Everywhere you turn, all you hear about is AI nowadays. With the rapid progressions in AI, it almost feels like one day it was non-existent and then all of a sudden it’s all around you. With various AI tools being bought into existence and adapting to various tools, there is a lot to explore in this field and a lot of benefits to be reaped from it.

‍

Have you ever had a clear image in your mind? A piece of creativity that you can describe down to the smallest element of it but cannot for the life of you put on paper? Every one of us has these moments, where we can see something in perfect clarity in our mind yet when we try to make it tangible, it still remains intangible. Generative AI has turned on tides on imagination and anything you can imagine is as good as true (for the most part).

‍

Generative AI has already disrupted the tech industry but even AI models are being “disrupted“ by other AI models. Diffusion models learn by corrupting their training models by adding noise and then reverse-engineering to retrieve the original image. This denoising process is then used to generate clean outputs with random inputs. AI-image generator tools like DALL-E and Midjourney use diffusion models to generate images. But anyone using the diffusion model owes it to one company and one company only - Stability.ai, the startup that created this model.

‍

AI by the people, for the people

Founded and bootstrapped by Emad Mostaque in 2020, stability.ai was founded because of Emad’s personal interest in AI and a lack of organization in the open-source AI community.

‍

“AI promises to solve some of humanity’s biggest challenges. But we will only realize this potential if the technology is open and accessible to all. Stability AI puts the power back into the hands of developer communities and opens the door for groundbreaking new applications. An independent entity in this space supporting these communities can create real value and change.” - Emad Mostaque (source)

‍

Stability.ai motto is AI by the people, for the people. Their aim is to provide a platform to ‘awaken’ humanity’s potential. Creator Emad said that he wants to solve the problem that most researchers face when creating AI, that is there is no feasible way to create an AI without any catches, or external influences. Stability.AI is an attempt to allow innovators to create and concentrate on their AI models without worrying about external factors. This has proven to be true as Stability. AI-supported research communities are actively producing ground-breaking AI models for image, language, code, audio, video, 3D content, design, biotech, and other scientific research.

‍

It is a generative AI with various models such as -

• Harmonai – This is used to create generative audio models. Harmonai's Dance Diffusion uses machine learning to create new music from scratch. Users can create audio from an existing model, upload and regenerate audio samples, or interpolate between two tracks. Users receive audio tracks composed entirely by the neural network after following the methods in the notebook. Although this seems impressive, it’s imperative to remember that Harmonai is still in its developmental stages and the ending result can be rather scratchy and grainy.
• CarperAI – which understands human preferences through automation.
• Clipdrop – which is the ultimate ecosystem of apps, plugins & resources for all creators. Clipdrop is an app suite that allows you to easily edit photographs with AI. - Extremely precise Background Removal. - Remove objects, text, flaws, or persons from photographs. - Instantly relight your photographs and doodles. - Use AI to teleport anything, anywhere.
• OpenBioML - OpenBioML is a decentralized, collaborative research community that believes open-source machine learning and open science can speed up biotechnology. OpenBioML aspires to be an open, collaborative research laboratory at the crossroads of machine learning and biology. From reviewing the most recent discoveries to collaborating on cutting-edge projects and replicating closed-source research they want to maximize artificial intelligence's positive influence in life sciences.
• DeepFloyd - DeepFloyd is a multimodal AI research lab that is working on the IF text-to-image generation model. The DeepFloyd team is part of the Stability AI project. IF is intended to outperform existing AI models in terms of generating text and captions in photos based on a prompt.
• MedARC - To address the grand issues in medical AI, stability AI is developing unique methodologies and technologies. MedARC's approach is required as a catalyst for advancement. MedARC designs and launches projects in collaboration with other researchers and medical professionals to create large-scale AI models for medicine, build interdisciplinary teams to address clinical needs, share models and datasets, publish results, establish best practices for reproducibility and transparency, build a public community for contributions and collaborations, and guide public policies surrounding AI development in medicine.

‍

This pioneer in diffusion modeling raised $101 million in October 2022 and will primarily be used to serve the custom versions of Stable Diffusion users at a larger scale and invest in more supercomputing power. Mostaque also said in an interview with TechCrunch that this capital will be used to grow the team to about 300 employees from 100. In March 2023, Stability AI also made its first acquisition, Init ML, with its flagship offering, Clipdrop, an image-generative tool that had over 15 million users at the time. Together these companies aligned on a common goal - democratic AI affecting lives.

‍

https://twitter.com/StabilityAI/status/1582053531686645760

‍

Stability.Ai is a very internal community, with voting rights offered to only employees, no billionaires, funds, governments, or anyone else. To ensure that they stay as an open source model and do not cede to outside controls, Stability AI intends to generate revenue by training "private" models for customers and serving as a broad infrastructure layer. It also provides a platform and API, DreamStudio, through which individual users can access its models — According to Mostaque, DreamStudio has over 1.5 million users who have created over 170 million images, and Stable Diffusion has over 10 million daily users "across all channels."

‍

What’s next for Stability AI?

It was in 2021 that stability.ai first launched stable diffusion, which is a prompt-based text-to-image generator that generates images on 5 gigabytes of VRAM as compared to the 40 or 20 gigabytes required by other models. It has grown so much that on 19/05/2023, stable diffusion announced that they are releasing a text-animation AI software called Stable Animation SDK in partnership with Krikey.ai.

‍

SDK will allow artists and developers to use their creations from stable diffusion to create intricate and stunning animations. SDK is compatible with all versions of stable diffusion and provides users with three models of generating animations:

• Users enter a text prompt (similar to Stable Diffusion) and adjust several parameters to generate an animation.
• Text input combined with first picture input: Users give an initial image that serves as the beginning point for their animation. The final output animation is created by combining a text prompt with an image.
• Input video + text input: Users supply an initial video on which to base their animation. They arrive at a final output animation that is also guided by a text prompt by altering various parameters.

‍

Community at Stability.ai

Stability.ai has a thriving and active community on Discord and they recently celebrated over 200K+ members in the community.

‍

‍

‍

Their whole discord page is divided into various sub-categories to make for easier navigation, from a start here section to information, events, co-creators, and even one for the new SDXL beta box and Stable Diffusion.

‍

The stable diffusion thread is extremely active with members coming together at various times to discuss their projects and to help others who are stuck or cannot figure out a particular feature.

‍

‍

They even have a community projects thread where members can share their projects using APIs, integrations, and various other apps. They even have a dreamers thread that artists can use to showcase their work with the community and receive valuable feedback.

‍

‍

‍

Stable Digest

Stable Digest is a bi-monthly community newsletter published by the Stable Foundation. Features interviews with major figures in the AI industry, community spotlights, and exclusive updates from across the Stable communities and beyond.

‍

Stable Digest is your essential companion in a lightning-fast world of constant AI innovation, shining a light on you, our community. It’s AI art & culture news by the community, for the community.

‍

This newsletter also features updates from the Discord community such as showcase winners and highlighted artwork to shine the spotlight on some of the more innovative creators in the community in the community.

‍

‍

‍

AI art & culture by the community, for the community

With modules like Dreamstudio and Stablestudio, Stability AI is leveraging its heavy funding round to provide a whole suite of AI tools for everything from image generation and music to editing and video creation. Emad has stayed true to himself and what he started and wants to continue democratizing AI for the masses. A community of creators, innovators, enthusiasts, freelancers - anyone is able to expand their boundaries with AI.

‍