Loading AI tools
Image-generating machine learning model From Wikipedia, the free encyclopedia
Stable Diffusion is a deep learning, text-to-image model released in 2022 based on diffusion techniques. The generative artificial intelligence technology is the premier product of Stability AI and is considered to be a part of the ongoing artificial intelligence boom.
![]() An image generated with Stable Diffusion based on the text prompt "a photograph of an astronaut riding a horse" | |
Original author(s) | Runway, CompVis, and Stability AI |
---|---|
Developer(s) | Stability AI |
Initial release | August 22, 2022 |
Stable release | SDXL 1.0 (model)[1]
/ July 26, 2023 |
Repository | |
Written in | Python[2] |
Type | Text-to-image model |
License | Creative ML OpenRAIL-M |
Website | stability![]() |
It is primarily used to generate detailed images conditioned on text descriptions, though it can also be applied to other tasks such as inpainting, outpainting, and generating image-to-image translations guided by a text prompt.[3] Its development involved researchers from the CompVis Group at Ludwig Maximilian University of Munich and Runway with a computational donation from Stability and training data from non-profit organizations.[4][5][6][7]
Stable Diffusion is a latent diffusion model, a kind of deep generative artificial neural network. Its code and model weights have been released publicly,[8] and it can run on most consumer hardware equipped with a modest GPU with at least 4 GB VRAM. This marked a departure from previous proprietary text-to-image models such as DALL-E and Midjourney which were accessible only via cloud services.[9][10]
Stable Diffusion originated from a project called Latent Diffusion,[11] developed in Germany by researchers at Ludwig Maximilian University in Munich and Heidelberg University. Four of the original 5 authors (Robin Rombach, Andreas Blattmann, Patrick Esser and Dominik Lorenz) later joined Stability AI and released subsequent versions of Stable Diffusion.[12]
The technical license for the model was released by the CompVis group at Ludwig Maximilian University of Munich.[10] Development was led by Patrick Esser of Runway and Robin Rombach of CompVis, who were among the researchers who had earlier invented the latent diffusion model architecture used by Stable Diffusion.[7] Stability AI also credited EleutherAI and LAION (a German nonprofit which assembled the dataset on which Stable Diffusion was trained) as supporters of the project.[7]
Stable Diffusion uses a kind of diffusion model (DM), called a latent diffusion model (LDM), developed by the CompVis group at LMU Munich.[13][8] Introduced in 2015, diffusion models are trained with the objective of removing successive applications of Gaussian noise on training images, which can be thought of as a sequence of denoising autoencoders. Stable Diffusion consists of 3 parts: the variational autoencoder (VAE), U-Net, and an optional text encoder.[14] The VAE encoder compresses the image from pixel space to a smaller dimensional latent space, capturing a more fundamental semantic meaning of the image.[13] Gaussian noise is iteratively applied to the compressed latent representation during forward diffusion.[14] The U-Net block, composed of a ResNet backbone, denoises the output from forward diffusion backwards to obtain a latent representation. Finally, the VAE decoder generates the final image by converting the representation back into pixel space.[14]
The denoising step can be flexibly conditioned on a string of text, an image, or another modality. The encoded conditioning data is exposed to denoising U-Nets via a cross-attention mechanism.[14] For conditioning on text, the fixed, pretrained CLIP ViT-L/14 text encoder is used to transform text prompts to an embedding space.[8] Researchers point to increased computational efficiency for training and generation as an advantage of LDMs.[7][13]
The name diffusion takes inspiration from the thermodynamic diffusion and an important link was made between this purely physical field and deep learning in 2015.[15][16]
With 860 million parameters in the U-Net and 123 million in the text encoder, Stable Diffusion is considered relatively lightweight by 2022 standards, and unlike other diffusion models, it can run on consumer GPUs,[17] and even CPU-only if using the OpenVINO version of Stable Diffusion.[18]
The earliest model was Latent Diffusion Model (LDM) released by CompVis.[11] The SD 1.1 up to SD 2.1 were all based on LDM series uses essentially the same architecture as the earlier. Its UNet backbone takes the following kinds of inputs:
Each run through the UNet backbone produces a predicted noise vector. This noise vector is scaled down and subtracted away from the latent image array, resulting in a slightly less noisy latent image. The denoising is repeated according to a denoising schedule ("noise schedule"), and the output of the last step is processed by the VAE decoder into a finished image.
