Better Than Midjourney: Openjourney Stable Diffusion
TLDRThe video introduces a method to run 'mid-journey' as a local service on a personal computer, enabling unlimited free image generation using an open-source model called 'Open Journey'. The tutorial guides through downloading the model from Hugging Face, installing it with Automatic 1111, and using various prompt sites to generate detailed and stylistic images without the need for a monthly subscription.
Takeaways
- π The speaker introduces an alternative to the mid-journey AI art generation platform, which requires a paid subscription.
- π The alternative allows users to run mid-journey as a local service on their PC, enabling unlimited free image generation.
- π οΈ To set this up, users need to have followed a previous tutorial on setting up Automatic 1111 on their local Windows machine.
- π Hugging Face is highlighted as a valuable resource for finding training datasets and models for stable diffusion, including an open-source mid-journey trained model.
- π The speaker demonstrates how to download and install the open Journey model into the local stable diffusion setup.
- π¨ The results from the local service are expected to be similar to those from mid-journey, but without the need for a subscription or limitations on image creation.
- π‘ The speaker suggests using prompt sites, such as Creative Indie, for mid-journey style prompts to generate specific types of images.
- π Prompt Hero is recommended for its collection of the best open Journey prompts and detailed information on how to achieve specific image results.
- ποΈ The process of refining prompts and details is emphasized as a way to achieve desired image outcomes with stable diffusion.
- πΈ The speaker shares examples of generated images, showcasing the level of detail and style replication possible with the local service.
- π’ The speaker invites viewers to share their generated images and engage in a Discord channel for further discussion and feedback.
Q & A
What is the main topic of the video?
-The main topic of the video is about running mid-journey, an AI-based image generation platform, as a local service on one's own PC to create unlimited free images.
What is mid-journey and what are its limitations?
-Mid-journey is an artificial intelligence stable diffusion website that allows users to connect to Discord and generate artwork. Its limitations include the requirement of a monthly subscription and a limited number of images that can be created based on the subscription plan.
What is the alternative to mid-journey presented in the video?
-The alternative presented is running mid-journey as a local service on one's own PC using a model called open Journey, which is an open-source mid-journey trained model available on hugging face.
How does one set up the local service for mid-journey?
-To set up the local service, one needs to follow a tutorial on setting up automatic 1111 on their local Windows machine, then download the checkpoint file for the open Journey model from hugging face and place it into the models directory inside the stable diffusion install.
What is hugging face and how is it used in this context?
-Hugging face is a website that hosts different training datasets, models, and other resources for AI and machine learning. In this context, it is used to find and download the open Journey model for use with the local mid-journey service.
What are some features of the open Journey model?
-The open Journey model is an open-source mid-journey trained model that can generate images similar to those produced by mid-journey. It allows for unlimited free image generation and can be used with a local version of stable diffusion hosted on one's own PC.
How can one find prompts for the open Journey model?
-One can find prompts by visiting websites like creativeindie and prompthero, which provide a list of mid-journey style prompts that can be used with the open Journey model in automatic 1111.
What are some examples of prompts tested in the video?
-Examples of prompts tested in the video include 'a surreal underwater world in the style of HR Giger', 'a historical epic battle scene in the style of Frank Frazetta', and 'a steampunk airship in the style of Jules Verne'.
What are the steps to refine the generated images?
-To refine the generated images, one can adjust the prompts, use negative prompts to exclude unwanted features, and modify settings like sampling steps, resolution, and the use of upscaling techniques like ESRGAN 4X.
How can users share their results and find more prompts?
-Users can join a Discord channel mentioned in the video to share their results and find more prompts by exploring websites like prompthero, which provides a section on the best open Journey prompts along with details on how they were generated.
What was the outcome of using the open Journey model?
-The outcome of using the open Journey model was the generation of highly detailed and intricate images in various styles, demonstrating the capability of the model to produce artwork similar to that of mid-journey without the need for a subscription.
Outlines
ποΈ Exploring Alternatives to Mid-Journey AI Art Generation
The paragraph discusses an alternative method to using the Mid-Journey AI platform for generating artwork. It highlights the limitations and costs associated with Mid-Journey, such as the need for a monthly subscription and limited image creation based on the plan. The speaker introduces a way to run Mid-Journey as a local service on one's PC, allowing for unlimited free image generation. This is made possible by training data from actual Mid-Journey images, and the results are expected to be similar. The speaker references a tutorial on setting up a local version of the AI model and guides the audience to Hugging Face, a website with various models and training data sets. The Open Journey model, an open-source version of Mid-Journey, is recommended for download and use with the local AI setup.
π Generating Detailed Art with Local AI Models
This paragraph delves into the process of generating detailed and intricate artwork using locally hosted AI models. The speaker shares their experience with different prompts and the resulting images, showcasing the capabilities of the AI in various styles, such as surreal underwater worlds, historical epic battles, and steampunk airships. The use of specific artists' styles in prompts is highlighted to achieve stylized images. The speaker also recommends websites that offer prompt-building guidance and curated lists of prompts for achieving desired results. The paragraph concludes with the speaker's excitement about the potential of this free alternative to Mid-Journey, encouraging the audience to share their own creations.
Mindmap
Keywords
π‘Artificial Intelligence
π‘Mid-Journey
π‘Stable Diffusion
π‘Discord
π‘Hugging Face
π‘Open Journey
π‘Prompts
π‘Upscale
π‘DPM Plus Plus
π‘Community and Collaboration
π‘Prompt Hero
Highlights
The introduction of a method to run mid-journey as a local service on a personal computer, enabling the generation of unlimited free images.
The comparison of mid-journey with an AI stable diffusion website, highlighting the limitations and costs associated with the latter.
The mention of a free trial with limited resources and various subscription plans for the AI stable diffusion website, including a $60/month Pro Plan.
The explanation of how to set up Automatic 1111 on a local Windows machine, which is a prerequisite for running mid-journey locally.
The introduction of Hugging Face, a website with numerous training datasets and models for stable diffusion, including an open source mid-journey trained model.
The simplicity of downloading and using the open Journey model directly with Automatic 1111 for stable diffusion.
The availability of various mid-journey models on Hugging Face, allowing users to experiment with different options.
The creation of a Discord Channel for sharing results and discussing improvements or alternative models.
The use of creativeindie website with a list of 644 mid-journey prompts for generating images with Automatic 1111.
The demonstration of generating images with specific styles, such as the surreal Underwater World in the style of HR Giger.
The detailed description of the process and settings used to generate images, including the use of DPM plus plus 2m sampling method and upscalers.
The presentation of the results, showcasing the intricate details and unique styles produced by the AI model.
The exploration of different prompts, such as historical epic battle scenes in the style of Frank Frazetta and steampunk airships in the style of Jules Verne.
The tip about websites that help with prompt building, such as Prompt Hero, which provides a section on the best open Journey prompts.
The example of using Prompt Hero to generate a highly detailed robot image, illustrating how specific prompts can yield targeted results.
The mixed results obtained from using a prompt for a young South Korean girl, demonstrating the potential need for refining prompts and settings.
The overall conclusion that running mid-journey locally offers a free alternative to the AI stable diffusion website, allowing for unlimited image generation.