Merge pull request #2655 from SillyTavern/image-swipes Image Generation: Add swipes for generated images

e4d045f03ecfc96323b800e8594907b111dc75c9

Cohee <18619528+Cohee1207@users.noreply.github.com>

Signed
6 files changed, +187 -10Ignore whitespace
public/index.html+5 -0
@@ -5802,6 +5802,11 @@
58025802 <div title="Caption" class="right_menu_button fa-lg fa-solid fa-envelope-open-text mes_img_caption" data-i18n="[title]Caption"></div>
58035803 <div title="Delete" class="right_menu_button fa-lg fa-solid fa-trash-can mes_img_delete" data-i18n="[title]Delete"></div>
58045804 </div>
5805+ <div class="mes_img_swipes">
5806+ <div title="Swipe left" class="right_menu_button fa-lg fa-solid fa-chevron-left mes_img_swipe_left" data-i18n="[title]Swipe left"></div>
5807+ <div class="mes_img_swipe_counter">1/1</div>
5808+ <div title="Swipe right" class="right_menu_button fa-lg fa-solid fa-chevron-right mes_img_swipe_right" data-i18n="[title]Swipe right"></div>
5809+ </div>
58055810 <img class="mes_img" src="" />
58065811 </div>
58075812 <div class="mes_bias"></div>
public/script.js+21 -1
@@ -459,6 +459,7 @@ export const event_types = {
459459 LLM_FUNCTION_TOOL_REGISTER: 'llm_function_tool_register',
460460 LLM_FUNCTION_TOOL_CALL: 'llm_function_tool_call',
461461 ONLINE_STATUS_CHANGED: 'online_status_changed',
462+ IMAGE_SWIPED: 'image_swiped',
462463};
463464
464465export const eventSource = new EventEmitter();
@@ -2112,6 +2113,7 @@ export function updateMessageBlock(messageId, message) {
21122113export function appendMediaToMessage(mes, messageElement, adjustScroll = true) {
21132114 // Add image to message
21142115 if (mes.extra?.image) {
2116+ const container = messageElement.find('.mes_img_container');
21152117 const chatHeight = $('#chat').prop('scrollHeight');
21162118 const image = messageElement.find('.mes_img');
21172119 const text = messageElement.find('.mes_text');
@@ -2127,9 +2129,27 @@ export function appendMediaToMessage(mes, messageElement, adjustScroll = true) {
21272129 });
21282130 image.attr('src', mes.extra?.image);
21292131 image.attr('title', mes.extra?.title || mes.title || '');
21302132 messageElement.find('.mes_img_container')container.addClass('img_extra');
21312133 image.toggleClass('img_inline', isInline);
21322134 text.toggleClass('displayNone', !isInline);
2135+
2136+ const imageSwipes = mes.extra.image_swipes;
2137+ if (Array.isArray(imageSwipes) && imageSwipes.length > 0) {
2138+ container.addClass('img_swipes');
2139+ const counter = container.find('.mes_img_swipe_counter');
2140+ const currentImage = imageSwipes.indexOf(mes.extra.image) + 1;
2141+ counter.text(`${currentImage}/${imageSwipes.length}`);
2142+
2143+ const swipeLeft = container.find('.mes_img_swipe_left');
2144+ swipeLeft.off('click').on('click', function () {
2145+ eventSource.emit(event_types.IMAGE_SWIPED, { message: mes, element: messageElement, direction: 'left' });
2146+ });
2147+
2148+ const swipeRight = container.find('.mes_img_swipe_right');
2149+ swipeRight.off('click').on('click', function () {
2150+ eventSource.emit(event_types.IMAGE_SWIPED, { message: mes, element: messageElement, direction: 'right' });
2151+ });
2152+ }
21332153 }
21342154
21352155 // Add file to message
public/scripts/extensions/stable-diffusion/index.js+117 -5
@@ -59,6 +59,7 @@ const initiators = {
5959 action: 'action',
6060 interactive: 'interactive',
6161 wand: 'wand',
62+ swipe: 'swipe',