Similar to the standard U-Net, the U-Net backbone used in the SD 1.5 is essentially composed of down-scaling layers followed by up-scaling layers. However, the UNet backbone has additional modules to allow for it to handle the embedding. As an illustration, we describe a single down-scaling layer in the backbone:
ResBlock
:
SpatialTransformer
, which is essentially the same as a multiheaded cross-attention layer found in a standard Transformer decoder. In the cross-attentional mechanism, the latent array itself serves as the query sequence, one query-vector per pixel. For example, if, at this layer in the UNet, the latent array has dimensions , then the query sequence has vectors, each of which has dimensions. The embedding vector sequence serves as both the key sequence and as the value sequence.In pseudocode,
def ResBlock(latent_array, time_embedding):
latent_array = conv_layer(latent_array)
time_embedding_processed = feedforward_network(time_embedding)
latent_array = latent_array + time_embedding_processed # Broadcasting assumed
latent_array = conv_layer(latent_array)
latent_array = latent_array + feedforward_network(time_embedding) # Broadcasting assumed
def downscaling_layer(latent_array, time_embedding, conditioning_sequence):
latent_array = ResBlock(latent_array, time_embedding)
# SpatialTransformer application
latent_array = SpatialTransformer(latent_array, conditioning_sequence)
return latent_array
The detailed architecture may be found in [19].
This architecture does not have self-attention mechanisms, although the cross-attention mechanisms can be modified into self-attention mechanisms.[20]
The XL version uses the same architecture,[21] except larger: larger UNet backbone, larger cross-attention context, two text encoders instead of one, and trained on multiple aspect ratios (not just the square aspect ratio like previous versions).
The SD XL Refiner, released at the same time, has the same architecture as SD XL, but it was trained for adding fine details to preexisting images via text-conditional img2img.
The 3.0 version[22] completely changes the backbone. Not a UNet, but a Rectified Flow Transformer, which implements the rectified flow method[23][24] with a Transformer.
The Transformer architecture used for SD 3.0 has three "tracks", for original text encoding, transformed text encoding, and image encoding (in latent space). The transformed text encoding and image encoding are mixed during each transformer block.
The architecture is named "multimodal diffusion transformer (MMDiT), where the "multimodal" means that it mixes text and image encodings inside its operations. This differs from previous versions of DiT, where the text encoding affects the image encoding, but not vice versa.
Stable Diffusion was trained on pairs of images and captions taken from LAION-5B, a publicly available dataset derived from Common Crawl data scraped from the web, where 5 billion image-text pairs were classified based on language and filtered into separate datasets by resolution, a predicted likelihood of containing a watermark, and predicted "aesthetic" score (e.g. subjective visual quality).[25] The dataset was created by LAION, a German non-profit which receives funding from Stability AI.[25][26] The Stable Diffusion model was trained on three subsets of LAION-5B: laion2B-en, laion-high-resolution, and laion-aesthetics v2 5+.[25] A third-party analysis of the model's training data identified that out of a smaller subset of 12 million images taken from the original wider dataset used, approximately 47% of the sample size of images came from 100 different domains, with Pinterest taking up 8.5% of the subset, followed by websites such as WordPress, Blogspot, Flickr, DeviantArt and Wikimedia Commons.[citation needed] An investigation by Bayerischer Rundfunk showed that LAION's datasets, hosted on Hugging Face, contain large amounts of private and sensitive data.[27]
The model was initially trained on the laion2B-en and laion-high-resolution subsets, with the last few rounds of training done on LAION-Aesthetics v2 5+, a subset of 600 million captioned images which the LAION-Aesthetics Predictor V2 predicted that humans would, on average, give a score of at least 5 out of 10 when asked to rate how much they liked them.[28][25][29] The LAION-Aesthetics v2 5+ subset also excluded low-resolution images and images which LAION-5B-WatermarkDetection identified as carrying a watermark with greater than 80% probability.[25] Final rounds of training additionally dropped 10% of text conditioning to improve Classifier-Free Diffusion Guidance.[30]
