v1.63.0 (AI Ways and Skills, Saves, Official Example Workflows, Privacy Mode)

• 画布左上角的立方体入口不再压在 ComfyUI 自己的菜单上面：以前它在最顶层，打开左上角的 C 菜单时会盖住“新建”那一行。现在它在画布上面、顶栏和菜单、对话框下面。README 的安装第 2 步也配上了图，写明装好后入口在哪、看不到时怎么办。

• The cube entry at the canvas's top left no longer sits over ComfyUI's own menus: it was on the very top layer and covered the "New" row of the C menu. It is now over the canvas and under the top bar, menus and dialogs. Step 2 of the README's install has a picture of where the entry is after installing and what to do when it can't be seen.

• 在设置里切换语言后，AI 工作台右边的展台、留言框和画布下方的存档条也马上换成新的语言。以前它们要刷新页面才变。

• After switching the language in Settings, the AI workbench's stage and message box and the saves bar under the canvas change at once too. They used to wait for a page reload.

• 浏览器里同时开着两个 ComfyUI 页面时，AI 干活不再时好时坏：以前两个页面会各抢一半动作，一个页面开了任务的标签页，另一个页面接着去跑却找不到它（"标签页关了"）。现在任务的动作先交给上次干活的那个页面，它关掉了别的页面才接手。

• With ComfyUI open in two browser tabs the AI no longer fails now and then: the two pages took half the actions each, one opened the task's tab and the other then could not find it ("the tab is closed"). Now the page that did the last action gets the next first, and another takes over only when it is gone.

• AI 用 run_workflow 时可以直接写图库里一张图的名字：一步打开这张图的工作流、改好提示词或种子、跑出来；同一张图再跑就接着用那个标签页。打开图的工作流时换一个新的工作流编号，原来那个标签页还开着也不会串；打开后的标记在加载之前就写进去，不会被 ComfyUI 收尾时冲掉（以前偶尔会说"标签页关了"或"没打开"）。

• With run_workflow the AI can name a picture of the gallery: its workflow opens, gets the new prompt or seed and runs, in one call, and a run of the same picture again reuses that tab. A picture's workflow opens with an id of its own, so the tab it ran on, still open, is not mixed up with it, and the tab's mark goes in before it loads, so ComfyUI finishing the tab does not drop it (it said now and then that the tab was closed or did not open).

• 偶尔打开 ComfyUI 时 AMB 整个没出来（刷新一下又好了）的问题修好了。ComfyUI 会同时加载插件的全部 js 文件，碰巧某个文件先跑时，会用到另一个还没准备好的变量，整个插件就加载失败；重启或更新后第一次打开最容易碰上。新加的测试会把每个文件都当成第一个加载试一遍。

• AMB is no longer missing now and then when ComfyUI opens (a reload brought it back). ComfyUI loads every js file of the plugin at once, and when one file happened to run first it used a value of another file not made yet, and the whole plugin failed; the first open after a restart or update met it most. A new test runs each file first.

• AI 看得到工具的回答了：AI 软件收到太长的回答（Antigravity 大约 4,000 字节以上）会存成文件不看，AI 就什么也没拿到，只好缩小范围重查（考试里为挑一个 LoRA 搜了 8 次）。现在搜模型的回答只留有用的（同类模型的加载节点只写一次，触发词列前四个，LoRA 带上作者示例图的权重），读大画布时后面的节点写成一行，其他列表放不下的列出名字，告诉 AI 怎么查剩下的。

• AI apps get to read the tools' answers: an app files an answer past its size (about 4,000 bytes in Antigravity) away unread, so the AI had nothing and searched again with smaller limits (eight searches for one LoRA in the exam). Model search now keeps what is used (a type's loaders said once, four trigger words, a LoRA's weight from its author's pictures), a big canvas gives its last nodes a line each, and any other list names what did not fit and says how to get it.

• 画布是空的时候，AI 跑图或读画布会被告知先打开一个工作流（存的、示例，或者图库里某张图自带的），不再让它去"加一个保存节点"。AI 查一张图的参数时，也会知道这张图的工作流能直接打开。

• On an empty canvas the AI is told to open a workflow first (a saved one, an example, or a gallery picture's own) instead of being told to add a save node. Reading a picture's settings, the AI learns that its workflow can be opened as it was.

• AI 发来只写着"test"这类试探的问题，不再放到你面前等回答，AI 会收到正确的问法。任务上的"出图 N 张"只数真正跑出来的图，AI 找到给你看的旧图不算。

• A question that only says "test" or the like no longer waits in front of you: the AI is told how to ask. A task's picture count counts the pictures its runs made, not old ones the AI found to show you.

• AI 想改图库里的某一张图时，可以直接打开那张图自带的工作流（open_workflow 写图片名），不用再打开原来的工作流卡，把换过的模型、LoRA 和提示词一样一样设回去。打开的是新标签页，你开着的那页不动。Gemini 实测，这类题从 28 次调用降到 21 次。

• When the AI changes a picture from the gallery, it can open the workflow that picture carries (open_workflow with the picture's name), instead of opening the saved card and putting back the model, LoRA and prompt that had been changed one by one. It opens in a new tab and leaves yours as it was. In a real Gemini run such a task went from 28 calls to 21.

• 在设置里切换语言后，AI 工作台的任务列表和时间线马上跟着换。以前有一部分要刷新页面才变。

• Switching the language in Settings now changes the AI workbench's task list and timeline at once. Part of them used to wait for a page reload.

• 任务上的“出图 N 张”不再把同一张图算两次。以前 AI 汇报时附上刚跑出的那张图，会被多数一次。

• A task's picture count no longer counts one picture twice. When the AI's report showed the picture its run had made, it was counted again.

