Browse Models
Launch a dedicated cloud GPU server running Laboratory OS to download and run Analog Diffusion using any compatible app or framework.
Must be used with a compatible app, notebook, or codebase. May run slowly, or not work at all, depending on local system resources, particularly GPU(s) and available VRAM.
Generate images and videos using a powerful low-level workflow graph builder - the fastest, most flexible, and most advanced visual generation UI.
Forge is a platform built on top of Stable Diffusion WebUI to make development easier, optimize resource management, speed up inference, and study experimental features.
Automatic1111's legendary web UI for Stable Diffusion, the most comprehensive and full-featured AI image generation application in existence.
Train your own LoRAs and finetunes for Stable Diffusion and Flux using this popular GUI for the Kohya trainers.
Analog Diffusion is a DreamBooth-based text-to-image model built on Stable Diffusion 1.5 that specializes in generating images with analog film photography aesthetics. Trained on diverse analog photographs, the model reproduces characteristic film grain, color palettes, and vintage visual qualities when prompted with the "analog style" activation token, supporting various subjects including portraits, environments, and objects.
Analog Diffusion is a text-to-image generative model developed using DreamBooth and trained on a wide array of analog photographs to capture and reproduce the distinctive aesthetic of analog film photography. This model aims to deliver images that reflect the color palette, texture, and subtle imperfections characteristic of traditional film, providing users with an accessible means to simulate analog styles within their synthetic image generation workflows. Detailed documentation and official model releases can be found on the Analog Diffusion Hugging Face page.
A header collage of eight model outputs, each demonstrating the analog style—featuring both historical and fictional figures, all rendered with distinctive analog film-like softness.
Analog Diffusion was engineered as a DreamBooth-based model, utilizing Stable Diffusion 1.5 as its foundational architecture. The integration of a Variational Autoencoder (VAE) within this architecture is crucial for encoding images into manageable latent space representations and decoding generated samples back to the image domain. The DreamBooth approach allows Analog Diffusion to specialize in analog visual characteristics by fine-tuning on a dedicated dataset, thereby producing outputs that consistently emulate analog photographic styles.
The model was trained with a dataset comprised of a diverse range of analog photographic samples. The training methodology centered around DreamBooth fine-tuning, enabling the model to robustly internalize the defining attributes of analog imagery such as grain structures, color shifting, and filmic contrast curves. This targeted approach allows for the simulation of analog effects independent of the subject matter, supporting high versatility in the generated outputs. The creator has shared specifics about the parameters used for example outputs, which include prompt structures and sampler settings.
The primary feature of Analog Diffusion is its ability to replicate the aesthetic of analog photography on arbitrary text prompts. To invoke this effect, users must include the activation token “analog style” in their prompt. The model provides additional controls for output sharpness and atmospheric haze; by including terms such as “blur” and “haze” in the negative prompt, users can emphasize image clarity, though this may attenuate the analog characteristics. Detailed guidance for prompt engineering is outlined in the model documentation.
Model-generated grid demonstrating the analog style across environments, from natural scenes and interiors to urban and dramatic landscapes. Prompt: 'analog style, snowy house at dusk, Christmas lights; analog style, volcanic eruption night; analog style, cozy attic room', and others.
Analog Diffusion is designed to handle a broad array of subjects, including portraits, characters, animals, and environments. The model’s stylistic treatment is preserved across these varied scenarios, consistently imparting analog hues, contrast profiles, and film-like artifacts. Representative collages shared by the author display outputs such as cinematic portraits, natural vistas, architectural scenes, and animal studies—all unified by the analog signature. Additional uncurated sample batches are available for review through the non-cherrypicked examples archive.
Collage of various subjects generated by Analog Diffusion, including a lion, armored figure, stylized portraits, and an owl, all with strong analog photographic coloration and texture. Prompts included 'analog style, portrait of a lion', 'analog style, person in armor, desert', among others.
While Analog Diffusion was trained exclusively on analog photographs, it has shown a propensity to generate unintended content in certain prompts. The creator recommends using negative prompting to mitigate the generation of unintended content. Additionally, a trade-off exists between maximizing sharpness (by requesting reduced haze and blur) and the preservation of analog authenticity, as increased clarity may diminish the intended filmic effect. Further technical notes and clarification on usage are detailed in the official documentation and release notes.
The model is suitable for creative text-to-image tasks where an authentically analog appearance is desired, such as concept art, synthetic photography, and digital moodboarding. The creator provides a user-accessible Gradio-based interface that allows real-time experimentation with prompts and style parameters. For technical integration, the model checkpoint is freely available for direct download, facilitating research and offline inference as required.