The model was trained using 256 Nvidia A100 GPUs on Amazon Web Services for a total of 150,000 GPU-hours, at a cost of $600,000.[31][32][33]
SD3 was trained at a cost of around $10 million.[34]
Stable Diffusion has issues with degradation and inaccuracies in certain scenarios. Initial releases of the model were trained on a dataset that consists of 512×512 resolution images, meaning that the quality of generated images noticeably degrades when user specifications deviate from its "expected" 512×512 resolution;[35] the version 2.0 update of the Stable Diffusion model later introduced the ability to natively generate images at 768×768 resolution.[36] Another challenge is in generating human limbs due to poor data quality of limbs in the LAION database.[37] The model is insufficiently trained to understand human limbs and faces due to the lack of representative features in the database, and prompting the model to generate images of such type can confound the model.[38] Stable Diffusion XL (SDXL) version 1.0, released in July 2023, introduced native 1024x1024 resolution and improved generation for limbs and text.[39][40]
Accessibility for individual developers can also be a problem. In order to customize the model for new use cases that are not included in the dataset, such as generating anime characters ("waifu diffusion"),[41] new data and further training are required. Fine-tuned adaptations of Stable Diffusion created through additional retraining have been used for a variety of different use-cases, from medical imaging[42] to algorithmically generated music.[43] However, this fine-tuning process is sensitive to the quality of new data; low resolution images or different resolutions from the original data can not only fail to learn the new task but degrade the overall performance of the model. Even when the model is additionally trained on high quality images, it is difficult for individuals to run models in consumer electronics. For example, the training process for waifu-diffusion requires a minimum 30 GB of VRAM,[44] which exceeds the usual resource provided in such consumer GPUs as Nvidia's GeForce 30 series, which has only about 12 GB.[45]
The creators of Stable Diffusion acknowledge the potential for algorithmic bias, as the model was primarily trained on images with English descriptions.[32] As a result, generated images reinforce social biases and are from a western perspective, as the creators note that the model lacks data from other communities and cultures. The model gives more accurate results for prompts that are written in English in comparison to those written in other languages, with western or white cultures often being the default representation.[32]
To address the limitations of the model's initial training, end-users may opt to implement additional training to fine-tune generation outputs to match more specific use-cases, a process also referred to as personalization. There are three methods in which user-accessible fine-tuning can be applied to a Stable Diffusion model checkpoint:
The Stable Diffusion model supports the ability to generate new images from scratch through the use of a text prompt describing elements to be included or omitted from the output.[8] Existing images can be re-drawn by the model to incorporate new elements described by a text prompt (a process known as "guided image synthesis"[50]) through its diffusion-denoising mechanism.[8] In addition, the model also allows the use of prompts to partially alter existing images via inpainting and outpainting, when used with an appropriate user interface that supports such features, of which numerous different open source implementations exist.[51]
Stable Diffusion is recommended to be run with 10 GB or more VRAM, however users with less VRAM may opt to load the weights in float16 precision instead of the default float32 to tradeoff model performance with lower VRAM usage.[35]
The text to image sampling script within Stable Diffusion, known as "txt2img", consumes a text prompt in addition to assorted option parameters covering sampling types, output image dimensions, and seed values. The script outputs an image file based on the model's interpretation of the prompt.[8] Generated images are tagged with an invisible digital watermark to allow users to identify an image as generated by Stable Diffusion,[8] although this watermark loses its efficacy if the image is resized or rotated.[52]
