source.md │ ├─ sources/ │ ├─ images/ │ ├─ audio/ article.mp3 │ ├─ versions/ │ └─ comments.json ├─ voices/ ├─ templates/ └─ prompts/
The film — 40 seconds
From notes to article.
No tracking. No analytics. No accounts.
No sign-inOpen it and start writing. There's no account to make.
Your library stays herePlain files, in a folder you choose, on this Mac.
Your own keysYou pick the provider for each job, and pay them directly.
Nothing reported backNo telemetry, no analytics, no crash reporting.
Support
Questions, answered.
Something not working, or an idea for the app? Email [email protected] with your Mac model and macOS version. Replies usually within two working days.
Which AI services does it work with?
OpenAI, or any OpenAI-compatible server, including one running on your own Mac. Anthropic Claude for writing. ElevenLabs for speech. You add your own API keys in Settings › Providers, and choose which model does which job in Settings › Models.
Does my writing leave my Mac?
Your library stays on this Mac, as plain files in a folder you choose. When you ask for a draft, an image or audio, the text that job needs goes to the provider you picked, under their terms. Comments are never sent unless your own prompt asks for them.
What can I bring in?
Markdown, plain text, CSV and JSON files, or a .zip export, which gets unpacked. Open Import Ideas with ⇧⌘I, drop your files in, give it a working title and press Go (⌘↩).
How long does a first draft take?
Usually 15 to 60 seconds, depending on the model and how much you bring in. The draft streams into the editor as it's written, and you can stop it at any time.
What's the difference between a template and a voice?
A template is the shape a piece takes: a short brief that says what each part has to do. A voice is how it sounds: your writing style, phrases to avoid, and the audio voice and image style that go with it. Keep them separate and any shape can be written in any voice.
Can it post to Substack or LinkedIn for me?
It creates a Substack draft, with text, images, videos and voiceover, for you to review and publish on Substack. You sign in on substack.com inside the app, so it never sees your password. For LinkedIn there's no API: it helps you write the post and a thumbnail, then copies the post and opens LinkedIn for you to paste.
Can I open my articles in other apps?
Yes. Each article is a folder with article.md, your sources, images, audio, earlier versions and comments.json. Voices, templates and prompts are plain files too. You can also export Markdown, HTML or plain text with ⇧⌘E.
Which Macs does it run on, and when is it out?
Macs with Apple silicon running macOS 14 or later. It's coming soon. Email [email protected] and we'll tell you when it's ready.
How your data is handled, in full: In My Words privacy policy.