## Description

When using the FAL.ai provider in LiteLLM, users can only access a small fraction of the models that FAL.ai actually offers. The provider uses hardcoded model-specific transformation classes rather than dynamically discovering models from FAL's API.

## Current Implementation

LiteLLM's FAL provider at `litellm/llms/fal_ai/image_generation/` currently has hardcoded support for **only these endpoints**:

| LiteLLM Model ID | FAL Endpoint | |------------------|-------------| | `fal-ai/nano-banana` | `fal-ai/nano-banana` | | `fal-ai/gemini-25-flash-image` | `fal-ai/nano-banana` | | `fal-ai/imagen4/preview` | `fal-ai/imagen4/preview` | | `fal-ai/recraft/v3/text-to-image` | `fal-ai/recraft/v3/text-to-image` | | `fal-ai/bria/text-to-image` | `bria/text-to-image/3.2` | | `fal-ai/flux-pro/v1.1` | `fal-ai/flux-pro/v1.1` | | `fal-ai/flux-pro/v1.1-ultra` | `fal-ai/flux-pro/v1.1-ultra` | | `fal-ai/flux-schnell` | `fal-ai/flux-schnell` | | `fal-ai/bytedance/seedream/v3/text-to-image` | `fal-ai/bytedance/seedream/v3/text-to-image` | | `fal-ai/bytedance/dreamina/v3.1/text-to-image` | `fal-ai/bytedance/dreamina/v3.1/text-to-image` | | `fal-ai/ideogram/v3` | `fal-ai/ideogram/v3` | |…