NovelAI Image Generation Intro Tutorial

In this paragraph, NAI-chan introduces the basics of using NovelAI Diffusion for image generation. It starts with a warm greeting and an overview of the Image Editor, where users begin creating anime-inspired art. The example used is generating an image of a 'flower field' and 'sunset,' while ensuring no humans are included in the scene. NAI-chan explains the significance of tags in prompts, emphasizing their role in guiding the AI to create more controlled and consistent results. Tags are separated by commas, with suggestions provided by the AI to make the process easier. This section also mentions a progress bar that helps users keep track of prompt length and introduces the concept of prompt structure, where elements placed earlier have more influence on the final image.

This paragraph expands on how to create more specific images by adjusting the prompt. NAI-chan demonstrates how to generate a girl by adding the tag '1girl' and describes the importance of tag placement in the prompt. Details such as camera angle ('from side'), outfit ('school uniform'), hair ('messy hair, brown hair'), and eye color ('green eyes') are added to create a custom character. The AI responds by generating the image based on the prompt. NAI-chan then introduces the seed feature, allowing users to tweak the prompt without entirely changing the image, providing a finer level of control over their creations.

In this section, NAI-chan explains how to enhance or reduce the influence of certain tags on the image by using curly brackets {} and square brackets [], respectively. Users can strengthen a tag like 'messy hair' by enclosing it in curly brackets or weaken elements like make-up using square brackets. A practical example shows how the AI responds to different strengths of the 'messy hair' tag, illustrating how the hair gets progressively messier as the tag strength increases. This section emphasizes the flexibility users have in shaping the AI's focus through tag manipulation.

Here, NAI-chan discusses how to manage elements in the image that the user didn’t ask for. By using the Undesired Content feature, users can exclude specific objects or details from appearing in the generated image. The example provided shows how adding 'petals' to the Undesired Content list prevents the AI from generating petals. The section also highlights how curly and square brackets can be applied in Undesired Content for more precise control.

This paragraph covers image size adjustments and generating multiple images at once. NAI-chan explains how different aspect ratios can impact the final result depending on the desired composition. Users can select how many images to generate simultaneously, though this will increase the cost in Anlas, NovelAI’s currency. The section also touches on subscription benefits for Opus tier users, who can generate unlimited images within certain parameters. Images generated appear in the History section for easy access, and unwanted images can be removed.

NAI-chan explains the importance of saving images manually, as NovelAI does not store any generated images for privacy reasons. Users are encouraged to download images they like before leaving the page to avoid losing their work. The Download ZIP option allows users to save all generated images at once. This paragraph also highlights that users retain ownership of their generated images and can use them freely. Moreover, users can recover the original prompt and settings by dragging and dropping the saved image back onto the site.

This section introduces users to more advanced settings, including steps, prompt guidance, sampler, and prompt guidance rescale. While the default values are optimized for general use, experimenting with these settings can lead to better results. NAI-chan reassures users that they can always reset the settings to default if something goes wrong during experimentation. The section encourages users to explore the official NovelAI documentation for in-depth tutorials on these settings.

In this final paragraph, NAI-chan wraps up the tutorial by encouraging users to experiment with NovelAI and create their own unique illustrations. The tutorial ends with a farewell message, wishing users good luck with their AI-powered creations and encouraging them to share their results.