Each txt2img generation will involve a specific seed value which affects the output image. Users may opt to randomize the seed in order to explore different generated outputs, or use the same seed to obtain the same image output as a previously generated image.[35] Users are also able to adjust the number of inference steps for the sampler; a higher value takes a longer duration of time, however a smaller value may result in visual defects.[35] Another configurable option, the classifier-free guidance scale value, allows the user to adjust how closely the output image adheres to the prompt.[30] More experimentative use cases may opt for a lower scale value, while use cases aiming for more specific outputs may use a higher value.[35]
Additional text2img features are provided by front-end implementations of Stable Diffusion, which allow users to modify the weight given to specific parts of the text prompt. Emphasis markers allow users to add or reduce emphasis to keywords by enclosing them with brackets.[53] An alternative method of adjusting weight to parts of the prompt are "negative prompts". Negative prompts are a feature included in some front-end implementations, including Stability AI's own DreamStudio cloud service, and allow the user to specify prompts which the model should avoid during image generation. The specified prompts may be undesirable image features that would otherwise be present within image outputs due to the positive prompts provided by the user, or due to how the model was originally trained, with mangled human hands being a common example.[51][54]
Stable Diffusion also includes another sampling script, "img2img", which consumes a text prompt, path to an existing image, and strength value between 0.0 and 1.0. The script outputs a new image based on the original image that also features elements provided within the text prompt. The strength value denotes the amount of noise added to the output image. A higher strength value produces more variation within the image but may produce an image that is not semantically consistent with the prompt provided.[8]
There are different methods for performing img2img. The main method is SDEdit[55], which first adds noise to an image, then denoises it as usual in text2img.
The ability of img2img to add noise to the original image makes it potentially useful for data anonymization and data augmentation, in which the visual features of image data are changed and anonymized.[56] The same process may also be useful for image upscaling, in which the resolution of an image is increased, with more detail potentially being added to the image.[56] Additionally, Stable Diffusion has been experimented with as a tool for image compression. Compared to JPEG and WebP, the recent methods used for image compression in Stable Diffusion face limitations in preserving small text and faces.[57]
Additional use-cases for image modification via img2img are offered by numerous front-end implementations of the Stable Diffusion model. Inpainting involves selectively modifying a portion of an existing image delineated by a user-provided layer mask, which fills the masked space with newly generated content based on the provided prompt.[51] A dedicated model specifically fine-tuned for inpainting use-cases was created by Stability AI alongside the release of Stable Diffusion 2.0.[36] Conversely, outpainting extends an image beyond its original dimensions, filling the previously empty space with content generated based on the provided prompt.[51]
A depth-guided model, named "depth2img", was introduced with the release of Stable Diffusion 2.0 on November 24, 2022; this model infers the depth of the provided input image, and generates a new output image based on both the text prompt and the depth information, which allows the coherence and depth of the original input image to be maintained in the generated output.[36]
ControlNet[58] is a neural network architecture designed to manage diffusion models by incorporating additional conditions. It duplicates the weights of neural network blocks into a "locked" copy and a "trainable" copy. The "trainable" copy learns the desired condition, while the "locked" copy preserves the original model. This approach ensures that training with small datasets of image pairs does not compromise the integrity of production-ready diffusion models. The "zero convolution" is a 1×1 convolution with both weight and bias initialized to zero. Before training, all zero convolutions produce zero output, preventing any distortion caused by ControlNet. No layer is trained from scratch; the process is still fine-tuning, keeping the original model secure. This method enables training on small-scale or even personal devices.