6263};
6364
6465const generationMode = {
@@ -2275,9 +2276,9 @@ async function generatePicture(initiator, args, trigger, message, callback) {
22752276 const quietPrompt = getQuietPrompt(generationType, trigger);
22762277 const context = getContext();
22772278
2278- // if context.characterId is not null, then we get context.characters[context.characterId].avatar, else we get groupId and context.groups[groupId].id
2279+ const characterName = context.groupId
2279- // sadly, groups is not an array, but is a dict with keys being index numbers, so we have to filter it
2280+ ? context.groups[Object.keys(context.groups).filter(x => context.groups[x].id === context.groupId)[0]]?.id?.toString()
2280- const characterName = context.characterId ? context.characters[context.characterId].name : context.groups[Object.keys(context.groups).filter(x => context.groups[x].id === context.groupId)[0]]?.id?.toString();
2281+ : context.characters[context.characterId]?.name;
22812282
22822283 if (generationType == generationMode.BACKGROUND) {
22832284 const callbackOriginal = callback;
@@ -3393,6 +3394,7 @@ async function sendMessage(prompt, image, generationType, additionalNegativePref
33933394 generationType: generationType,
33943395 negative: additionalNegativePrefix,
33953396 inline_image: false,
3397+ image_swipes: [image],
33963398 },
33973399 };
33983400 context.chat.push(message);
@@ -3535,7 +3537,9 @@ async function sdMessageButton(e) {
35353537 const $mes = $icon.closest('.mes');
35363538 const message_id = $mes.attr('mesid');
35373539 const message = context.chat[message_id];
3538- const characterFileName = context.characterId ? context.characters[context.characterId].name : context.groups[Object.keys(context.groups).filter(x => context.groups[x].id === context.groupId)[0]]?.id?.toString();
3540+ const characterFileName = context.groupId
3541+ ? context.groups[Object.keys(context.groups).filter(x => context.groups[x].id === context.groupId)[0]]?.id?.toString()
3542+ : context.characters[context.characterId]?.name;
35393543 const messageText = message?.mes;
35403544 const hasSavedImage = message?.extra?.image && message?.extra?.title;
35413545 const hasSavedNegative = message?.extra?.negative;
@@ -3579,10 +3583,23 @@ async function sdMessageButton(e) {
35793583
35803584 function saveGeneratedImage(prompt, image, generationType, negative) {
35813585 // Some message sources may not create the extra object
35823586 if (typeof message.extra !== 'object' || message.extra === null) {
35833587 message.extra = {};
35843588 }
35853589
3590+ // Add image to the swipe list if it's not already there
3591+ if (!Array.isArray(message.extra.image_swipes)) {
3592+ message.extra.image_swipes = [];
3593+ }
3594+
3595+ const swipes = message.extra.image_swipes;
3596+
3597+ if (message.extra.image && !swipes.includes(message.extra.image)) {
3598+ swipes.push(message.extra.image);
3599+ }
3600+
3601+ swipes.push(image);
3602+
35863603 // If already contains an image and it's not inline - leave it as is
35873604 message.extra.inline_image = message.extra.image && !message.extra.inline_image ? false : true;
35883605 message.extra.image = image;
@@ -3621,6 +3638,99 @@ async function writePromptFields(characterId) {
36213638 await writeExtensionField(characterId, 'sd_character_prompt', promptObject);
36223639}
36233640
3641+/**
3642+ * Switches an image to the next or previous one in the swipe list.