• AI 看工作流清楚多了（Gemini 实测发现的）：列表以前一次两万多字，AI 软件直接存成文件不看了，现在卡片多时只给名字、路线和备注。每张卡的路线说明以前一直是空的，现在有了；按模型找卡片也恢复了。官方示例总会列出名字，用中文（图生图、两段放大）也搜得到。

• AI apps see your workflows better (found in a real Gemini run): the list was over 20,000 characters, which an AI app files away unread, and is now names, routes and notes when there are many. Each card's route line was always empty and now says the route, and finding cards by model works again. Official examples are always named, and found by their Chinese names too.

• 新任务里先写了留言的话，任务列表就用这句话当标题，不再都叫“接手任务”。

• A new task you wrote to takes its name from what you wrote, instead of the AI's "take the task".

• 两个 Anima 两段放大示例写明了：要多张图就多跑几次，别调批次。Anima 的 VAE 会把一批图当成视频帧，第二段只剩一张。

• The two Anima two-pass examples now say to run several times for several pictures, not to raise the batch: Anima's VAE takes a batch for video frames, and the second pass is left with one.

• 当前节点的提示词卡片现在都标着“正面提示词”或“负面提示词”：从图片存下来的也按它原来接在采样器哪个口上来认，以前只有工作流来源的才标，图片和工作流同时有时标注还会丢。

• Current node's prompt cards all say positive or negative now: values kept from a picture are told by what they were wired to in that workflow. Before, only a workflow's said so, and the label was lost when a picture held the same text.

• 回到一个存档时，输出节点上的图也跟着换：画布是哪个存档，节点上就是哪个存档跑出来的图。跑了几次就都排在节点里，和批量出图一样，新的在前，最多 16 张。这个存档没有图时节点空着，不留别的存档的图，免得看错。只对这个版本之后跑的图有效。

• 回到存档时数字有时不亮、撤销多出一步的问题修好了：画布上有 Preview Image 这类节点时，每隔一次就会这样。

• Going back to a save brings its pictures back too: the output nodes show what that save made. Several runs sit on the node together as a batch does, the newest first, up to 16. A save with no picture leaves the node empty rather than showing another save's. This works for runs from this version on.

• Going back to a save no longer loses the lit number or leaves an empty undo step every other time, as it did with a node like Preview Image on the canvas.

• 发版前把每个页面按中英文、深浅色和 480 窄宽走了一遍，修了一批小问题。浅色主题下，模型整理页的标题和文件名、任务上“接着做”的标签原来几乎看不见。窄窗口下，任务的“接着哪个任务”标签伸出窗口、× 点不到，设置页的主题按钮把页面撑出横向滚动，模型页的搜索框压住标题，工作流详情的“新预设”被切掉。

• 英文界面的列表和冒号不再是中文标点，“1 saves”这类单复数对了，左边栏的 AI 工作台写作 “AI”，一条漏翻的提示补上了。深色主题下浅紫色按钮和角标上的字改成深色，看得更清楚。

• A round through every page in both languages, all themes and at 480 fixed small things. In the light theme, the Tidy page's headings and file names and a task's "goes on in" chip could hardly be read. In a narrow window, a task's "goes on from" chip ran out of the window with its × out of reach, the theme buttons in Settings pushed the page sideways, the Models search sat over the title, and "New preset" was cut off.

• The English window no longer shows Chinese punctuation in lists and after labels, "1 saves" and the like have the right plural, the rail says "AI" for the AI workbench, and one line that was left in Chinese is translated. In the dark themes, text on pale accent buttons and badges is dark now, so it reads better.

• 侧栏下面的工具按常用程度排了：当前节点在最上面，然后是模型检查，模型扫描在最下面。

• The rail's tools go by how often they are used: Current node first, then Model Check, with the scan last.

• 当前节点的参数卡清爽了：来源不再一条一行，改成按工作流、搭配、图片、这里存的分开的小标签，同一渠道几条合成一个，点标签直接过去，几条的话先列出来。

• 当前节点现在也列出搭配里同类节点的数值，可以对比和套用；普通搭配给提示词框，节点结构搭配给每个框。

• Current node's cards are easier to read: where a set came from is a chip per channel (workflow, combo, picture, saved here) instead of a line per source, a press goes there, and several in a channel are listed first.

• Current node now lists the values combos hold for that node type, to compare and put in: a plain combo's prompt on a prompt box, every box of a node structure.

• 给 AI 挑模型时和模型页一样好找了：类型用平常的叫法，有基础模型的分类条，版本和名字分开写，同一模型的几个版本合成一张卡，点版本就能换。

• Picking models for the AI works like the Models page: types by their usual names, base model chips, the version apart from the name, and a model's versions on one card, changed from its version.

• 工作流页多了“官方示例”：文生图、两段放大、图生图、局部重绘，打开时自动选一个你已有的模型。两段放大有 SDXL / SD 1.5、Anima 和推荐的 Anima · Tile 锁图三个版本，缺节点包或模型的版本会写明缺什么。新装的人和 AI 都有现成路线可用；AI 没有合适的已存工作流时会用它们。

• The Workflows page has "Official examples": text to picture, two passes, picture to picture and repainting an area, each opening on a model you already have. Two passes comes for SDXL / SD 1.5, for Anima, and as the recommended Anima held by Tile; a version says what pack or model it lacks. New installs and the AI have routes to start from; the AI gets them when no saved workflow fits.

• 可以自己选做法了：在“做法”里点“这次用它”，只对当前任务生效。AI 会收到，这一类活 AMB 直接按你选的搭，AI 想用别的也不行。

• You can pick the way yourself: "Use it this time" in Ways, for the current task only. The AI is told, and Anomalous builds your pick for that kind of job, whichever one the AI asks for.

