
New in Social: teach ScriptCut how a client writes and how their covers look, and every caption or cover you make for them afterward follows those rules automatically.
Inside a client folder's Client kit (Social > Accounts), you can now set two things once per client: how they write, and how their covers look. Paste a handful of that client's own published captions and ScriptCut reads them and writes down their style in plain English: sentence length, whether they write as we or you, emoji and hashtag habits, how they open a caption, what they never say. Upload a few covers or thumbnails they have already published and it writes down their visual rules the same way: colours, how much text sits on a cover and where, whether a person appears, the register. There is also a one-time client brief, who they are, who they sell to, what they must never claim, that sits in front of everything ScriptCut makes for them.
Every new client used to mean starting from a blank AI with no sense of their voice or their look. You'd write a caption, then rewrite it to sound like them. You'd design a cover, then redo the text placement because it didn't match their feed. Client Voice and Look moves that work to a one-time setup instead of a fix-it pass on every single post.
Say you run social for a coffee brand. You paste in ten of their past captions and upload five covers they've posted before. ScriptCut reads both and shows you its notes as editable text: short sentences, always we, one emoji per caption, never mentions price. You can change or delete any line before it's used. From then on, every caption and cover you generate for that client starts from those rules instead of a blank page. Learning again later, say after the brand shifts tone, updates the existing rules rather than replacing them, so one more good example refines what's already agreed.
Set it up once for a client and try it at ScriptCut.
In Social > Accounts, inside a client folder's Client kit, where you set Captions rules, Covers rules, and the client brief.
A handful of that client's own published captions for the writing style, and a few covers or thumbnails they've already published for the visual style.
No. ScriptCut writes down the rules as plain, editable text you can read and change. It does not train a model, and the example images are read once and not kept.
Yes. Both the captions rules and the covers rules are shown as editable text, so you can change or delete any line.
Learning again updates the existing rules rather than replacing them, so adding one more good example refines what's already agreed.
It's a one-time note on who the client is, who they sell to, and what they must never claim, and it sits in front of everything ScriptCut makes for that client.
That client behaves exactly as before, with no style rules applied.