invokeai-mastodon-node

An InvokeAI custom Invocation node, "Post to Mastodon", that appears in the Workflow Editor and lets you publish a generated image to Mastodon directly from a workflow graph.

This node has zero third-party dependencies. It only makes a plain HTTP POST (stdlib urllib) to a small local bridge service — the actual posting logic (NSFW detection, caption/hashtag generation, Mastodon upload) lives in the companion project, invokeai-mastodon, which you run separately as invokeai-mastodon serve. This split keeps InvokeAI's own Python environment untouched — no TensorFlow, no Mastodon.py, nothing installed alongside InvokeAI's pinned dependencies.

Full step-by-step tutorial with screenshots: docs/tutorial.md

Install

In InvokeAI's UI: Nodes → Custom Nodes → Install Node Pack → Git Repository URL, paste:

https://github.com/Vitexus/invokeai-mastodon-node

Or manually:

git clone https://github.com/Vitexus/invokeai-mastodon-node <INVOKEAI_ROOT>/nodes/invokeai-mastodon-node

Restart InvokeAI. Its log should show Loaded 1 node pack from <root>/nodes: invokeai-mastodon-node.

Use

1. Start the bridge service (see the invokeai-mastodon README for full setup, including Mastodon credentials and NSFW/caption config):

   invokeai-mastodon serve --config config.toml --port 8765

   It binds to 127.0.0.1 only by default — it has no authentication, so never expose it beyond localhost.

2. In InvokeAI's Workflow Editor, add a node and search "mastodon" — Post to Mastodon appears under the Mastodon category. Wire an image into its Image input, leave Service Url at the default unless you changed the bridge's port, and toggle Dry Run for testing without actually posting.

3. Run the graph. The node's output is a status string (or the resulting Mastodon post URL), visible in the node's Outputs panel.

License

MIT