Caption Generation
Writing captions is the part of social media that stops people posting. Greenwood Social drafts them for you — either from the media you're posting, or from what your business does.
From your media
Attach an image or video and we write the caption from what's actually in it. For video that means sampling frames and, where there's speech, working from a transcript — so the caption describes the content rather than guessing from a filename.
From your business
You can also generate posts with no media at all, written from your business profile — what you do, who you serve, how you talk. Ask for a batch and you get a mix: most posts leading with something useful, a smaller number mentioning what you sell.
That profile is built during onboarding and improves as you add material. Anything you've given us — documents, your website — grounds what gets written, which is what keeps it from reading like generic filler.
It writes for the platform
Generation knows which platform a post is for and writes to it — within that platform's character limit, with hashtags where they help and none where they're ignored. A caption drafted for TikTok isn't the same shape as one drafted for LinkedIn.
It also writes in the language of your business profile rather than defaulting to English.
Steering the output
If a batch isn't right, say what to change — warmer, shorter, less salesy, more technical — and regenerate. The feedback applies to tone and content, and you can keep adjusting until it sounds like you.
Everything generated is a draft. Nothing publishes until you schedule it, and every caption is editable first — in the create-post form or the bulk grid.
Good to know
- Generation is grounded in what you've given us. It won't invent statistics, customer names, or claims about your business that aren't in your profile.
- Read what it writes before you schedule it. It's a strong first draft, not a substitute for knowing your own business.
- If a generation request is slow or a model is unavailable, we fall back to a second provider rather than failing outright — so a batch rarely comes back empty.
Where it works
In the app when creating a post or filling a bulk session, and during onboarding. Working through an AI assistant over MCP, the assistant writes the copy itself and we publish it — same result, different author.