Skip to content
TomoriBot

Image Generation

TomoriBot can generate images from a text prompt or by editing a reference image. Use /generate image, or just ask her (“draw me a red panda drinking coffee”).

What She Can Do

Section titled “What She Can Do”
  • Text-to-image — generate from a prompt.
  • Image-to-image — edit or restyle a whole reference image.
  • Inpainting — redraw a specific region while preserving the rest.
  • Outpainting — extend the canvas beyond the original frame.
  • Customizable aspect ratios.
  • Reference images can come from message attachments, stickers, emojis, or user/persona avatars. Point her at a message, or name a user/persona to pull in their avatar as a reference.

Which editing modes are available depends on the backend. Text-to-image and image-to-image work on the built-in cloud providers (Google, Vertex, OpenRouter), while inpainting and outpainting are provided by local ComfyUI custom endpoints and are gated by that endpoint’s declared capabilities. Whatever the backend can’t do is simply hidden from her, so she won’t offer a mode your setup doesn’t support.

When she generates an image, she uses your persona’s Physical Appearance context plus default positive and negative tags (where the backend supports negative prompts). The result is delivered as a Discord media gallery with generation-time details, including any referenced users or personas.

Tag Customization

Section titled “Tag Customization”

Every tag source above is editable, each at a different scope. All of these open a modal pre-filled with the current tags, so you edit in place:

  • /persona image-tags — the selected persona’s Physical Appearance tags (how she looks). Requires the Manage Server permission.
  • /personal image-tags — your own appearance tags, applied when a generation references you. Follows you across every server (see Personalization).
  • /config image-tags default-positive and /config image-tags default-negative — the server-wide default tags added to (or steered away from) every generation. Negative tags only take effect where the backend supports negative prompts. Submitting the modal with an empty box resets that list to the built-in defaults.

Setup

Section titled “Setup”
  1. Configure an image model with /model image.
  2. Make sure image generation is allowed — it’s gated by the imagegen_enabled capability (/capabilities).
  3. Ask her to generate, or run /generate image.

Provider Support

Section titled “Provider Support”

Native image generation is available on Google, Vertex AI, Vertex AI Express, OpenRouter, Z.ai, NVIDIA NIM, and NovelAI (anime-styled; native inpainting is built and coming soon, currently disabled while edge-blending is refined). For the full support matrix and how to add a provider, see Providers & Models.

For local image generation with your own hardware via ComfyUI, see Setup: ComfyUI.