AI Model Integration Policy

AI Model Integration Policy

This page defines the scope and principles for integrating AI models into Darktable.

Darktable is intended to remain a tool for photographic development and enhancement. AI integration must support the improvement of captured images without altering their semantic content or introducing elements that were not present in the original scene.

Execution and Privacy Requirements

All AI models must run entirely locally on the user’s machine. Integration of models that require access to external network services (e.g., cloud-based inference, remote APIs, or online dependencies) is not permitted.

This ensures that image data remains private and under the user’s control at all times, and guarantees that all processing is reproducible, deterministic, and independent of third-party services or connectivity.

Supported Use Cases

The following categories of AI-assisted tools are considered appropriate:

1. Editing Assistance

AI may be used to assist user-driven workflows without modifying image content autonomously.

• Assistance in creating and refining drawn masks.
• Edge detection or subject segmentation to accelerate manual selections.

2. Image Restoration and Correction

AI models may correct defects or limitations introduced by capture devices or aging media, provided they do not alter the scene content.

Examples include:

• Denoising: reducing sensor noise in low-light or high-ISO images.
• Super-resolution (upscaling): improving perceived sharpness or resolution without hallucinating new structures.
• Deblurring: compensating for motion blur or slight focus issues.
• Artifact removal: reducing compression artifacts (e.g., JPEG blocking).
• Dust and scratch removal: cleaning scanned film or prints.
• Color restoration: correcting fading in digitized old photographs.
• Content-aware object removal (restricted): removing objects may be acceptable when the resulting area is reconstructed exclusively from existing image data.
  • Example: removing a sensor dust spot or a small distracting element using surrounding pixels (inpainting based only on local context).
  • Constraint: the reconstruction must not introduce new semantic elements or structures that were not already present elsewhere in the image.

Unsupported Use Cases

The following types of AI models are explicitly out of scope:

• Generative object replacement (not allowed): filling removed areas with newly generated content that is not derived from existing image regions is not permitted.
  • Example: removing a person and synthesizing a plausible background (e.g., generating new buildings, textures, or scenery not visible in the original frame).
• Object insertion: adding new elements not present at capture time.
  • Example: adding animals, people, or artificial objects.
• Scene modification: replacing or significantly altering parts of the image.
  • Example: changing the sky, altering backgrounds, or modifying landscapes.
• Generative transformations: stylistic or structural changes that reinterpret the image.
  • Example: converting a photograph into a painting-like result or changing seasons.

Guiding Principle

All AI-assisted processing must preserve the integrity of the captured scene.

The resulting image should remain a faithful representation of what existed at the time of capture. Enhancements are acceptable only when they compensate for technical limitations or restore degraded data, not when they introduce or remove meaningful content.