Stability provides an online image generation service called DreamStudio.[59][60] The company also released an open source version of DreamStudio called StableStudio.[61][62] In addition to Stability's interfaces, many third party open source interfaces exist, such as AUTOMATIC1111 Stable Diffusion Web UI, which is the most popular and offers extra features,[63] Fooocus, which aims to decrease the amount of prompting needed by the user,[64] and ComfyUI, which has a node-based user interface, essentially a visual programming language akin to many 3D modeling applications.[65][66][67]
Version number | Release date | Notes |
---|---|---|
1.1, 1.2, 1.3, 1.4[68] | August 2022 | All released by CompVis. There is no "version 1.0". 1.1 gave rise to 1.2, and 1.2 gave rise to both 1.3 and 1.4.[69] |
1.5[70] | October 2022 | Initialized with the weights of 1.2, not 1.4. Released by RunwayML. |
2.0[71] | November 2022 | Retrained from scratch on a filtered dataset.[72] |
2.1[73] | December 2022 | Initialized with the weights of 2.0. |
XL 1.0[74][21] | July 2023 | The XL 1.0 base model has 3.5 billion parameters, making it
around 3.5x larger than previous versions.[75] |
XL Turbo[76] | November 2023 | Distilled from XL 1.0 to run in fewer diffusion steps.[77] |
3.0[78][22] | February 2024 (early preview) | A family of models, ranging from 800M to 8B parameters. |
Key papers
Training cost
Stable Diffusion claims no rights on generated images and freely gives users the rights of usage to any generated images from the model provided that the image content is not illegal or harmful to individuals.[81]
The images Stable Diffusion was trained on have been filtered without human input, leading to some harmful images and large amounts of private and sensitive information appearing in the training data.[27]
More traditional visual artists have expressed concern that widespread usage of image synthesis software such as Stable Diffusion may eventually lead to human artists, along with photographers, models, cinematographers, and actors, gradually losing commercial viability against AI-based competitors.[82]
Stable Diffusion is notably more permissive in the types of content users may generate, such as violent or sexually explicit imagery, in comparison to other commercial products based on generative AI.[83] Addressing the concerns that the model may be used for abusive purposes, CEO of Stability AI, Emad Mostaque, argues that "[it is] peoples' responsibility as to whether they are ethical, moral, and legal in how they operate this technology",[10] and that putting the capabilities of Stable Diffusion into the hands of the public would result in the technology providing a net benefit, in spite of the potential negative consequences.[10] In addition, Mostaque argues that the intention behind the open availability of Stable Diffusion is to end corporate control and dominance over such technologies, who have previously only developed closed AI systems for image synthesis.[10][83] This is reflected by the fact that any restrictions Stability AI places on the content that users may generate can easily be bypassed due to the availability of the source code.[84]
Controversy around photorealistic sexualized depictions of underage characters have been brought up, due to such images generated by Stable Diffusion being shared on websites such as Pixiv.[85]
In June of 2024, a hack on an extension of ComfyUI, a user interface for Stable Diffusion, took place, with the hackers claiming they targeted users who committed "one of our sins", which included AI-art generation, art theft, promoting cryptocurrency.[86]
In January 2023, three artists, Sarah Andersen, Kelly McKernan, and Karla Ortiz, filed a copyright infringement lawsuit against Stability AI, Midjourney, and DeviantArt, claiming that these companies have infringed the rights of millions of artists by training AI tools on five billion images scraped from the web without the consent of the original artists.[87] The same month, Stability AI was also sued by Getty Images for using its images in the training data.[88]
In July 2023, U.S. District Judge William Orrick inclined to dismiss most of the lawsuit filed by Andersen, McKernan, and Ortiz but allowed them to file a new complaint.[89]
Unlike models like DALL-E, Stable Diffusion makes its source code available,[90][8] along with the model (pretrained weights). It applies the Creative ML OpenRAIL-M license, a form of Responsible AI License (RAIL), to the model (M).[91] The license prohibits certain use cases, including crime, libel, harassment, doxing, "exploiting ... minors", giving medical advice, automatically creating legal obligations, producing legal evidence, and "discriminating against or harming individuals or groups based on ... social behavior or ... personal or personality characteristics ... [or] legally protected characteristics or categories".[92][93] The user owns the rights to their generated output images, and is free to use them commercially.[94]
Seamless Wikipedia browsing. On steroids.
Every time you click a link to Wikipedia, Wiktionary or Wikiquote in your browser's search results, it will show the modern Wikiwand interface.
Wikiwand extension is a five stars, simple, with minimum permission required to keep your browsing private, safe and transparent.