• 点“做法”马上出来，不再每次重画：打开 AI 工作台时就提前读好，再点时停在你上次看的位置，展开的详情也还在；开关一个做法也不会跳回顶部。

• Ways opens at once and no longer draws itself twice: it is read when the AI workbench opens, and comes back where you left it, details still open; turning a way on or off no longer jumps to the top.

• 新任务的页面清爽了：只留一句发给 AI 的话和一个复制按钮。要先交代的话和图都放下面的留言框，不再有重复的两行。连过 AI 以后也不再看到“第一次用？”；从没连过时，先告诉你怎么连。

• The new task page is plain now: one sentence for the AI and a Copy button. What to say first and pictures go in the message box below, with no second place for them. Once an AI has connected, "First time?" is gone; until then, how to connect comes first.

• AI 工作台的任务列表：一条线上的任务排在一起，后续任务缩进挂在最早的任务下面，一眼看出谁接着谁。每个任务行上多了 ⋯，改名、隐私模式、从列表移除不用先打开任务。移除中间的任务时，线不会断。

• The AI workbench's task list keeps a line of tasks together: follow-ups hang under the first task, so you see what goes on from what. Each task row has a ⋯ for rename, privacy mode and remove, without opening the task first. Removing a task in the middle does not ...

v1.62.4 (Fix: opened at the computer's LAN address, changes were refused)

• ComfyUI 只监听局域网地址（比如 --listen 192.168.1.5）时，在运行 ComfyUI 的这台电脑上用这个地址打开，现在也算本机：扫描、改设置、填代理都能用了。以前它被当成别的设备，只能浏览。
• 设置页的代理和“测试连接”从别的设备打开时会说清原因：只能在运行 ComfyUI 的电脑上改。以前代理总说“格式写错了”，测试说“连不上 Civitai”，其实格式和网络都没问题。被拒绝的提示也会指向窗口顶上的“仅浏览”：点开有从别的设备也能用的办法（服务器上的 allow_remote_changes 文件）。
• With ComfyUI listening on its LAN address only (say --listen 192.168.1.5), opening it at that address on the computer it runs on now counts as this computer: scans, settings and the proxy work. Before, it was taken for another device and could only look.
• Opened from another device, the settings page's proxy and Test connection now say why nothing happened: only the computer ComfyUI runs on can do that. Before, the proxy always said the format was wrong and the test said Civitai could not be reached, though neither was the case. Refusals also point to the View only chip at the top of the window, which says how to use the plugin from other devices too (the allow_remote_changes file on the server).

v1.62.3 (Fix: ComfyUI-Manager took other packs' nodes for AMB's)

• ComfyUI-Manager 不再把别的插件的节点算到 AMB 头上：AMB 没有节点，是测试文件里的假节点表让它误以为有，于是缺 Power Lora Loader 等节点时会推荐装 AMB。Manager 重新扫描后生效。AMB 本身用起来没有变化。
• ComfyUI-Manager no longer takes other packs' nodes for AMB's: AMB has no nodes, but node tables in its test files looked like some, so a missing Power Lora Loader and the like could offer AMB. Takes effect once Manager scans again. AMB itself works as before.

v1.62.2 (Fix: workflow cards squeezed into strips)

• 工作流很多时，卡片不再被压成细条，列表正常滚动。
• With many workflows, the cards are no longer squeezed into thin strips; the list scrolls.

v1.62.1 (Fix: Anomalous sometimes did not start)

• 修好一个偶发的启动故障：有时打开 ComfyUI，AMB 整个没加载，悬浮按钮和窗口都没有，刷新一下又好了。
• Fixed a startup failure that came now and then: ComfyUI opened with no Anomalous at all, no button and no window, until a reload.

v1.62.0 (AI Workbench, AI Builds Workflows on the Canvas, Repaint an Area, Materials Side Panel)

• 音频库搬进了“角色语音”页：页面顶上的标签现在是 角色 ｜ 配音 ｜ 音频库，左边栏少了一项。加了 AI 工作台以后左边栏有 10 项，数字键只有 1–9，音频库就没键可按了；现在 9 打开角色语音，并回到你上次停的那个标签。

• The audio gallery moved into Voices: the tabs at the top read Characters | Voice-over | Audio, and the rail is one entry shorter. With the AI workbench the rail had ten entries for nine number keys, leaving the gallery without one; 9 now opens Voices on the view used last.

• AI 能拿刚出的图接着改了：搭“在这张图上重画”的工作流时，加载图像节点也认输出文件夹里的图（xxx.png [output]，ComfyUI 本来就支持）；以前 input 文件夹里没有这张图就被拦下。左侧栏的“模型”“图库”不再显示成英文 models / gallery。

• The AI can redraw a picture it just made: a Load Image it builds now takes an output picture (name.png [output], as ComfyUI itself does) instead of being refused when the input folder lacks it. The rail shows Models / Gallery again instead of the raw keys.

• 修复完成后的任务接续：AI 带着原任务 ID 查看结果、补面板，也会留在刚完成的任务里，不会另开空任务或把面板放进另一个任务。错误标记只在对应重试完成并成功后消除；等待审批、仍在运行、已取消，以及另一个节点或设置的成功都不算修好。

• Fixed follow-ups carrying an explicit task ID: reads and panels immediately after completion stay with that task instead of creating an empty task or placing a panel in another one. Error recovery requires a completed successful retry of the recorded target; pending approval, running or cancelled attempts and success on another node or setting do not clear it.

• 参数名自动纠正现在也用于审批和任务记录：你批准后，AI 改用正确参数名重试不再重复问，换模型的结果卡片也不会漏掉。展台和底部提问条同步同一道题的草稿，刷新时保留原来的光标、选区和编辑位置。

• Corrected argument names now carry through approval and task recording: retrying with the proper name reuses an unused approval, and model changes keep their result cards. The stage and bottom question strip share each answer draft; refresh preserves the caret, selection and editing surface.

