Character Commentator

When it reshapes the image for an emotion, that emotion is noted for you to know.

There is now a button, compress avatars. I suggest clicking it to scrunch big pngs down to much smaller webps.
Bridge also updated!

An option, starting false(so head into the settings if you want it).
When on, whenever someone comments and they have an image, it will figure out what emotion they're expressing and then make a new image based on the old image with the new emotion in play.
This does cost a little anlas each time, hence it being a setting.
It does not save the emotion images created.

One, way better at working with lorebook names that don't match character names.
Two, bigger, huge! You can now add a 'style', there's a button at the bottom. This bit of text is added to all images made for this specific story. This lets you have more control, and thus more coherence between avatars. Note, this is V5, so sentences are alrighty.

An error toast to help if it appears a mistake was made that needs fixing.

Should stop trying to re-do characters with a space in their name.

Major update, now if it sees an avatar without an image, it tries to make one. A bridge script is needed(violent/tampermonkey) for this to work, but it has no friction. Enjoy pictures with no clicks. Unless you want to retry.

Small one, focus character is now per-story instead of global, so you can have different focuses per story.

There’s now a button next to the avatar button in the action row.
If the current character is not locked, it shows Only This Character.
If that character is already locked as the sole speaker, it shows Random Speakers.

2.6.5 -- It's much sharper about keeping itself to dialogue only now, and not trying to sneak in actions.

2.5.8 -- This one's a big one. Sit down. No, all the way. Okay. Remember those Data URIs? Confusing stuff. Forget them. Toss it right out. Any time someone speaks and they do NOT have an image, a set image button will be present. Give it an image, and off into the lorebook it goes. Someone WITH an image speaks? Clear image is right there. No DataURIs to even think about. Enjoy!

Attempt to avoid actions and keep it to spoken words. Not 100%, can't be.

Minor fix to make sure no sentence fragments are left at the end.

You can now enter a model, so it can speak with the new xialong! Seems to work reasonably well. Try it out!

Tightened up the vertical spacing to fit more in the window. Looks better, to me. Hopefully it does to you too.

There is now an X to click to close the hypebot window. We're done with you, go away.

New option! Turn it on and it'll lean towards people who would be thinking harder about what's going on in the story instead of just being a random shot of them all.

Characters can now use *s in their replies and it will italicize and/or bold them appropriately without you thinking about them.

Give each character equal chance to appear.

New buttons to demand who speaks next if you prefer. Option added to go with it!

Fixed checking categories! Now that actually works, oops. But also yay!

You can now define a category(folder) and it will assume all entries in that category are characters, without looking for Type: Character in them. This is an option in the extension.

Oops, now they'll remember their first comment, the one you started chatting about. It'll be right there, in the chat.

Welcome to a whole new world! A character comments, you can click chat and gab at them about what they just said. Prod, or just be silly with them. Your characters, come to life.

A big one! Now you can make a lorebook, by default Avatars, but there's config to change that. Within it, each line should be 'person name: URI data of an image' Then, when that person is hyped, that image will display. Simple as that. I strongly mega-encourage you don't let that lorebook ever become active, or you'll waste a lot of tokens. Since I'm not aware of any lorebook limits, you can shove in pictures of everyone. I recommend 256x256 webps though.

It will be more eager to get bigger in its responses, still uses the maximum.

Longer outputs are possible, and it tries to fill the max tokens you set.

It gets less confused about multi-word character names and will check for both first and full name mentions of the character. Console debug tuned down. If, while generating, it decides it should not generate, it will not hide the bot if visible. It just won't update it. No jumpscares as you're working.

It scans a bit harder. Still doesn't see headers properly(oops), but is a tiny bit better at spotting lorebooks. Having Type: character actually in the lorebook will help a lot.

Remove buttons, fix closing of bot sometimes.

Initial release: Avatar image for all character comments (leave empty for default HypeBot avatar). Must be a base64 data URI starting with 'data:image/' - convert images at: https://www.base64-image.de/