Fibo-Edit is Bria AI's structured image editing model. Combine JSON parameters + precise masks to make deterministic, reproducible edits with minimal drift. 100% licensed training data for commercial-safe outputs.
Fibo-Edit is Bria AI's JSON + Mask + Image structured instruction image editing model, built for controllable, reproducible, production-ready workflows.
Define precise edit parameters with structured JSON prompts. No more guessing what the AI will do—specify exactly what you want changed.
Target specific regions with pixel-perfect masks. Edit only what you need while preserving the rest of your image intact.
Same input, same output—every time. Fibo-Edit eliminates prompt drift with reproducible results suitable for production pipelines.
Experience structured AI image editing directly in your browser. Upload an image, define your edit, and see deterministic results instantly.
See what Fibo-Edit can do: recolor, relight, replace materials, swap backgrounds, and more—all with precise JSON control.
Recolor: Change product colors while preserving texture and lighting
Relight: Add professional studio lighting to any product shot
Background Replace: Swap backgrounds for contextual product placement
Material Swap: Transform textures and materials with JSON parameters
Fibo-Edit serves developers, designers, e-commerce teams, and ComfyUI power users with production-ready image editing capabilities.
Transform product photography at scale. Generate color variants, swap backgrounds for lifestyle scenes, add professional lighting—all with consistent, reproducible results.
One product → 6 color variants in seconds
Create campaign variations without the prompt lottery. Fibo-Edit's deterministic output means you can iterate on creative concepts while maintaining brand consistency.
Summer vibes, warm tones, outdoor setting
Professional, cool tones, studio backdrop
Holiday theme, festive elements, warm lighting
Build image editing into your applications with Fibo-Edit's comprehensive API. Structured JSON requests, async job handling, and webhook support for production workloads.
import fal_client
result = fal_client.subscribe(
"fal-ai/bria/fibo-edit/edit",
arguments={
"image_url": "https://...",
"prompt": "recolor to blue",
"mask_url": "https://..."
}
)
print(result["image"]["url"])Fibo-Edit has official ComfyUI Partner Nodes. Build complex editing workflows with node-based visual programming and export reproducible JSON workflows.
Load Image Node
Input source image
Fibo-Edit Node
JSON + Mask processing
Save Image Node
Export edited result
Fibo-Edit supports a comprehensive set of image editing operations, all controllable through structured JSON parameters.
Add professional studio lighting, adjust shadows, and change light direction without reshooting.
Change object colors while preserving textures, reflections, and composition integrity.
Swap objects via text descriptions. Replace items in scene while maintaining visual coherence.
Transform textures and materials—matte to glossy, wood to metal, fabric to leather.
Apply artistic styles and visual treatments while preserving subject identity.
Target specific regions with pixel-perfect masks. Edit precisely what you need, nothing more.
Fibo-Edit combines three inputs to produce deterministic, controllable image edits.
Provide the image you want to edit. Supports PNG, JPG, WebP formats up to 4096px.
Specify your edit using structured JSON parameters. Define color changes, lighting, materials, or text-based replacements.
For targeted edits, create a mask to specify exactly which region should be modified.
Receive your edited image. Same inputs always produce the same output—no prompt lottery.
Fibo-Edit uses Visual GenAI Language (VGL) for precise, interpretable edit commands. Here's how to structure your JSON prompts.
image_url
URL to source image (required). Must be publicly accessible.
prompt
Natural language edit instruction. Be specific: "recolor the shirt to navy blue".
mask_url
Optional. URL to mask image defining edit region. White = edit, Black = preserve.
num_inference_steps
Denoising steps (default: 50). Higher = better quality, slower.
guidance_scale
Prompt adherence (default: 5). Higher = stronger edit effect.
