Flux 2 Lora Gallery Image to Image
Input
Hint: Drag and drop files from your computer, images from web pages, paste from clipboard (Ctrl/Cmd+V), or provide a URL.
Customize your input with more control.
Flux 2 Full Portrait LoRA | [image-to-image]
Black Forest Labs' FLUX.2 LoRA Gallery delivers specialized face-to-full-portrait generation at $0.021 per compute second. Trading single-purpose optimization for flexible prompt control, this endpoint extends cropped facial images into complete body portraits while maintaining facial identity. Built for developers who need rapid character visualization without full model training.
Use Cases: Character Design Prototyping | Profile Photo Extension | Portrait Composition Testing
Performance
At $0.021 per compute second, FLUX.2 LoRA Gallery targets specialized face extension workflows where speed matters more than comprehensive editing capabilities.
| Metric | Result | Context |
|---|---|---|
| Inference Steps | 40 steps (default) | Configurable 1-50 range for speed/quality tradeoff |
| Cost per Compute Second | $0.021 | Approximately 48 seconds of compute per $1.00 on fal |
| Batch Generation | Up to 4 images | Single API call generates multiple variations |
| LoRA Scale Control | 0-2 range | Adjustable face preservation strength (default 1.0) |
| Guidance Scale | 2.5 default | 0-20 range for prompt adherence control |
Specialized Face Extension Architecture
FLUX.2 LoRA Gallery uses a fine-tuned LoRA adapter specifically trained for face-to-body extension tasks. Unlike general image-to-image models that handle arbitrary edits, this endpoint optimizes for a single transformation: extending cropped facial photos into full body portraits while preserving facial identity.
What this means for you:
-
Single-input simplicity: Accepts one cropped face image via URL, no complex masking or region specification required
-
Identity preservation control: LoRA scale parameter (0-2 range) lets you dial facial likeness strength up or down based on how closely you want the output to match the input face
-
Prompt-driven body context: Natural language descriptions ("professional attire", "casual outdoor setting") control clothing, pose, and background without additional reference images
-
Reproducible outputs: Seed parameter enables consistent regeneration for iterative refinement workflows
Technical Specifications
| Spec | Details |
|---|---|
| Architecture | FLUX.2 with face-to-portrait LoRA adapter |
| Input Formats | Single image URL (cropped face) |
| Output Formats | PNG, JPEG, WebP |
| Image Size | Configurable (defaults to input dimensions) |
| License | Commercial use permitted |
API Documentation | Quickstart Guide | Enterprise Pricing
How It Stacks Up
FLUX.1 [dev] Image to Image ($0.030 per inference) – Flux 2 Full Portrait LoRA trades general-purpose editing flexibility for specialized face extension at 30% lower cost. FLUX.1 [dev] handles arbitrary image transformations with broader prompt interpretation, ideal for complex scene editing beyond portrait generation.
Flux 2 Image to Image – The base Flux 2 edit endpoint prioritizes semantic understanding for natural language edits across any image type. Flux 2 Full Portrait LoRA narrows focus to face-to-body transformations with identity preservation controls unavailable in the general endpoint.
FLUX.1 Kontext [pro] – Kontext emphasizes spatial consistency and multi-image coherence for complex scene composition. Flux 2 Full Portrait LoRA optimizes single-face extension workflows where identity preservation matters more than multi-element spatial relationships.