Screenwriting software stopped being ambitious in 1991. We are not asking for permission to restart it.
Keep reading
The problem was solved, and then everyone stopped.
Final Draft shipped in 1991 and solved a real problem. Screenplay format is unforgiving, margins are law, and before that software you set them by hand or you got read by nobody. It worked. It became the standard. Fair enough.
Then something strange happened: that was treated as the finish line. For three decades the category has iterated on the same object — a word processor that knows where the tab stops go. Collaboration was bolted on. Cloud sync was bolted on. Revision colours, production reports, a darker theme. All of it useful, none of it a new idea about what the software is for.
- 1991Screenplay formatting is automated. The hard part is genuinely solved.
- 2000sCompetitors arrive with the same model at a lower price.
- 2010sThe browser and real-time collaboration are added to the same object.
- 2020sText generation arrives, and gets attached as a sidebar that has not read your script.
- NowThe software still does not know who your characters are.
Look at what a writer actually does between the first card and the shooting draft, and almost none of it is formatting. It is holding a structure in your head. It is remembering that the thing she says on page 74 contradicts what she refused to say on page 12. It is renaming a character and finding her in fourteen other places. It is receiving notes in an email, a Word file and a voice memo, and reconciling them against a PDF that has already moved on.
Every one of those is a job the software could do and does not. So the writer does it, at midnight, in a second document that exists only to track the first one.
Admin is not a small tax. It is where the draft dies.
The cost is not measured in hours. It is measured in the scenes that never get written because the energy went somewhere else first.
You did not get into this to maintain a document.
Writers are not asking for a machine that writes for them. Most of them actively do not want one, and they are right — a tool that produces pages you did not mean is worse than no tool. What they are asking for is much less glamorous and much harder to build: software that has actually read the thing.
That distinction is the entire product. A model that autocompletes your sentence is a party trick. A system that knows the shape of your script, and can tell you that your protagonist has not spoken since page 40, is a collaborator you can leave the room with.
A script is not a text file. We stopped treating it like one.
Underneath Typewrite, a screenplay is not a stream of characters with formatting attached. It is a structure the software can reason about: scenes with identity, characters that are entities rather than repeated strings, references that resolve to a real node instead of matching text.
That sounds like an implementation detail. It is the whole difference.
Sees: seven words. Can check the spelling and set the margins. "Daniel" is a string that also appears 31 other times.
Sees: Mara speaking to Daniel in scene 14. Knows they last shared a scene on page 12, that it ended on the same accusation, and that this is the first time she says it out loud.
Once the software holds that structure, the useful things stop being features and start being consequences. The breakdown is not a document somebody types at midnight — it falls out of the script and moves when the scene moves. The board, the outline, the treatment and the pages are not four files that drift apart; they are four views of one document. Notes attach to a scene rather than arriving as a PDF and an email.
And the AI, when you want it, is reading the same structure you are. That is why we call it a reader and not an assistant. It is off by default, it is one line at a time, and the editor is a complete product without it.
The script is where the whole production starts.
Everything downstream of a screenplay is derived from it. The breakdown, the schedule, the budget, the boards, the call sheets — all of it is somebody reading the script and typing its contents into a different system. Then the script changes, and somebody types it all again.
The industry has accepted this as the natural cost of making things. It is not natural. It is what happens when the document at the top of the chain is dumb, so every tool below it has to re-derive by hand what the script already knew.
Fix the top of the chain and everything below it gets cheaper.
That is the ambition, and we would rather say it out loud than pretend we are only building an editor. Typewrite is the entry point, not the destination. Control the document where creative work actually begins, make it structured and honest, and you earn the right to touch what comes after it.
The creative industries have never had the equivalent of what accounting or design got: a connected system built by people who use it. They got isolated tools that do not talk to each other, sold at enterprise prices to studios and at hostile prices to everyone else. We think that is a category-sized mistake, and we intend to spend the next decade on it.
Start with the writer, because the writer is first and because the writer has been served worst.
Ambition without limits is just noise.
Anyone can announce that they are going to change an industry. The part that is worth reading is what someone rules out. These four are the ones we will be held to.
We will not train on your pages
Your drafts are never training data, on any plan, including the free one. There is no setting to get wrong because there is no setting.
We will not hold your work hostage
PDF and Fountain export at any time, with the full version history, including after you cancel. Leaving has to be as easy as arriving.
We will not price out the people who need it
Students write free, permanently. The free tier is a real tier and not a countdown. Early access pricing is locked for the life of the plan.
We will not ship a ghostwriter
The AI reads, tracks and answers. It does not produce pages you did not mean. It can be turned off entirely, and the editor stands without it.
Who is saying this.
Typewrite began during a screenwriting semester at NABA, and the first users were classmates who had pages due on Monday. It was built because the tools in the room were not good enough, by someone who had to hand in the same assignment.
It is not a pseudonym, not a Discord handle and not a team you cannot find. If you write to the address below, a person answers.
Fade out.
If any of this sounded right, the rest is on the product page.
Wave 01 opens September 2026. Your price stays locked for as long as you keep the plan.