{
"image_url": "https://example.com/product.jpg",
"prompt": "recolor the sneakers to deep navy blue, keep the white sole",
"mask_url": "https://example.com/mask.png",
"num_inference_steps": 50,
"guidance_scale": 5,
"seed": 42
}{
"image": {
"url": "https://fal.media/files/...",
"content_type": "image/png",
"width": 1024,
"height": 1024
},
"seed": 42,
"timings": {
"inference": 3.245
}
}Masks let you target specific areas for editing while preserving the rest of your image. Here's how to create effective masks for Fibo-Edit.
White regions = areas to edit
Black regions = areas to preserve
Mask dimensions must match your source image exactly. Mismatched sizes cause errors.
Soft mask edges create smoother blending. Hard edges can cause visible seams.
| Error | Cause | Solution |
|---|---|---|
| Mask size mismatch | Mask resolution ≠ image resolution | Resize mask to match source image exactly |
| Edit bleeding | Mask edges too soft/large | Use tighter mask boundaries |
| Visible seams | Mask edges too hard | Add 2-5px feathering to edges |
| No edit applied | Mask is all black (no edit region) | Ensure white regions mark areas to edit |
Get started with Fibo-Edit API in minutes. Choose your preferred language and copy the code.
# Fibo-Edit API via fal.ai
curl -X POST https://queue.fal.run/fal-ai/bria/fibo-edit/edit \
-H "Authorization: Key $FAL_KEY" \
-H "Content-Type: application/json" \
-d '{
"image_url": "https://example.com/input.jpg",
"prompt": "recolor to vibrant blue",
"mask_url": "https://example.com/mask.png",
"num_inference_steps": 50,
"guidance_scale": 5
}'# pip install fal-client
import fal_client
# Set FAL_KEY environment variable or pass api_key
result = fal_client.subscribe(
"fal-ai/bria/fibo-edit/edit",
arguments={
"image_url": "https://example.com/input.jpg",
"prompt": "recolor to vibrant blue",
"mask_url": "https://example.com/mask.png",
"num_inference_steps": 50,
"guidance_scale": 5
},
with_logs=True
)
print(result["image"]["url"])// npm install @fal-ai/serverless-client
import * as fal from "@fal-ai/serverless-client";
fal.config({
credentials: process.env.FAL_KEY
});
const result = await fal.subscribe("fal-ai/bria/fibo-edit/edit", {
input: {
image_url: "https://example.com/input.jpg",
prompt: "recolor to vibrant blue",
mask_url: "https://example.com/mask.png",
num_inference_steps: 50,
guidance_scale: 5
},
logs: true
});
console.log(result.image.url);per image on fal.ai
average inference time
job queue with webhooks
Complete reference for Fibo-Edit API request and response schemas.
| Parameter | Type | Required |
|---|---|---|
image_url |
string | Yes |
prompt |
string | Yes |
mask_url |
string | No |
num_inference_steps |
integer | No (default: 50) |
guidance_scale |
float | No (default: 5) |
seed |
integer | No (random) |
| Field | Type | Description |
|---|---|---|
image.url |
string | Output image URL |
image.width |
integer | Image width in pixels |
image.height |
integer | Image height in pixels |
seed |
integer | Seed used for generation |
timings |
object | Inference timing data |
For production workloads, use the async pattern with webhooks to process large batches efficiently.
import fal_client
import asyncio
async def process_batch(images):
"""Process multiple images asynchronously"""
tasks = []
for img in images:
# Submit job to queue
handler = fal_client.submit(
"fal-ai/bria/fibo-edit/edit",
arguments={
"image_url": img["url"],
"prompt": img["prompt"],
"mask_url": img.get("mask")
}
)
tasks.append(handler)
# Wait for all results
results = []
for handler in tasks:
result = handler.get() # Blocks until complete
results.append(result["image"]["url"])
return results
# Usage
batch = [
{"url": "https://...", "prompt": "recolor to red"},
{"url": "https://...", "prompt": "recolor to blue"},
{"url": "https://...", "prompt": "recolor to green"}
]
outputs = asyncio.run(process_batch(batch))Fibo-Edit has official ComfyUI nodes released January 2026. Build visual workflows with node-based programming.