• 修复任务接续与等待回答：按你最近一次未读回复接续任务，后台进度更新不再打乱顺序；选择题回答也能接回原任务，仍有工具正在执行的任务不会被新会话自动接走。读取答案不指定面板时也会正确处理超时；到时间自动选择或恰好收到你的回答，都会在当次调用里交给 AI。等选择题时发的留言和图片不再被读走却漏传。

• Fixed task resumption and answer waiting: resume by the latest unread user reply, including panel answers, rather than background updates; a task with a tool still running is excluded from automatic takeover. Reading answers without a panel now handles expired questions too. Timeout choices and user answers arriving at that boundary reach the AI in the same call; messages and pictures sent while it waits for a question are no longer consumed without delivery.

• 任务列表更准：AI 改正过来的错（比如参数写错、下一次就对了）不再让整个任务显示成“出错”，只有 AI 没跨过去的才算，红标改叫“有一步没做成”；AI 说做完以后马上再看一眼结果，这一眼记在原来的任务里，不再多出一个只有一步的空任务；接着给你看前后对比、问你满不满意，也回到原来那个任务里（任务重新打开），不再另起一个；你点“新任务”写好的任务一直等到 AI 来接（或者你再开一个新的），不会过了 12 小时悄悄失效、而页面还写着“等 AI 接手”。

• AI 少走弯路、少猜：

  • AI 把必填参数写成了常见的另一个名字（比如打开工作流时把 workflow 写成 name、看图时把 image 写成 file）时，AMB 直接照它的意思做，并告诉 AI 正确的名字，省一次来回。只认考试里见过的这几种叫法；别的名字照旧报错、列出该写什么，免得猜错意思。
  • 时间线里 AI“放大看图”那一步显示成代码名，改成了中文名。
  • 去掉了“给 AI 的工具”里的 日常：所有工具都直接给。Antigravity、Claude Code、Cursor、VS Code 这些软件会自己按需读工具说明，AMB 再做一层“现查现用”只会让 AI 多绕路（一次实测 29 次调用里 18 次花在查工具和转一道上）。
  • AMB 更新以后，AI 软件还拿着旧的工具列表来调用时，任务标题下面会出现一行黄字，告诉你怎么让它重新连接；AI 那边也收到同样的提醒。AI 软件重新取了列表，黄字自己消失。
  • AI 加 LoRA 或换模型时，AMB 对比 LoRA 和主模型的底模：一个是 SD 1.5、一个是 SDXL 这种用不了的，会提醒 AI 并告诉它按底模去找；Pony 的 LoRA 用在 Illustrious 上这种能用但可能效果差的，也会轻轻提一句。只提醒，不拦着。
  • AI 看模型详情时，能看到作者示例图用的采样器、步数、CFG、尺寸，LoRA 还有作者用的权重（来自扫描时存下的 Civitai 资料），还有版本名。
  • 跑图被 ComfyUI 拒绝时，AI 收到的不再是原样的英文报错，而是“哪个节点、缺什么、该用哪个工具修”。
  • 用 Gemini 3.8 Flash 实际考了 7 道常见任务（出图、换模型加 LoRA、问你再改、缺模型、找图看参数、搭放大分支、修断线），失败多半出在第一次用某个工具时参数写错，于是：参数写错时，AI 收到的报错会列出这个工具该写哪些参数，下一次就对；同一个东西在各工具里叫同一个名字（写提示词和跑图都叫 prompt，打开工作流和跑图都叫 workflow）；等待时间写超了（比如 120 秒）按最长的 55 秒等，不再报错；按类型找模型时写单数（checkpoint、lora）也能找到。
  • 很久以前（两个多小时前）留下、AI 一直没收到的留言或标注，不再把一个新对话拉进那个旧任务。
  • 关卡窗口新增 等不到回答时：停下等我（默认）或 按推荐的继续。选后者时，到了等待时间你还没答，AI 标了推荐项的选择题就按推荐的选，展台和“决策”里都写着“到时间，按推荐选了”；没有推荐项的问题、你已经动手写了回答的问题、关卡里“先问我”的事，永远不会替你答。

• The task list says what is true: an error the AI got past (a wrong argument right on the next call) no longer marks the whole task as failed, only one it did not get past does, and the red mark reads "A step did not go through"; a look the AI takes right after saying it is done goes into that task instead of making an empty one-step task, and a before and after or a "happy with it?" it shows next opens that task again instead of starting another; a task you start with New task waits until an AI takes it (or you start another), instead of quietly lapsing after 12 hours while its page still says it waits.

• The AI takes fewer detours and guesses less:

