Models Info is a metadata downloader and preview manager for Stable Diffusion Checkpoints and LoRAs. It integrates directly into the SD WebUI Extra Networks cards interface to query online services, fetch thumbnail images, sort models into clean subfolders, and provide quick file management actions.
- Queries online databases to automatically retrieve base model tags, version labels, trained trigger words, and external model links.
- Downloads official preview images and places them alongside your models as local thumbnails.
- Compresses large preview images down to optimized JPEG files (
.preview.jpg) to save disk space.
- Injects a base model selection dropdown directly into the Extra Networks header toolbar.
- Filters model card grids instantly to show only models matching the selected architecture (such as SD 1.5, SDXL, Flux, Pony, SD 3.5) without reloading the page.
- Groups unmatched or custom-made assets under the "Unknown" filter option for easy library cleanup.
- Scans model tags and automatically routes new LoRAs into organized subfolders matching their actual usage categories.
- Built-in categories include:
- Character: For specific fictional characters or real celebrities.
- Style: For artistic styles, paintings, or aesthetic filters.
- Clothing: For outfits, fashion accessories, or costume designs.
- Concept: For generic themes, abstract concepts, or specialized objects.
- Pose: For active poses, body angles, or action-based stances.
- Background: For environments, scenery, or background elements.
- Utility: For training assistants, helpers, and noise offsets.
This extension integrates with two major model platforms:
- Civitai: The global hub for community-created Stable Diffusion models.
- Arcenciel: A hosting platform providing model indexing.
Both APIs are queried in your preferred order with automatic fallbacks if a model is only hosted on one platform.
Manage your model library directly on individual Extra Networks cards:
|
1 | Base Model Badge: Displays the model's architecture (e.g. Illustrious, SDXL). |
| 2 | Open Link (🌐): Open the model's webpage on Civitai or Arcenciel. | |
| 3 | Inject Trigger Words (🏷️): Append the trigger words into the prompt. | |
| 4 | Edit Metadata (⚙️): Manually customize the model's metadata. | |
| 5 | Rename Model (✏️): Rename the model and associated files. | |
| 6 | Delete Model (🗑️): Delete the model and associated files. | |
| 7 | Model Name: The filename or custom-assigned title of the model asset. | |
| 8 | Version Badge: Displays the model's version label (e.g. v1.0, main, booru). |
Go to Settings ➔ Models Info in the WebUI to customize the extension behavior:
- Civitai API Key / Arcenciel API Key: Optional keys to authenticate requests and download private, restricted, or early-access models.
- Primary Metadata Source: Select which service is queried first. If a model is not found on your primary choice, it automatically checks the secondary platform.
- Preview Image Width: Select downscaling dimensions for preview thumbnails (256px, 512px, 1024px, 2048px, or keep original). Smaller sizes load faster in the browser.
- Auto-sort LoRA into subfolders: Enable or disable sorting LoRAs into the built-in categories upon download.
- Maximum NSFW level: Controls model preview filters:
None: Blocks all mature content and only allows clean previews.Soft: Hides explicit or adult-rated previews.Mature: Hides extreme content only.X: Allows all preview content.
- Max concurrent API requests: Adjusts how many parallel downloads run at once to respect platform rate limits.
The model might not be uploaded to Civitai or Arcenciel, or its file hash does not match any entry in their databases. If this happens, you can use the Edit Metadata action button to manually add tags, trigger words, and a custom preview image.
No. The organizer moves both the model file and its companion sidecars together. The WebUI updates its internal directory structure to match when you refresh or restart the interface.
Technically yes, but running both simultaneously is not recommended. Doing so causes redundant API queries, duplicate downloaded images, and visual clutter on the cards. For the best experience, I recommend using only one.
I'm open to adding support for more model hosting websites, provided they have a public, open API that allows looking up model metadata directly by file hash (such as SHA-256). Platforms that only support slow text-based searches or lack a public API cannot be integrated.
