Generative Fill is unavailable in 32-bit HDR documents because Adobe Firefly’s cloud diffusion models operate exclusively within standard dynamic range (SDR) integer color spaces. When working on high dynamic range files—such as 32-bit floating-point .exr, .hdr, or 32-bit TIFF assets—the Generative Fill button is greyed out and generative options disappear from the Contextual Task Bar. Diffusion models cannot synthesize scene-referred floating-point luminance values above standard display white (1.0). To use Generative Fill on an HDR asset, you must downsample the canvas to a 16-bit integer space or use a non-destructive round-trip workflow.
Fast Fix
Go to Image > Mode and select 16 Bits/Channel. In the HDR Toning window that opens, set the Method to Exposure and Gamma to preserve your scene’s lighting values without tone-mapping color shifts. Click OK. Generative Fill will activate immediately. Run your generation, then composite the generated layer back into your 32-bit master file.
Why 32-Bit Floating-Point Files Invalidate Generative Models
To understand why 32-bit documents disable generative tools, it helps to examine how dynamic range is represented in digital images:
Integer Color Spaces (8-bit and 16-bit) ---> Supported
├── Fixed range between 0 (Pure Black) and 255/32768 (Display White).
└── Firefly's neural networks are trained natively on these standardized boundaries.
Floating-Point Color Spaces (32-bit HDR) ---> Unsupported
├── Open-ended luminance values (values can exceed 1.0, representing actual light intensities).
└── Diffusion models cannot predict or match unbounded physical luminance mathematically.
If Photoshop allowed Generative Fill to execute inside a native 32-bit floating-point environment, the server would return an SDR integer pixel block. Inserting an SDR block into an HDR composite produces severe visual artifacts:
- The generated area looks completely clipped or dull when viewed under HDR exposure adjustments.
- Brightness and exposure sliders fail to adjust the generated content and the original 32-bit plates equally.
- Color grading operations in linear space cause visible seams along the selection boundaries.
For a look at how other color modes affect AI availability, see Why AI Tools Disappear When Working in “Grayscale” Mode.
Diagnostic Flow: Document Mode Verification
Before attempting workarounds, confirm whether bit depth is the only restriction affecting your canvas:
Generative Fill Disabled
│
├── Check Document Header Tab: Does it say "(RGB/32)"?
│ │
│ ├── YES ──► Root cause: 32-bit floating-point color restriction.
│ │ Solution: Use the Non-Destructive Round-Trip Workflow below.
│ │
│ └── NO ──► Check if mode is "(RGB/8)" or "(RGB/16)".
│ If Generative Fill remains greyed out on standard bit depths:
│ See [Why the "Generative Fill" Button is Grayed Out (M4 Mac Fix)](<http://www.adobeopsguide.com/generative-ai/interface-fixes/generative-fill-button-grayed-out-m4-mac>).
If generative tools are missing from the Crop Tool as well, confirm your settings using Why “Generative Expand” is Missing from the Crop Tool.
The Non-Destructive 32-Bit Round-Trip Workflow
If you are a VFX artist, 3D lighter, or landscape photographer who must deliver a final 32-bit master document, you should not permanently downsample your working master. Instead, use this round-trip procedure to generate content in 16-bit and re-integrate it into your 32-bit workspace:
32-Bit Pipeline
┌─────────────────────────────────────────────────────────────┐
│ 1. Master 32-bit File (Unbroken high dynamic range values) │
└──────────────────────────────┬──────────────────────────────┘
▼ (Crop & Export Selection)
┌─────────────────────────────────────────────────────────────┐
│ 2. Temporary 16-bit Integer File │
│ - Downsampled via Exposure & Gamma │
│ - Run Generative Fill / Generative Expand │
└──────────────────────────────┬──────────────────────────────┘
▼ (Copy Generated Layer)
┌─────────────────────────────────────────────────────────────┐
│ 3. Paste Back to 32-bit Master Canvas │
│ - Convert layer to 32-bit space │
│ - Match exposure using 32-bit Exposure Adjustment Layer │
└─────────────────────────────────────────────────────────────┘
Step 1: Isolate the Target Area
- In your 32-bit document, make a selection around the area you need to modify or expand using the Lasso Tool (L).
- Press Ctrl+Shift+C (Windows) or Cmd+Shift+C (macOS) to create a merged copy of the selected region.
- Create a new document (Ctrl+N / Cmd+N). Photoshop will automatically set the canvas dimensions to match your clipboard data.
- Set the color mode of the new document to RGB Color and 16 Bits/Channel. Click Create.
- Paste your selection into the new document (Ctrl+V / Cmd+V).
Step 2: Execute Generative Fill
- With your 16-bit document open, make your selection across the area needing generative filling or background expansion.
- In the Contextual Task Bar, the sparkle icon and Generative Fill button will now be active. If the sparkle icon is still missing, consult How to Fix the “Missing Sparkle Icon” on AI Features.
- Enter your prompt or leave the prompt empty to sample surrounding textures. Click Generate.
- Pick your preferred variation from the Properties panel.
Step 3: Composite Back into the 32-Bit Master
- Select the generated layer in your 16-bit document.
- Right-click the layer in the Layers panel and select Duplicate Layer.
- In the Destination dropdown, select your original 32-bit master document. Click OK.
- Return to your 32-bit document. The generated layer will appear in your layer stack.
- If the generated element appears slightly darker than surrounding HDR light sources, add an Exposure adjustment layer directly above the pasted layer.
- Alt-click (Windows) or Option-click (macOS) between the Exposure adjustment layer and the generated layer to create a Clipping Mask.
- Adjust the Exposure slider until the highlights in the generated layer match the luminance of the surrounding 32-bit scene.
Alternative: Direct In-Place Downsampling
If your document does not require floating-point light values for VFX rendering pipelines and only needs to be viewed on standard displays:
- Go to Image > Mode > 16 Bits/Channel.
- When the HDR Toning dialog appears, select the Method dropdown.
- Do not choose Local Adaptation if you want to avoid flattening overall scene contrast. Instead, choose Exposure and Gamma.
- Leave Exposure at
0.00and Gamma at1.00, then click OK. - When prompted, choose Don’t Flatten to preserve your existing layer stack.
- The entire canvas is now operating in a 16-bit integer space. Generative Fill and Generative Expand can now be used natively across all layers.
Retouching 32-Bit Assets Without Mode Conversion
If your production pipeline forbids converting documents to 16-bit—even temporarily—cloud generative tools cannot be used.
However, you can still perform object removal and scratch cleanup using local tools:
- The native Remove Tool can operate within certain high-bit workflows when configured for on-device processing.
- Because local tools process pixels on your workstation hardware rather than sending integer arrays to cloud servers, localized texture reconstruction can often be handled directly on the canvas.
- To configure your tools for local processing, read How to Use the “Remove Tool” Locally When Gen-Fill is Blocked.
- If your GPU struggles to process high-resolution buffers during these operations, review Why HDR Renders Fail on SDR GPU Workstations.