3643+ * @param {object} args Event arguments
3644+ * @param {any} args.message Message object
3645+ * @param {JQuery<HTMLElement>} args.element Message element
3646+ * @param {string} args.direction Swipe direction
3647+ * @returns {Promise<void>}
3648+ */
3649+async function onImageSwiped({ message, element, direction }) {
3650+ const context = getContext();
3651+ const animationClass = 'fa-fade';
3652+ const messageImg = element.find('.mes_img');
3653+
3654+ // Current image is already animating
3655+ if (messageImg.hasClass(animationClass)) {
3656+ return;
3657+ }
3658+
3659+ const swipes = message?.extra?.image_swipes;
3660+
3661+ if (!Array.isArray(swipes)) {
3662+ console.warn('No image swipes found in the message');
3663+ return;
3664+ }
3665+
3666+ const currentIndex = swipes.indexOf(message.extra.image);
3667+
3668+ if (currentIndex === -1) {
3669+ console.warn('Current image not found in the swipes');
3670+ return;
3671+ }
3672+
3673+ // Switch to previous image or wrap around if at the beginning
3674+ if (direction === 'left') {
3675+ const newIndex = currentIndex === 0 ? swipes.length - 1 : currentIndex - 1;
3676+ message.extra.image = swipes[newIndex];
3677+
3678+ // Update the image in the message
3679+ appendMediaToMessage(message, element, false);
3680+ }
3681+
3682+ // Switch to next image or generate a new one if at the end
3683+ if (direction === 'right') {
3684+ const newIndex = currentIndex === swipes.length - 1 ? swipes.length : currentIndex + 1;
3685+
3686+ if (newIndex === swipes.length) {
3687+ const abortController = new AbortController();
3688+ const swipeControls = element.find('.mes_img_swipes');
3689+ const stopButton = document.getElementById('sd_stop_gen');
3690+ const stopListener = () => abortController.abort('Aborted by user');
3691+ const generationType = message?.extra?.generationType ?? generationMode.FREE;
3692+ const dimensions = setTypeSpecificDimensions(generationType);
3693+ const originalSeed = extension_settings.sd.seed;
3694+ extension_settings.sd.seed = Math.round(Math.random() * Number.MAX_SAFE_INTEGER);
3695+ let imagePath = '';
3696+
3697+ try {
3698+ $(stopButton).show();
3699+ eventSource.once(CUSTOM_STOP_EVENT, stopListener);
3700+ const callback = () => { };
3701+ const hasNegative = message.extra.negative;
3702+ const prompt = await refinePrompt(message.extra.title, false, false);
3703+ const negativePromptPrefix = hasNegative ? await refinePrompt(message.extra.negative, false, true) : '';
3704+ const characterName = context.groupId
3705+ ? context.groups[Object.keys(context.groups).filter(x => context.groups[x].id === context.groupId)[0]]?.id?.toString()
3706+ : context.characters[context.characterId]?.name;
3707+
3708+ messageImg.addClass(animationClass);
3709+ swipeControls.hide();
3710+ imagePath = await sendGenerationRequest(generationType, prompt, negativePromptPrefix, characterName, callback, initiators.swipe, abortController.signal);
3711+ } finally {
3712+ $(stopButton).hide();
3713+ messageImg.removeClass(animationClass);
3714+ swipeControls.show();
3715+ eventSource.removeListener(CUSTOM_STOP_EVENT, stopListener);
3716+ restoreOriginalDimensions(dimensions);
3717+ extension_settings.sd.seed = originalSeed;
3718+ }
3719+
3720+ if (!imagePath) {
3721+ return;
3722+ }
3723+
3724+ swipes.push(imagePath);
3725+ }
3726+
3727+ message.extra.image = swipes[newIndex];
3728+ appendMediaToMessage(message, element, false);
3729+ }
3730+
3731+ await context.saveChat();
3732+}
3733+
36243734jQuery(async () => {