  • When a call writes its one needed argument under a known other name (name for open_workflow's workflow, file for image), Anomalous takes it as meant and tells the AI the right name, saving a round trip. Only the names seen in the exams count; any other is answered with what the tool takes, so no meaning is guessed.
  • The timeline names the AI's "look closer at a picture" step instead of showing its code name.
  • The tool set's Daily choice is gone: AI apps get every tool. Antigravity, Claude Code, Cursor and VS Code load tool descriptions as they need them, so looking tools up through Anomalous only added detours (in one run 18 of 29 calls).
  • When an AI app still calls by the tool list of an older Anomalous, a yellow line under the task header says how to reconnect it, and the AI is told too; it goes once the app fetches the list again.
  • Adding a LoRA or swapping a model, Anomalous compares the base models of the LoRA and the main model: one that cannot work (SD 1.5 on SDXL) is pointed out with how to find a fitting one; one that loads but may look off (a Pony LoRA on Illustrious) gets a softer note. A note only, never a block.
  • Model details give the AI the sampler, steps, CFG and size the author's example pictures used, and for a LoRA the weight (from the Civitai record a scan keeps), and the version name.
  • When ComfyUI refuses a run, the AI gets which node, what is missing and the tool that fixes it, instead of ComfyUI's raw report.
  • Gemini 3.8 Flash took seven everyday tasks as an exam (a run, a model and LoRA swap, asking you first, a missing model, a picture's settings, an upscale branch, a broken link); most failed calls were a tool's first use with wrong arguments, so: an argument error now lists what the tool takes, so the next call is right; one thing has one argument name across tools (prompt for set_prompt as for run_workflow, workflow for open_workflow as for run_workflow); a wait longer than 55 s waits 55 s instead of failing; a model type in the singular (checkpoint, lora) finds them too.
  • A message or marks the AI never got, left more than two hours ago, no longer draws a new chat into that old task.
  • The Gates window has When I don't answer: Wait for me (the default) or Go with the recommendation, where a choice the AI marked a recommended option for takes it once your waiting time runs out, shown on the stage and under Decisions; a question without a recommendation or one you began to answer, and anything a gate asks first, is never answered for you.

• AI 问你的问题更好答：选项是一按就发出去的按钮，右边有箭头，问题下面写着“点一个就发给 AI”，点了马上变成选中的样子；让 AI 定 挪到问题右上角，不再跟选项挤在一起被当成“确认”；“都不是”那一行默认收起成 都不是，我自己说…，关卡卡片上的原因也收起成 说说想怎么改…，点了才出输入框。展台、留言框上方的提问条、关卡卡片都一样。关卡窗口里几条说明改对了：改画布也管加节点、连线、新建工作流；跑完以后那几句话在留言框上方；“先问我”和按钮上的字对上了。写了一半的回答不再被清掉：AI 又问了新问题、读走了另一个回答时，字和光标都还在，那个问题一直排在提问条最前面；窄窗口里你正在打字时，新问题只点亮展台标签，不把你切过去；用输入法按回车选字时，不再把半句回答发出去。

• Questions from the AI are easier to answer: options are buttons that send at a press, with an arrow and "One press sends it to the AI" above them, and the pressed one shows at once; Let the AI decide sits at the question's top right instead of under the options where it read as a confirm button; a line of your own stays folded as None of these, I'll say it…, and the Gates card's reason as Say how to change it…. The same on the stage, the question strip over the message box and the Gates card. Gates window texts fixed: changing the canvas includes adding nodes, wiring and new workflows; the lines after a run are over the message box; the intro names "Ask me" as the button does. A half-typed answer is no longer wiped when the AI asks something new or reads another answer: the text and the cursor stay, and that question stays first in the strip; in a narrow window a new question lights the stage tab instead of taking you there while you type; Enter that picks a word in an input method no longer sends half an answer.

• 局部重绘前先对一下范围：你圈得不一定准，AI 看了标好的图可以挪动、改形状（方框 / 椭圆）、往外扩、把边缘做软，然后在展台把 你圈的 和 AI 调整后 并排给你看，你点 用 AI 调整的 / 用我圈的 / 我重新圈，或者写一句想怎么改。选好以后 AMB 把带这个遮罩的图放进 input 文件夹，AI 收到文件名和一句常用重绘节点的提示；画布上怎么加节点、怎么连线仍然由 AI 自己搭，不套固定模板。

• Agree on the area before repainting: your marks are rough, so the AI can move, reshape (box or ellipse), grow and soften the area and show Your marks beside The AI's area on the stage; you pick Use the AI's / Use mine / I'll mark it again or write what to change. Anomalous then puts the picture with that mask into the input folder and the AI gets its name with a short note on the usual repaint nodes; how to build the repaint on the canvas stays the AI's, no fixed template.

• AI 问你问题后会等你：关卡窗口新增 等我回答（1 分钟 / 3 分钟 / 10 分钟，默认 3 分钟）。一次调用最多等 55 秒，你没答它就接着等，直到这个时间；你在回答框或留言框里打字时，停手一分钟内都算还在答。AI 写完说明或放出展台面板后会被提醒：收工前先问你下一步做什么、要不要开始（跑图、下载、扫描时除外），说明里写了问句的，提醒它改用提问。问题上的“这题跳过”改叫 让 AI 定，AI 收到后自己决定、记下决定，不再把同一个问题再问一遍。

• The AI waits for your answer: the Gates window has Waiting for your answer (1, 3 or 10 minutes, 3 by default). One call waits up to 55 s and the AI then waits on until that time, and while you type in an answer or the message box until a minute after your last key. After a note or a panel it is reminded to ask what next or whether to go ahead before it ends its turn (not while a run, download or scan goes on), and to put a question in its note into ask_user. "Skip this one" is now Let the AI decide: the AI decides, records it and does not ask again.

• 记录里不再出现“你在 AMB 里改了：保存图像 $$canvas-image-preview”：那是 ComfyUI 显示预览图时自己改的内部字段，不是你的改动，以前会被当成你的改动发给 AI。

• The record no longer says you changed a Save Image node's `$$can...

v1.61.1 (Delete a Character)

• 角色卡上加了橘色的 删除角色 按钮（Anomalous_TTS 议题 #2）：确认以后，角色的整个文件夹（模型、参考音频、设置）移到系统回收站，需要时能从回收站还原；U 盘、网络盘上没有回收站就不删。设置坏了、读不出来的角色也能删。
• Character cards have an orange Delete character button (Anomalous_TTS issue #2): after a confirm, the character's whole folder (models, reference audio, settings) goes to the Recycle Bin, where it can be restored; on a drive without one nothing is deleted. Broken characters can be deleted too.

