Stable Diffusion 3 HANDS ON! How Good Is It Really?
TLDRStability AI has recently released Stable Diffusion 3 and Stable Diffusion 3 Turbo, accessible exclusively via API through a partnership with Fireworks AI, an API platform offering hosting and fast access to such models. With a focus on open generative AI, Stability AI plans to make the model weights available for self-hosting to members soon. The API pricing is relatively high, with costs around $10 per thousand credits, making Stable Diffusion 3 about 32 times more expensive per image generation compared to Stable Diffusion XL 1.0. Despite this, the reviewer managed to set up Stable Diffusion 3 beta on Pixel Doo within 3 hours, allowing users to generate images with optional negative prompts and choosing between the two versions of the model. The reviewer tested various prompts from press releases and found the image quality to be consistent with the examples on Stability AI's website, indicating that the images displayed were not overly cherry-picked. The model's prompt adherence was impressive, and the reviewer suggests that the quality of Stable Diffusion 3 lives up to its hype. Users interested in trying it out can do so with a Pro Plan on Pixel Doo, which starts at $9.95 per month for unlimited generations and access to other features.
Takeaways
- π Stable Diffusion 3 and Stable Diffusion 3 Turbo have been released by Stability AI and are available via API.
- π€ Stability AI has partnered with Fireworks AI, an API platform for hosting and providing fast access to models like Stable Diffusion.
- π Model weights for self-hosting will be made available with a Stability AI membership in the near future.
- π» The user managed to get Stable Diffusion 3 beta up and running on Pixel Doo within 3 hours.
- π° The pricing for the API is relatively high, costing about $10 per thousand credits.
- π’ Generating an image with Stable Diffusion 3 costs 6 to 12 credits per image, making it 32 times more expensive than Stable Diffusion XL 1.0.
- π A Pro Plan subscription starting at $9.95 per month offers unlimited image generation on Pixel Doo.
- π· The quality of images generated by Stable Diffusion 3 seems to be consistent with those displayed on the website, suggesting less cherry-picking.
- π Text coherence in images generated by Stable Diffusion 3 has room for improvement, as seen with the examples provided.
- π The Turbo model is faster but results in lower quality images compared to the standard model.
- π¨ Prompt adherence for positive prompts is quite good, potentially reducing the need for negative prompts.
- π± For hands-on experience and to generate unlimited images, a Pro membership on Pixel Doo is required.
Q & A
What is Stable Diffusion 3 and how was it recently made available?
-Stable Diffusion 3 is an AI model developed by Stability AI. It was recently released and made available via an API in partnership with Fireworks AI, an API platform that provides hosting and fast access to AI models like Stable Diffusion.
What is the significance of the API pricing for Stable Diffusion 3?
-The API pricing for Stable Diffusion 3 is relatively high, costing about $10 per thousand credits. Each image generated with Stable Diffusion 3 requires 6 to 12 credits, making it approximately 32 times more expensive to generate an image compared to Stable Diffusion XL 1.0.
How quickly can Stable Diffusion 3 generate an image?
-Stable Diffusion 3 is quite fast in generating images. The model can produce an image in a short amount of time, as demonstrated in the script where the image generation process was completed while the speaker was still talking.
What is the difference between Stable Diffusion 3 and Stable Diffusion 3 Turbo?
-Stable Diffusion 3 Turbo is a faster but lower quality version of the Stable Diffusion 3 model. While it generates images more quickly, the resolution and detail of the images produced by the Turbo model are not as high as those from the standard model.
How does Stable Diffusion 3 handle text in images?
-Stable Diffusion 3 has shown improvements in handling text within images. However, it still struggles with text coherence at times, as evidenced by the examples provided in the script where the text in the generated images did not always match the prompt perfectly.
What is the process for generating an image with Stable Diffusion 3?
-To generate an image with Stable Diffusion 3, a user needs to provide a prompt. Optionally, a negative prompt can also be provided. The user can then choose between Stable Diffusion 3 and Stable Diffusion 3 Turbo, and the image is generated based on these inputs.
How does one access Stable Diffusion 3 for image generation?
-To access Stable Diffusion 3, one needs to sign up for a Pro Plan on Pixel Dojo, which starts at $9.95 a month. This plan offers unlimited image generations and access to other features such as a creative upscaler and other Stable Diffusion models.
What is the commitment of Stability AI regarding the model weights of Stable Diffusion 3?
-Stability AI has committed to making the model weights of Stable Diffusion 3 available for self-hosting to those with a Stability AI membership in the near future, aligning with their commitment to open generative AI.