36253735 await addSDGenButtons();
36263736
@@ -3759,6 +3869,8 @@ jQuery(async () => {
37593869 }
37603870 });
37613871
3872+ eventSource.on(event_types.IMAGE_SWIPED, onImageSwiped);
3873+
37623874 eventSource.on(event_types.CHAT_CHANGED, onChatChanged);
37633875
37643876 await loadSettings();
public/style.css+23 -0
@@ -4569,6 +4569,7 @@ a {
45694569 image-rendering: -webkit-optimize-contrast;
45704570}
45714571
4572+.mes_img_swipes,
45724573.mes_img_controls {
45734574 position: absolute;
45744575 top: 0.1em;
@@ -4578,9 +4579,16 @@ a {
45784579 opacity: 0;
45794580 flex-direction: row;
45804581 justify-content: space-between;
4582+ align-items: center;
45814583 padding: 1em;
45824584}
45834585
4586+.mes_img_swipes {
4587+ top: unset;
4588+ bottom: 0.1rem;
4589+}
4590+
4591+.mes_img_swipes .right_menu_button,
45844592.mes_img_controls .right_menu_button {
45854593 filter: brightness(90%);
45864594 text-shadow: 1px 1px var(--SmartThemeShadowColor) !important;
@@ -4589,16 +4597,20 @@ a {
45894597 width: 1.25em;
45904598}
45914599
4600+.mes_img_swipes .right_menu_button::before,
45924601.mes_img_controls .right_menu_button::before {
45934602 /* Fix weird alignment with this font-awesome icons on focus */
45944603 position: relative;
45954604 top: 0.6125em;
45964605}
45974606
4607+.mes_img_swipes .right_menu_button:hover,
45984608.mes_img_controls .right_menu_button:hover {
45994609 filter: brightness(150%);
46004610}
46014611
4612+.mes_img_container:hover .mes_img_swipes,
4613+.mes_img_container:focus-within .mes_img_swipes,
46024614.mes_img_container:hover .mes_img_controls,
46034615.mes_img_container:focus-within .mes_img_controls {
46044616 opacity: 1;
@@ -4612,6 +4624,17 @@ body:not(.caption) .mes_img_caption {
46124624 display: none;
46134625}
46144626
4627+.mes_img_container:not(.img_swipes) .mes_img_swipes,
4628+body:not(.sd) .mes_img_swipes {
4629+ display: none;
4630+}
4631+
4632+.mes_img_swipe_counter {
4633+ font-weight: 600;
4634+ filter: drop-shadow(2px 4px 6px black);
4635+ cursor: default;
4636+}
4637+
46154638.img_enlarged_holder {
46164639 /* Scaling via flex-grow and object-fit only works if we have some kind of base-height set */
46174640 min-height: 120px;
src/endpoints/images.js+1 -1
@@ -82,7 +82,7 @@ router.post('/list/:folder', (request, response) => {
8282 }
8383
8484 try {
8585 const images = getImages(directoryPath, 'date');
8686 return response.send(images);
8787 } catch (error) {
8888 console.error(error);
src/util.js+20 -3
@@ -382,14 +382,31 @@ function removeOldBackups(directory, prefix) {
382382 }
383383}
384384
385-function getImages(path) {
385+/**
386+ * Get a list of images in a directory.
387+ * @param {string} directoryPath Path to the directory containing the images
388+ * @param {'name' | 'date'} sortBy Sort images by name or date
389+ * @returns {string[]} List of image file names
390+ */
391+function getImages(directoryPath, sortBy = 'name') {
392+ function getSortFunction() {
393+ switch (sortBy) {
394+ case 'name':
395+ return Intl.Collator().compare;
396+ case 'date':
397+ return (a, b) => fs.statSync(path.join(directoryPath, a)).mtimeMs - fs.statSync(path.join(directoryPath, b)).mtimeMs;
398+ default:
399+ return (_a, _b) => 0;
400+ }
401+ }
402+
386403 return fs
387404 .readdirSync(pathdirectoryPath)
388405 .filter(file => {
389406 const type = mime.lookup(file);
390407 return type && type.startsWith('image/');
391408 })
392409 .sort(Intl.CollatorgetSortFunction().compare);
393410}
394411
395412/**