KeyFlow功能Features › AI 助手AI assistant
AI 助手：在任何应用里，按一个键AI assistant: one key, in any app
不用切到聊天窗口。选中文字、说一句话、问一个问题，结果回到你正在用的地方。文本、助理、语音、生图共用一个 Credit 余额和一份记忆。No switching to a chat window. Select text, say a sentence or ask a question, and the result lands where you're working. Text, assistant, voice and images share one credit balance and one memory.
文字项：⌃1 翻译、⌃2 优化、⌃3 解释Text items: ⌃1 translate, ⌃2 polish, ⌃3 explain
在任何应用里选中文字，按一个键。结果弹出在选区旁边，可以直接修改：↩ 替换原文，⌘C 拷贝，⌘R 重新生成，或换一个指令处理同一段文字。Select text in any app and press one key. The result pops up next to the selection and is editable: ↩ replaces the original, ⌘C copies, ⌘R regenerates, or switch to another item for the same text.
- ＋ 自定义指令：每条都能改名字、快捷键和提示词，比如「总结要点」「改成美式英文」Add your own items, each with its own name, hotkey and prompt: “summarize”, “American English”, and so on
- 「继续优化」：对结果再说一句要求；「更新指令」：把满意的改法写回指令，下次直接就对“Refine” adds a follow-up instruction; “Update prompt” writes a good refinement back into the item so next time is right the first time
- 识别文字（⌥/）的结果也能一键交给 AI 优化OCR results (⌥/) can be handed to AI with one click
助理：⌥Space，一条一直连着的对话Assistant: ⌥Space, one continuous conversation
屏幕顶部的胶囊里问任何问题，回答流式出现，需要时自动联网并附来源。有选中文字时自动带上。回答完 ↩ 追问，Esc 关掉，10 分钟内再打开还在。Ask anything in the capsule at the top of the screen. Answers stream in, search the web when needed and cite sources. Selected text comes along automatically. Press ↩ to follow up, Esc to close; reopen within 10 minutes and it's still there.
- 能帮你记待办、写笔记、把笔记拆成待办，做过的都能一键撤销It can add to-dos, write notes and split a note into to-dos, and everything it does can be undone in one click
- ⌘N 新对话，Tab 回到上一个，⌘[ / ⌘] 在历史里前后走，⌘↩ 打开对话历史窗口⌘N starts a new conversation, Tab returns to the previous one, ⌘[ / ⌘] walk through history, ⌘↩ opens the history window
- 新问题和最近的对话有关时，会提示「接着聊？」，Tab 接上When a new question relates to a recent conversation it offers to continue; press Tab to pick it up
- 对话只由你删除，不会自动清理Conversations are deleted only by you, never automatically
语音输入：按住 Fn 说话Voice input: hold Fn and talk
在任何输入框里按住 Fn（或右 ⌥）说话，实时看到文字；松开后插入整理好的句子：去掉「就是说、然后呢」，补上标点，人名和术语按你的词汇表写对。中英混说也没问题。Hold Fn (or right ⌥) in any text field and talk; words appear as you speak. Release and a tidy sentence is inserted: fillers removed, punctuation added, names and terms spelled the way your vocabulary says. Mixed Chinese and English works too.
- 使用云端识别，中文尤其准，随时可改为「仅本机识别」Cloud recognition is used, especially accurate for Chinese, and “on-device only” is a switch away
- 语音输入不花 Credit，云端识别也包含在内（合理使用）Voice input never spends credits, cloud recognition included (fair use)
- 说的话可以直接进笔记、进待办，或者当作助理的问题Speech can go straight into a note, a to-do, or a question for the assistant
生图：⌃0Images: ⌃0
选中一段描述，或者刚拷贝一张图，按 ⌃0。查看器里可以「继续优化」：换风格、改比例、再来一版，每一版都保留。Select a description, or copy an image, and press ⌃0. In the viewer you can refine: change the style or aspect ratio, try another version, and every version is kept.
- 三个速度档：极速、标准、精细；生图 35 Credit / 张（精细档 60）Three speed tiers: fast, standard and max; images cost 35 credits each (60 on max)