Summary
Prompt in, file on disk.
Most image tools hand back a link you then have to download. This one takes the folder you name and writes the picture there, so the result of the call is a file you can already open.
What it is
An image generator wired to Volcengine's Doubao Seedream model. You give it a prompt, a filename, a destination folder and an aspect ratio; what comes back is a picture already saved to disk rather than a link you then have to fetch.
What you get
- Text-to-image generation from a prompt written in Chinese or in English
- Square, landscape and portrait shapes — 1:1, 4:3, 16:9, 3:4 and 9:16 — chosen per call rather than fixed at setup
- The file written into a folder you name, with the path and its permissions checked before generation starts
- Retries on a failed model call, and a specific message when it still fails
Requirements
Python 3.10 or newer with uv, and a Volcengine Ark API key from console.volcengine.com. The key goes in your shell environment; the server runs locally and writes into whatever folder the call points it at. One image per call — the model does not batch.