I use ChatGPT to help draft posts or brainstorm replies, but the workflow of copying tweets, pasting them into ChatGPT, asking it to reply, and then pasting it back into Twitter was driving me crazy. Also, it almost always sounds like a generic robot
So, I built Tweetback, a Chrome/Edge/Firefox extension that brings your API key directly into the X/Twitter interface.
I wanted to see how far I could push the AI to actually understand context and sound human. Here is what it can do:
"Cloning" Writing Styles
Instead of spending hours writing system prompts, you just type in any public X username. The AI analyzes their past posts to understand their cadence, slang, and formatting, and essentially "clones" their personality for your drafts.
Vision AI & Context
Twitter is highly visual. If you try to reply to a meme using standard text AI, it fails. I integrated Vision AI so the extension actually "looks" at the images in a tweet before drafting a reply. It can also generate AI images to attach to your posts natively.
BYO-Key (No Wrapper Subscriptions)
I hate when developers slap a $20/mo subscription on a basic OpenAI wrapper. This runs entirely on your own API Key (OpenAI, Anthropic, etc.), so you only pay base token costs. (Costs less than 1$ for 50,000 replies).
The Copilot Workflow
It adds UI buttons directly under tweets. You click a tone (or your cloned persona), and it generates a draft right there in the reply box. I also added some automated background modes for the power users (keyword tracking, auto-replying, auto-posting watchlists).
I’d love to hear from people who use AI for social media:
- Do you prefer fully automated trigger-based systems for social media, or human-in-the-loop?
- What feature is missing here that would make this a no-brainer for your workflow?
Website: https://tweetback.ai
submitted by