需要 / Needs Anomalous TTS v1.1.1：旧版 TTS 不显示这个按钮，其他照常 / with an older one the button is not shown and nothing else changes.

v1.61.0 (Quick Search, Canvas Menu, Combos From Pictures, Tidier Model Folders, Multi-Character Voice-over)

Needs Anomalous TTS v1.1.0 for the multi-character voice-over.

🔎 Quick search on the canvas: Alt+B (画布快速搜索)

• Press Alt+B over the canvas (or right-click the empty canvas → AMB → Quick search…) and type: combos, models and saved prompts that match show at once, with their pictures. A model is found by its name, file, base model or trigger words; a combo by its name, models or prompt; a prompt by its text.
• Enter puts it where your pointer was: a combo as its group of nodes and a model in its loader node, both following the pointer until you click; a prompt as a new prompt node. ↑ ↓ pick, Tab shows only combos, models or prompts, Esc closes.
• The shortcut is a ComfyUI command (AMB quick search), so it can be changed in ComfyUI's keybinding settings like any other. Nothing is read until the bar opens.

🖱️ AMB in the canvas's right-click menu (画布右键菜单)

• Right-click a node → AMB offers what Current node does for it, without opening the browser first: Open in AMB (a prompt node opens Prompts on it, any other node Current node), Change model by picture, Insert a LoRA before / after, Apply saved values, Pictures made with this model, Model details, Save this node's values and Save as combo. Each entry shows only where it applies: a checkpoint loader has no "before", a prompt node no model entries.
• Right-click the empty canvas → AMB opens any of AMB's pages.
• Make it yours: Settings → Canvas → Right-click menu. Switch each entry on or off and move it up or down; a line above each list shows what the menu will offer. Two cards, each with a small drawing of the menu, choose whether AMB's entries sit together under "AMB" (the default, ComfyUI's menu gets one line more) or are listed in the menu. Adjust this menu… at the end of the submenu leads there. Save as combo moved into the AMB submenu with the rest.
• In the model picker, Esc now closes the picker wherever the focus is, and only the picker: it used to close the browser behind it when the search box had lost the focus.
• The floating AMB button no longer sits on the head of a model picker opened from this menu (with the browser closed).
• Fixed: when a model folder is a link (a junction to another drive, say), a combo kept from a picture and other lookups of saved model paths got a wrong place for the model, or failed.

🧩 A combo's page, redone (搭配详情重做)

• A head that says what it is: the picture (the one you kept it from, with a See the picture button; else the main model's cover), the name in large type with a ✎ Rename button beside it, and chips for its base model, main model and LoRAs. Put on canvas and More (delete) sit under them.
• Only what is chosen: the main model and the LoRAs are short rows with their covers; Change and + Add a LoRA open the grid of compatible models only when you want it, instead of every model of that base filling the page. ✕ takes a LoRA out; the numbers say their order.
• No Save button: every change (models, name, prompt) saves itself, as it already did. Fixed: a tag changed in the bilingual view was not saved into the prompt.

★ See only the kept pictures (图库只看收藏)

• ★ Kept beside the gallery's search lists only the pictures you kept as a workflow, combo or prompt (the ★ on their cards). Works with the search and with Pictures / Videos. Pointing at a card's ★ says what it was kept as.

📌 A picture kept as a combo keeps its nodes as they were (图片存搭配，节点原样记住)

• Keep → Its combo now keeps the nodes that made the picture's model and prompts: the model and CLIP loaders, the whole LoRA chain (Power Lora Loader and other packs' nodes too), CLIP skip and model patches, the positive and negative prompt nodes, with every value as the picture had it. The sampler, size and decoding are left out, so Put on canvas gives a group you wire to your own sampler, like a combo saved from the canvas.
• It used to keep only the main model, the LoRAs' names and the positive prompt: LoRA strengths came back as 1.0, and the negative prompt and CLIP skip were lost. A combo only remembers; it never changes a value (now one of the plugin's rules). Combos kept before stay as they are.
• Models are named by where they are on this computer; one this computer lacks is kept by the picture's name and said after keeping. A workflow without a sampler to read from still gives the simple kind.
• Fixed: after a picture was kept (as a workflow, combo or prompt) or its parameters compared, the prompt boxes of its workflow could stay on ComfyUI's page and show at its top left, over the tabs. The workflow read off the picture is now taken away once it is read.

🩺 Model Check and ComfyUI-Manager's list (模型检查与 Manager 清单)

• A Download by name from ComfyUI-Manager's model list now also says where the list's file is (the site and repository), to help you decide.
• A list entry that does not say its type is no longer offered for a node whose type is known.
• Without ComfyUI Manager, Model Check says it would add one more place to download from when a model's source is not found.
• A foundation component (VAE, text encoder, CLIP Vision…) no source was found for gets Search Hugging Face (its mirror when that is on): the search page for its file name, for you to pick from.
• The rules of Model Check (what counts as the same model, the confidence levels, foundation components, download sources and where ComfyUI-Manager fits) are now written in one place: docs/guides/model-check-rules.md.

⭐ Combos kept from a picture are easy to tell apart (图片存的搭配一眼认得出)

• The card shows the picture you kept it from, with the main model's cover as a small round badge in its corner, instead of the model's cover alone (several combos of one model used to look the same).
• A name that says which one it is: the first LoRA (or the main model) and the prompt's first two telling tags, like Ahri · fox ears, kimono, skipping words nearly every picture has (1girl, solo, looking at viewer…).
• Rename it right away: the message after keeping shows the picture and the name in a box; type a new one and press Enter. A combo's page has a Rename button too.

📒 Prompt cards look like notebook pages (提示词卡片换成笔记本样式)

• In Prompt Studio each card is a page with a bound spine in its colour (green positive, red negative), its title, a Positive / Negative label and its first words as small tags instead of a cut-off line of text. The card open as the reference wears a ribbon; Common and My prompts are index tabs with how many cards they hold. Dark and light themes alike.