What was the reviewer's overall impression of the image quality generated by Stable Diffusion 3?
-The reviewer found that Stable Diffusion 3 mostly lived up to the hype, with the images generated being of high quality and not too far off from the examples displayed on the website. The prompt adherence was considered good, and the reviewer suggests that negative prompts might not be necessary due to the quality of the positive prompt results.
How does the reviewer suggest one can experiment with the model?
-The reviewer suggests playing around with different prompts and possibly using negative prompts to refine the image generation process and achieve better results.
What are some of the unique image prompts tested by the reviewer?
-The reviewer tested a variety of unique prompts, including an anthropomorphic tortoise on a subway, a man with a retro TV for a head in the desert, a cardboard box with a specific phrase, an alien spaceship shaped like a pretzel, a kangaroo holding a beer with ski goggles singing, and a cheeseburger on a throne-like toilet in a royal chamber.
What is the reviewer's final verdict on Stable Diffusion 3 Turbo?
-The reviewer found that while Stable Diffusion 3 Turbo is quicker, the quality of the images generated is lower and more cartoonish, suggesting that for better quality, the standard model of Stable Diffusion 3 is preferable.
Outlines
π Stable Diffusion 3 and Turbo Release with API Availability
Stability AI has released Stable Diffusion 3 and its Turbo version, but with a catch - they are only accessible via API. They've partnered with Fireworks AI, an API platform that offers hosting and swift access to models like Stable Diffusion. The company has pledged to make the model weights available for self-hosting to members soon. The API has a relatively high cost, with credits needed for usage, and generating an image with Stable Diffusion 3 is significantly more expensive than with Stable Diffusion XL 1.0. The speaker quickly implemented Stable Diffusion 3 beta on Pixel Doo, allowing users to generate images with various options and examples provided. The quality of the generated images is put to the test against prompts from press releases to ensure they are not cherry-picked.
π¨ Testing Image Quality and Prompt Adherence of Stable Diffusion 3
The video script discusses the quality and prompt adherence of Stable Diffusion 3 and its Turbo model. The speaker shares initial results of image generation using various prompts, noting that the standard model generally produces higher quality images than the Turbo model, which appears more cartoonish and of lower resolution. The speaker also emphasizes the importance of prompt adherence, especially in generating images with text, which has been a challenge for AI generators. Despite some text generation inconsistencies, the overall performance of Stable Diffusion 3 is considered to be up to the hype, with good prompt adherence and image quality. The speaker suggests that with the improved performance, negative prompts might be less necessary. The audience is encouraged to try out the models on Pixel Doo, which requires a Pro membership for unlimited image generations.
Mindmap
Keywords
π‘Stable Diffusion 3
π‘API
π‘Fireworks AI
π‘Model Weights
π‘Pixel Doo
π‘Prompt
π‘Negative Prompt
π‘Credits
π‘
π‘Pro Plan
π‘Text Coherence
π‘Cherry-Picking
Highlights
Stability AI has released Stable Diffusion 3 and Stable Diffusion 3 Turbo, but only available via API.
Partnership with Fireworks AI for hosting and fast access.
Commitment to open generative AI, with model weights to be available for self-hosting with a Stability AI membership.
Stable Diffusion 3 beta was operational on Pixel Doo within 3 hours.
API pricing is high at about $10 per thousand credits.
Generating an image with Stable Diffusion 3 is 32 times more expensive than Stable Diffusion XL 1.0.
Paid Pro Plan starts at $9.95 per month for unlimited usage of Pixel Dojo.
Prompt adherence for Stable Diffusion 3 seems very accurate.
Text coherence in images generated by Stable Diffusion 3 is notably improved compared to previous versions.
Stable Diffusion 3 Turbo model generates images more quickly but with lower quality.
The generated images from Stable Diffusion 3 do not appear to be cherry-picked.
Negative prompts were not used in the tests, but could be an area for further exploration.
Stable Diffusion 3 Turbo model is faster but still struggles with text accuracy.
Pixel Doo offers a Pro membership for $9.95 a month, including unlimited generations and access to all Stable Diffusion models.
Stable Diffusion 3 generally lives up to the hype with quality outputs.
The reviewer suggests that the positive prompt adherence is so good that negative prompts may not be necessary.
The reviewer will continue to add more features to Pixel Doo as time progresses.
Viewers are encouraged to comment with their thoughts on Stable Diffusion 3 and Stable Diffusion 3 Turbo.