Uploads and workflows from your assistant

Once LUVI is connected to Claude or ChatGPT, your assistant can upload a file from your computer into a project, and even build a workflow on top of it.

Attach the file to the chat and say which project it should go to. We asked for this perfume photo to go into our Perfume Promo project.

If code execution with network access is on in Claude, the assistant uploads the file itself, and the photo shows up in your library, inside that project.

Without that, in ChatGPT for example, the assistant gives you an upload link instead. It opens LUVI's Upload Files window with your project already selected.

Click Choose files, pick the photo, confirm, then tell the assistant it's done. It finds the file in your library on its own.

Now for the workflow. Describe what you want: we asked it to put the bottle on a black marble counter, then turn that into a five-second video.

The assistant creates an empty workflow, checks which inputs each model takes and wires up the nodes. LUVI checks the graph before saving it; if a connection is missing or wrong, nothing is saved.

Then it works out the cost, about 814 credits for this one, and hands you a link to the workflow.

The link opens on the canvas, with the photo, two prompts, two models and the result nodes all in place. You can open and change any node.

Running it is up to you: you're the one who presses Run.

If you'd rather leave that to the assistant, open the Arm for MCP panel, set a credit ceiling and a time limit, and press Arm. It can then run the workflow within that limit, and the permission lapses on its own when time's up.

When the run finishes, the image and the video appear in the result nodes and in your library.

In short: attach your file, describe what you want, open the link to check it, and press Run.