Flux-2-Klein-9B Edit with LoRA delivers higher-fidelity, instruction-based edits combined with custom LoRA adapters. Best for premium stylized edits where lighting, textures, and identity must be preserved while applying a specific look.
About this model
Flux-2-Klein-9B Edit with LoRA delivers higher-fidelity, instruction-based image edits combined with custom LoRA adapters. Built on the 9B base for stronger detail retention and lighting consistency, it shines on premium stylized edits where the result must look polished and on-brand while still respecting the original scene.
The model accepts 1 to 3 reference images, an editing prompt, and up to 3 LoRA adapters with per-LoRA scale weights. Pick this variant when the cost difference vs. the 4B is justified by needing the cleanest possible textures, lighting, and per-pixel detail in the output.
Cost analysis
| Provider | Cost | Notes |
|---|---|---|
| muapiapp | $0.03 per generation | Flat per-image price for the 9B Edit-with-LoRA variant, regardless of LoRA count. |
| Fal.ai | Not available | Flux-2 Klein 9B Edit with LoRA is not generally available on Fal.ai at the time of writing. |
| Replicate | Not available | Flux-2 Klein 9B Edit with LoRA is not generally available on Replicate at the time of writing. |
Flat per-image price for the 9B Edit-with-LoRA variant, regardless of LoRA count.
Flux-2 Klein 9B Edit with LoRA is not generally available on Fal.ai at the time of writing.
Flux-2 Klein 9B Edit with LoRA is not generally available on Replicate at the time of writing.
* Competitor pricing is estimated based on similar model architectures and usage tiers.
Configuration schema
| Parameter | Type | Description | Default |
|---|---|---|---|
| Prompt | string | Editing instruction describing the desired change. | Add ornate gold filigree to the costume while preserving the pose, lighting, and background scene. |
| Image URLs | array | List of 1-3 reference image URLs to edit. | https://d3adwkbyhxyrtq.cloudfront.net/webassets/videomodels/flux-2-klein-9b-edit-in.jpg |
| LoRA List | array | Up to 3 LoRA adapters to apply during the edit. | [object Object] |
| Aspect Ratio | Enum (7 options) | The aspect ratio of the generated image | 1:1 |
Editing instruction describing the desired change.
Add ornate gold filigree to the costume while preserving the pose, lighting, and background scene.List of 1-3 reference image URLs to edit.
https://d3adwkbyhxyrtq.cloudfront.net/webassets/videomodels/flux-2-klein-9b-edit-in.jpgUp to 3 LoRA adapters to apply during the edit.
[object Object]The aspect ratio of the generated image
1:1Developer documentation
Provide reference images: Pass 1 to 3 image URLs in images_list.
Describe the edit: Write a focused prompt that targets specific elements to change while leaving the rest of the scene intact.
Attach LoRAs: Add up to 3 LoRA adapters in lora_list, each with path and scale.
Pick an aspect ratio: Set aspect_ratio to control framing.
Submit and refine: Call flux-2-klein-9b-edit-lora. If the LoRA influence is too strong, lower its scale; if the edit doesn't go far enough, raise the scale or sharpen the prompt.
Frequently asked
Choose 9B Edit when you need higher detail retention, cleaner lighting, and stronger prompt adherence. Choose 4B Edit when cost and throughput matter more than per-image fidelity.
Between 1 and 3 reference images per request, passed in `images_list`.
Yes. The `lora_list` field uses the same `path` + `scale` schema, so the same adapters work on both the edit and text-to-image endpoints.