🗂️ Tidier model folders (模型文件夹更清爽)

• Settings → Models → Files that come with models: what a scan found (.info), your names and notes (.anomalous.json) and Civitai's picture kept aside can go in a hidden .amb folder inside each model folder instead of beside every model, leaving only the model and its cover there. Covers stay beside the model, where ComfyUI and other plugins look. Choosing a place says how many files will move, then moves them (only moves; switch back any time); a file that cannot move still works where it is.
• No more duplicate cover copies: scans used to keep a copy of Civitai's picture identical to the cover for every model. Now the cover is noted as Civitai's and a copy is kept only when you put your own cover in its place (so you can still bring Civitai's back). Check and tidy in the same place sends the existing duplicates to the Recycle Bin.

🎭 Several characters in one script (多角色配音)

• Every line has its character: on Voice-over, click a character in the list on the left, then paste: the lines are its. To give lines to another, pick them (a click; Ctrl / ⌘ adds one, Shift a run; Esc lets go) and choose who reads them on the right, or click the name on a line for a menu of characters. Each line shows who reads it in that character's colour, and its emotions are that character's. The Who speaks bar at the top is gone: the list does its job.
• Three layers: the whole script, a character, a line (整份剧本 / 角色 / 句子，三层分清): the left side picks, the side panel changes. Nothing picked: the whole script, what everyone reads with, and each character in it. A character clicked in the list: that character in this script (only its lines). Picked lines: only them, with who reads them. A path on top, Whole script › Aris › Line 2, says which layer you are changing and leads back up (Esc too). Every layer has the same three rows, speed, tone and language, and each row says where its value comes from: this line (yellow), the character (purple), the whole script or the default (grey), or the character's usual setting; a click on it goes to that layer, ↺ lets this layer follow the one above. The more particular wins: a line over its character over the whole script over the usual settings. The settings that last beyond this script (the tones, a character's own settings) are links at the foot, apart from the rest. The tone can be heard on that line right there; Numbers shows how it compares and its numbers. With many characters, Read by lists the ones in this script (with their colours) and the rest under Other characters, a menu with a search box; a card's name menu gets the same search box.
• Drafts (草稿箱): every script is kept as a draft on its own, half a second after a change, in ComfyUI's user folder (so clearing the browser or using another one loses nothing). A bar over the script names it (a click renames it; unnamed, its first line names it) and has New (a blank script, this one stays), Drafts and Close script. Drafts shows a card per script, the last changed first, with who reads it and how many lines: open, rename, copy, delete. Deleted scripts wait in Recently deleted for 30 days. A reload opens the script that wa...

v1.60.2 (Fix: the browser opens whatever the plugin folder is called)

💬 From a Bilibili comment (来自 B 站评论)

• The browser opens when the plugin folder has another name (插件文件夹改了名也能打开): a plugin unpacked from a ZIP of GitHub sits in Anomalous_Model_Browser-main; the browser's look did not load there, so a click on Model Browser seemed to do nothing (the browser lay under the canvas, without its look) and that button showed in the top bar although the floating button was chosen. The plugin now finds its files under whatever name its folder has. Updates to the plugin also reach the page at once there again, as with the usual name.

v1.60.1 (Cloud Servers, Generation Presets, Full Window, Faster Scan Page)

🔍 Scanning (扫描)

• The scan page opens at once, also with many models: it shows the last counts right away, says it is updating them, and shows the fresh ones when they come. What each model file's header says is now kept on disk (user/anomalous/model_heads.json, by file, read again only when the file changes), so a big library or a network drive is not read again after every restart either; requests that come while the page counts share that one count.
• See the models found on Civitai: the scan page's "Found on Civitai" count opens a list of them, with each model's Civitai name and base model, a box to find one by name, and a way to open each.
• Before an online scan, the plugin checks whether Civitai can be reached, on both its addresses. When it cannot, it asks instead of waiting on every model: Scan from the files (base models found, no covers or trigger words; a later scan fills them in), Set a proxy (opens Settings → Models → Network), or Cancel. A check that reached Civitai is trusted for five minutes.
• Civitai's other address (civitai.red) when civitai.com does not answer: some networks reach only the second, a cloud server in mainland China among them (tested on AutoDL: civitai.com timed out, civitai.red answered in about a second). Scans, import and Model Check's lookups and Settings → Network → Test ask civitai.red then, so models still get their names, trigger words and base models; the test says which address answered. The covers are on Civitai's picture server, which such networks often cannot reach either: after two covers in a row that would not download, a scan fetches no more covers instead of waiting on each.

🧩 Combos (搭配)

• A combo's models are cover cards, as in the model library: the cover on top, the name below. In a combo saved from picked nodes, the slot (Main model, LoRA…) is written on the cover and a click on the card changes the model; in a default combo, the main models and LoRAs to pick from are cover cards side by side. They used to be rows as wide as the window, with Change far from the name (模型卡片不再横向拉满整个窗口).

🎙️ Character voices (角色语音)

• Update Anomalous_TTS from GPT-SoVITS Settings (在 GPT-SoVITS 设置里更新 Anomalous_TTS): a new Anomalous_TTS version line shows the installed version; Check for updates looks up its published stable releases (only then does it go online), Update to v… switches to the newest one and offers to restart ComfyUI, and Back to v… undoes it. Anomalous_TTS now publishes numbered stable releases (the first is v1.0.0), so an update never lands on unfinished work. Only an Anomalous_TTS installed with git from its GitHub repository is changed (as ComfyUI Manager installs it), only from the computer ComfyUI runs on, and never over files you changed. When an import needs a newer Anomalous_TTS, the message points here.

