Generating an image

A generation starts with the character's trigger word followed by a description of the scene. The character itself is never described.

The trigger word

Each character has one trigger word. Set it in the dashboard's Characters tab under Trigger: (the same field is Command Trigger in the Edit Bot modal). Changes apply immediately.

The trigger is usually the character's name — /spongebob — but any word works, including /create or /generate.

A generation in Telegram

/spongebob running through Times Square at night time

Describe the scene, not the character

The model already knows the character. Describe only what is around it.

Slash Required For Generation

By default the leading slash is optional: spongebob at the beach generates, and so does /spongebob at the beach.

Turn on Slash Required For Generation in Settings → Features (Bot Utilities) and the bot only responds to triggers that start with /. In /menu the same switch is ⌨️ Require Forward Slash.

The switch

SettingValue
DefaultOff
CoversGenerations and edits, including reply-edits, photo captions and album captions.

Warning

For generations the switch is per chat. For edits it applies to the whole bot.

The first generation

From finished training to a verified bot

  1. Wait for the dashboard to mark training complete.
  2. Open a DM with the bot, or an authorized group.
  3. Send the trigger word with a short prompt.
  4. Check that the character is correct before opening the bot to the group.

If the character is wrong, submit a revision before launch. Monthly plans include free revision rounds; the Day Pass does not.

Where the image goes

Every generated image is saved to the character's gallery site.

Turning off the Image Captions switch in Bot Utilities removes the caption and the View On Website button from delivered images. The image still goes to the gallery site.