Image Generation

TomoriBot can generate images from a description or edit a reference image. Ask her in chat or use /generate image. For manual NovelAI generation, use /novelai generate image.

Tip

If a chat request does not produce an image, try /generate image. It calls the image model directly, even when the chat model cannot use tools.

Editing and references

TomoriBot can use attachments, stickers, emojis, and user or persona avatars as references. Mention a user or persona to use their avatar, or their saved character reference image when they have one. Depending on your image provider, she can also restyle an image, redraw part of it (inpainting), or extend it beyond its edges (outpainting).

Tip

To edit an existing picture, reply to the message containing it and describe the change. This makes the picture available even if the original message is far up the channel.

Available editing modes depend on the provider. Inpainting and outpainting require a local ComfyUI endpoint that supports them. The nsfw option hides the result behind a Discord spoiler; it does not change provider content filters.

Tag customization

Appearance and default tags can guide image generation. For example, if you ask “Generate an image of me hugging you,” TomoriBot can use your saved appearance and the persona’s appearance to describe both of you in the image.

• /config > Persona > Image Generation Details: the persona’s appearance tags and character reference image. Default Tomori characters come with one. Giving the persona a new avatar removes it, so upload your own reference for the new look.
• /personal config: your appearance tags and character reference image. TomoriBot skips your reference image when your privacy level is Full. See Personalization.
• /config > Models > Image Generation Defaults: server-wide positive and negative tags. Negative tags depend on provider support.

Tag lookup

Danbooru Search helps TomoriBot find a character’s appearance or outfit tags before generating an image, including cosplay requests. The Default + Image Generation Expert system prompt preset encourages her to look up tags and choose the appropriate prompt style.

Setup and providers

Configure an image model under /config > Models > Switch Models and enable image generation under /config > Permissions. Available models depend on your provider. See Providers & Models for the provider list, or Setup: ComfyUI for local generation.