• Generation presets (生成预设): GPT-SoVITS Settings has three presets, Steady (fewer slips, flatter), Standard (the old default) and Lively (livelier, slips now and then): pick the default one for every character, and one for each pair of languages, such as a Japanese character speaking Chinese (until now every such line was made steadier without asking; that is the rule you start with). With nothing changed, voices sound exactly as before.

• Presets you can hear and compare (听得到、比得出的预设): pick a preset and see where it sits between steadier and livelier next to your default (a ring for the default, a dot for the one picked), with a sentence such as "a little livelier than your default (Standard)". ▶ Listen to this one has a character read a line with that preset; every take uses the same seed, so only the preset changes, and the presets already heard stay as buttons to click back and forth. Folded under it are the numbers as bars against your default, the way the Current node card shows a change (old → new, ↑ ↓); each has its plain name and the usual one beside it (top_k, top_p, temperature, repetition_penalty), so values from the GPT-SoVITS WebUI are easy to match. Click a number to change it; a changed built-in number shows its built-in value, and Reset puts it back. Make default sets the preset every character uses. + New makes your own preset from a copy of the one picked, with a name of your own (such as 念旁白 for narration); it can be picked everywhere the built-in ones can, renamed, and deleted (the question says what uses it; those places go back to the general setting).

• Modes on Voice-over (配音时直接选模式): under How it is read, Auto (the presets and the character's own settings decide, line by line) or any preset for the whole script, shown against Auto the same way and heard from the ▶ on each line: it reads that line in the mode picked, and the modes already heard on it stay on the side as buttons to compare. The scale and the numbers fold away there, so the narrow side stays short. A preset's numbers can be changed right there and are saved into that preset. Links lead to the presets and to the character's generation settings. This replaces the folded fine-tuning numbers, which showed the node's defaults rather than what the presets used; Save as default now keeps the language and speed only.

• A character's own generation settings (角色特调): a new Generation settings button on each character card sets the preset and speed for the character as a whole and for each language it speaks, with the numbers themselves when a preset is not enough. Under each row are the numbers its lines will get, blue where they differ from the general presets. The node follows these settings too, so a workflow in ComfyUI sounds like generating here; numbers changed on the node itself still go first. Needs Anomalous_TTS v1.1.0 (update it from the same dialog).

• More room on Voice-over (配音页更宽敞): each line is one row (number, text, its buttons), so more lines fit; a bar between the lines and the generating side can be dragged to share the width (remembered; double-click for the default, about half each); Into your own workflow is folded to one line; who reads and their voices share one row, and the how-to line shows only until there are lines. When the side is wide enough, the modes' scale and numbers are shown open.

🪟 The window (窗口)

• Full window (全屏): a new button beside Dock makes the plugin fill the whole ComfyUI page, and turns it back; a double-click on the title bar does the same. It is remembered.

☁️ ComfyUI on a cloud server (云服务器上使用)

• How far it goes: fully supported on overseas cloud servers (RunPod…). On servers in mainland China (AutoDL…) only in part: browsing, offline scans and tidying work, but Civitai is mostly out of reach there, so lookups are unreliable and covers rarely download (海外服务器完整支持；国内服务器只能部分支持). Not fully tested yet: one AutoDL server so far, no overseas one. The cloud guide ends with a section for AI assistants (what is going on, the steps, how to check, what to do when it did not work, what to tell the user about safety), so a user can hand the link to their AI.
• On a rented cloud server (AutoDL, RunPod…) the page is always opened from another device, so since v1.58.1 the plugin only let it look: scans, downloads, moves and saves were refused. Now there are two ways: open ComfyUI through an SSH tunnel (the address starts with 127.0.0.1; it counts as the computer ComfyUI runs on, nothing to change), or make an empty file named allow_remote_changes in the plugin folder on the server and restart ComfyUI. With the file, every device that can open the page may change things; the page can neither make nor remove it, and the MCP endpoint stays for this computer only. How, and the risk: docs/guides/cloud-server.md.
• The View only chip beside the page title can be pressed: it says what is refused and the two ways above, with a Copy button for the one line to paste into the server's terminal (the plugin's real path on that server already filled in), so there is no folder to find. With the file there it reads Remote changes on, in yellow, and its dialog gives the line that turns it off again.
• Models that are links (模型是快捷方式/链接时): on a cloud server such as AutoDL every model in the models folder is a link into the platform's shared model disk. Scans wrote the info beside the link, but the cards read it from the disk behind it: a scanned model showed as not scanned, under the disk's file name (6b4679…) instead of its own, and editing, deleting, opening its folder or finding it from a saved workflow could be refused as "outside the models folder". A linked model is now taken as named, with its covers and info beside the link; edits, renames and deletes act on the link, never on the shared disk; moving one in Tidy moves the link instead of copying the model disk's file onto the server. Names leaving the folder (..) are refused as before.
• Models linked from another tool keep their info: a model that is a link with no info beside it shows the info beside the file it links to (A1111's Civitai Helper leaves .civitai.info there). Only read there; covers, edits and scans stay beside the link.

v1.60.0 (Model Import, Tidy, Parameter Compare, Workflows Rebuilt, Base Model Filter, Settings Tabs) — 2026-10-22

📥 Drop a model in, it goes to the right folder (拖进来自动归位)

• Models page → + Import, or drop model files anywhere on the models page: each file gets a card saying what it is — Checkpoint, LoRA, VAE, text encoder, diffusion model, ControlNet, Embedding, CLIP Vision, upscale model — with its base model, how sure that is, and one line on what such a model is for. It is read from the file's header only; .ckpt/.pt files, which can carry code, are never opened (you pick their type).
• Where it will go is shown and can be changed; by default it follows the download setting (by base model, such as loras/Illustrious)....