ModelScope icon

ModelScope

Interact with ModelScope API-Inference services

Actions3

Vision Model → Vision Chat

AI-generated

Summary

Sends a prompt and one image to a selected ModelScope vision model through chat completion. Choose an image URL or an incoming binary property; the operation returns the model response with token usage and image-source metadata.

Inputs

  • Model (required) — Select a vision model; defaults to Qwen/Qwen3.8-Flash-Next.
  • Image Source (required) — Choose Image URL (default) or Binary.
  • Image URL (required) — When Image Source is Image URL, provide the URL of the image to analyze.
  • Binary Property (required) — When Image Source is Binary, name the incoming item's binary property containing the image; defaults to data.
  • Prompt (required) — Question or instruction for the image; defaults to “请描述这张图片的内容”. Blank prompts fail.
  • Temperature — Controls response randomness from 0 to 2; defaults to 0.7.
  • Max Tokens — Maximum generated tokens, from 1 to 8192; defaults to 2048.

Output shape

A single JSON object per input item, returned on the main output.

Includes the chat completion fields (such as id, model, choices, and usage), status, processing time, input/output/total token counts, and completion timestamp. Also includes image_source and image metadata: URL input is reported in image_url; binary input reports the binary property name, MIME type, and byte count. The node requires ModelScope API credentials. If Continue On Fail is enabled, a failed item returns an error field instead.

Links

Discussion