Search for "Bria" in ComfyUI Manager and install the official Bria nodes package.
cd ComfyUI/custom_nodes
git clone https://github.com/Bria-AI/ComfyUI-BRIA-API
pip install -r requirements.txtTest Fibo-Edit instantly on Comfy Cloud without local setup.
BRIA Image Edit
Full Fibo-Edit capabilities with mask support
BRIA Outpainting
Extend images beyond original boundaries
BRIA Background Remove
RMBG 2.0 integration for background removal
Pre-built ComfyUI workflows for common Fibo-Edit use cases. Download, import, and customize.
Batch recolor products while preserving textures and lighting.
Replace backgrounds with lifestyle scenes for e-commerce.
Add professional studio lighting to flat product shots.
Convert materials—matte to glossy, wood to metal, etc.
Fibo-Edit open weights are available on Hugging Face for local inference and research.
CPU offloading supported for systems with less VRAM. Commercial use requires Bria AI license.
Choose the deployment model that fits your infrastructure and compliance requirements.
Fastest setup. Use fal.ai or Bria Platform APIs for serverless inference with no infrastructure management.
Run Fibo-Edit on your own hardware using Hugging Face weights and Diffusers pipeline.
Private deployment on your cloud (AWS, Azure, GCP) with Bria enterprise support.
Tips for optimizing Fibo-Edit inference speed and managing costs at scale.
| Volume | Cost |
|---|---|
| 1,000 images | $40 |
| 10,000 images | $400 |
| 100,000 images | $4,000 |
| 1,000,000 images | Contact for volume pricing |
How Fibo-Edit compares to Photoshop Generative Fill, Adobe Firefly, and Stable Diffusion inpainting.
| Feature | Fibo-Edit | Photoshop Gen Fill | Adobe Firefly | SD Inpaint |
|---|---|---|---|---|
| Edit Control | ✓ JSON + Mask | Text prompt | Text prompt | Text + Mask |
| Deterministic Output | ✓ Yes | ✗ No | ✗ No | ✗ No |
| Licensed Training Data | ✓ 100% | ✓ Yes | ✓ Yes | ✗ Mixed |
| API Access | ✓ REST API | ✗ No | ✓ Limited | ✓ Various |
| Open Weights | ✓ Hugging Face | ✗ No | ✗ No | ✓ Yes |
| ComfyUI Support | ✓ Official Nodes | ✗ No | ✗ No | ✓ Yes |
| Commercial Indemnity | ✓ Via Bria | ✓ Yes | ✓ Yes | ✗ No |
Traditional AI image editors produce unpredictable results. Fibo-Edit's structured approach eliminates this.
Traditional AI editors interpret text prompts probabilistically. The same prompt can produce wildly different results each time—we call this "prompt drift" or the "prompt lottery."
"Make it more blue" → Different shades, different areas affected, inconsistent results across runs.
{"prompt": "recolor to #1E40AF navy blue", "mask": "shirt_region.png"} → Same input = same output, every time.
3 runs, 3 different results
Traditional AI
3 runs, identical results
Fibo-Edit
Ready-to-use JSON templates for common editing tasks. Copy, customize, and deploy.
{
"prompt": "place product on white marble surface with soft shadows",
"guidance_scale": 6
}Replace plain backgrounds with lifestyle contexts for product photography.
{
"prompt": "add warm studio lighting from upper left, soft shadows",
"guidance_scale": 5
}Add professional studio lighting to flat or poorly-lit product shots.
{
"prompt": "recolor to deep navy blue, preserve texture and reflections",
"mask_url": "[product_region_mask]"
}Generate SKU color variants from a single product photo.
{
"prompt": "transform to brushed metallic silver finish",
"mask_url": "[surface_mask]"
}Convert surfaces between materials—matte to glossy, plastic to metal, etc.
