IDM

IDM is an image generation method using internal distribution matching and feature-moment alignment.

IDM is an image generation approach that augments internal distribution matching with feature-moment matching. It uses the generator's exponential moving average copy to supply a shared feature space for generated and reference samples. The method keeps the full distribution matching correction and compares pooled feature statistics across samples. It relies on the learned generator, text encoder, and variational autoencoder for inference, keeping feature extraction strictly for training. The technique improves aggregate metrics and text preservation for few-step image generation.

IDM at a glance

Best for: Researcher
Pricing: Free
Released: October 2026
Category: AI image tool
Website: inclusion-ai.org

Key features

Augments internal distribution matching with feature-moment matching
Uses exponential moving average feature space for generated and reference samples
Compares pooled feature statistics across samples
Eliminates the need for an external feature encoder
Applies feature mean and coordinate variance adjustments
Retains full distribution matching correction

Use it for

Generating images in one or few steps
Rendering dense text in generated images
Comparing feature statistics between generated and reference samples

Alternatives to IDM

Image or AI — It checks Content Credentials, metadata, and uses a pattern model for evidence.