Stability has made significant progress in the last ten months since the release of its Stability Diffusion Model. Since then, the model has undergone several changes and improvements, resulting in a highly usable and practical product. The recent launch of OpenAI’s GPT model has led to a rise in AI-based apps and services, and Stability AI is no exception. In fact, the company has recently announced the release of the upgraded SDXL 0.9 Model, which promises to make the existing products even better.
The revised model has enabled the company to scale and develop new architectures that address the significant hands issue. Previously, the model struggled to handle photorealistic hand images, but with the recent update, it can now handle them with increasing accuracy and precision. To achieve this, the company utilized a refined dataset of high-quality images that outperformed OpenAI’s Clip Image-to-Text model.
Stability AI’s commitment to constant improvement and innovation is evident in its recent achievements. With the release of the SDXL 0.9 Model, the company is poised to make even more significant strides in the field of AI-based apps and services, and we cannot wait to see what they accomplish next.
Stability AI’s Stable Doodle
Stability AI recently released an improved model called Stable Doodle. Whether you’re an artist, designer, illustrator, student, or just someone who loves to doodle, this sketch-to-image tool can help you turn your rough sketches into pleasing images using the advanced Stable Diffusion model.
Stable Doodle uses the Stable Diffusion XL model, which boasts 77 million parameters, and a powerful T2I adapter to generate images with prompts. While it recognizes its inherent limitations, the algorithm analyses the outline of an image to generate images. Users can also enter prompts in a way familiar to those who use image-generation models like DALL-E, Playgrounds.ai, Midjounry, or others.
How to Use Stability AI’s Stable Doodle
Stable Doodle is available on Web-based app. Android and iOS.
- To use it, open the app and draw anything you like.
- Next, draw your doodle on the canvas as prompted by Stability AI.
- Enter the style you want the image to have in the text box, then click “Generate” to start generating the image.
- Remember, the better your doodle, the better the image generated.
- Once you have found the image you want, click on it to download.
- If you find the image quality low, use the dropdown menu in the upper navigation and select “Upscale” to improve it.
Experiment with different doodles and use a pen pad or stylus to improve your drawing skills. You will achieve fine-tuned control for high-quality images.
Please note that this app is still in beta and may have some biases and inaccuracies. Additionally, users have a daily limit and can choose from 14 different styles.
Stability AI’s Clipdrop Tools
Stability AI has recently launched a range of tools, including Stable Doodle, Image Upscaler, Relight, Remove Background, Reimage XL, Text Remover, and Uncrop. These tools offer a range of features, such as:
- Cleanup: Remove unwanted objects and text from your images.
- Upscale: Enhance the quality of your images by removing compression, reducing noise, and sharpening.
- Relight: Add a natural glow to your images.
- Remove Background: Remove the background from your images with precision, even if the edges are sharp or the image is complex.
- Reimage XL: Change the surroundings of your images and create different variants of the same image.
- Stable Diffusion XL: Create text from images. Enter your prompt and choose the style you prefer.
- Text Remover is focused on removing text from images with precision.
- Uncrop: Expand the size of your images.
- Dreamstudio.ai: This is also released by Stability AI which is a combination of all these models to generate and edit the images.
Stability AI expects commercial success with these revolutionary tools. Last month, the company raised $25 million through convertible notes, bringing its total funding to $125 million and valuing it at $1 billion.
Risks and Benefits
Stability has introduced digital watermarking, but the algorithm still has some biases in the images generated from the trained data. Additionally, Stability is no longer taking data directly from the internet; instead, they refine the data before training the model. Stability is expected to be released at the next level. As an open-source model, anyone can contribute and use the project to leverage the Stability AI research project.
The company still needs to release some of its models. Stability AI is still working on invisible watermarking and Authenticity.org with lawmakers on national datasets that reflect countries’ cultures and other related issues. As it emerges so rapidly, Stability is looking for standards to use appropriately.
There are several models to try, such as Cliprop’s Stable Doodle, Uncrop, and Reimagine XL. To integrate these into your apps and services, please consider checking out the Stable Animation SDK, API, and other resources available in the Stability AI documentation.