Solutions for the most common Fibo-Edit issues. Search or browse by error type.
Error: "Mask dimensions do not match image dimensions"
Solution: Resize your mask to exactly match your source image resolution. If your image is 1024x1024, your mask must also be 1024x1024.
Cause: Mask is all black (no edit region defined) or prompt is too vague.
Solution: Ensure white regions in your mask cover the areas you want to edit. Make your prompt more specific: "recolor the shirt to red" instead of "change color".
Error: "Unable to fetch image from URL"
Solution: Ensure your image URL is publicly accessible (not behind authentication). Use direct image URLs, not webpage URLs. Test by opening the URL directly in a browser.
Error: "CUDA out of memory"
Solution: Fibo-Edit's 8B model requires 24GB+ VRAM for full inference. Enable CPU offloading in your inference script, or reduce image resolution. Alternatively, use the cloud API.
Cause: Hard mask edges without feathering.
Solution: Add 2-5 pixels of Gaussian blur or feathering to your mask edges in Photoshop, GIMP, or your image editor of choice.
Track Fibo-Edit updates, new features, and improvements.
Official Bria nodes for ComfyUI now available. Includes BRIA Image Edit, Outpainting, and Background Remove nodes.
Initial release of Fibo-Edit model. 8B parameter DiT architecture with JSON-native structured editing. Open weights on Hugging Face, API on fal.ai.
Fibo-Edit deployment on Replicate platform for additional API options.
Official documentation, guides, and community resources for Fibo-Edit.
Official API reference, authentication, and endpoint documentation.
Interactive playground, pricing, and fal.ai API integration guide.
Partner node introduction and workflow setup guide.
Source code, issues, and community contributions.
Common questions about Fibo-Edit licensing, usage, and capabilities.
Fibo-Edit offers open weights under CC BY-NC 4.0 for non-commercial use. For commercial usage, you need a license from Bria AI. API access via fal.ai costs $0.04 per image.
Yes. Fibo-Edit is trained on 100% licensed data, making outputs commercial-safe. For commercial API usage with full legal indemnification, obtain a license from Bria AI.
Fibo-Edit uses structured JSON parameters instead of probabilistic text prompts. This means deterministic outputs—same input always produces the same output. Combined with precise mask control, you get production-ready results without "prompt drift."
Yes. Official Bria ComfyUI Partner Nodes were released January 2026. Install via ComfyUI Manager or from the Bria GitHub repository. You can also test instantly on Comfy Cloud.
Fibo-Edit's 8B parameter model requires 24GB+ VRAM for full inference. CPU offloading is supported for systems with less VRAM but will be slower. Recommended settings: 50 inference steps, guidance scale 5.
Fibo-Edit supports PNG, JPG, and WebP formats. Maximum resolution is 4096px on the longest side. For best results, use square images (1024x1024) or common aspect ratios.
Images processed via fal.ai are not stored permanently—they're deleted after processing. For enterprise users with strict data requirements, Bria offers private VPC deployment options.
Get weekly Fibo-Edit tutorials, JSON templates, and workflow tips delivered to your inbox.
No spam. Unsubscribe anytime. We respect your privacy.
Join the Fibo-Edit community on GitHub, Discord, and X to share workflows and get help.
Learn Fibo-Edit with in-depth tutorials, comparisons, and use-case guides.
Complete introduction to Fibo-Edit: architecture, capabilities, and use cases.
TutorialGet started with Fibo-Edit API in 5 minutes with copy-paste code examples.
TutorialDeep dive into VGL structured prompts for deterministic editing.
GuideInstall and use official Bria nodes in your ComfyUI workflows.
Use CaseGenerate SKU variants, swap backgrounds, and add studio lighting at scale.
ComparisonHead-to-head comparison of features, pricing, and use cases.
Quick links to all Fibo-Edit resources and documentation.