Skip to content

Your Installed Models

The installed models dropdown is your personal model library. Every model you add from search or by endpoint ID appears here, organized by category with thumbnails and quick actions.

Click the model selector at the top of the Generate tab to open the dropdown. Models are displayed as individual cards matching the visual style of search results.

Models are organized under category headers — IMAGE TO VIDEO, TEXT TO VIDEO, TEXT TO IMAGE, etc. Category filter chips inside the dropdown let you jump to a specific category without scrolling.

Each card in the dropdown shows:

  • Thumbnail — captured when the model is first added via search. Models added without thumbnails get a lazy backfill the next time the dropdown opens.
  • Model name — the display name from fal.ai’s metadata
  • Category badge — abbreviated category label

Click any model card to select it. The dropdown closes and the Generate tab updates with that model’s form — inputs, parameters, media requirements, and cost estimate all reflect the selected model.

The primary way to add models. Use Model Search to find a model and click +Add. The schema, pricing, and thumbnail are fetched automatically.

For models not yet indexed in search, or custom models deployed on fal.ai:

  1. Open the model dropdown
  2. Use the Add by ID field
  3. Enter the fal.ai endpoint path (e.g., fal-ai/kling-video/v3/pro/text-to-video)
  4. The plugin fetches the schema and registers the model

This is useful for private models, newly released models not yet in the search index, or custom fine-tuned models deployed to fal.ai.

  1. Open the model dropdown
  2. Hover over the model card you want to remove
  3. A remove option appears on hover
  4. Click it — a confirmation overlay appears on the card: “Remove model?”
  5. Confirm to remove

The card animates out and the dropdown stays open. The model is removed from your library but can always be re-added from search.

Removing a model does not delete your generation history or cost data for that model.

Thumbnails are sourced from fal.ai’s model metadata when available. They provide a visual reference for quickly identifying models in the dropdown — especially useful when you have many models installed.

If a model is added without a thumbnail (e.g., via endpoint ID), the dropdown attempts to backfill it from fal.ai on the next open.

Any model deployed on fal.ai with a valid OpenAPI schema can be added to modelBridge. This includes:

  • Public models from fal.ai’s catalog
  • Private models deployed to your fal.ai account
  • Fine-tuned models with custom endpoints

The plugin reads the schema and generates the UI automatically — no special handling needed. If the model uses standard fal.ai field naming, media detection and validation work out of the box.

Your installed models are stored with dual persistence:

  • localStorage — fast reads during normal operation
  • Disk file — survives CEP cache clears and Premiere Pro updates

Model data includes the parsed schema, pricing information, media requirements, thumbnails, and any learned constraints. All of this persists through plugin updates.