需要 Android 10 及以上手机，并先安装 Haminn。模型可以由你自己的电脑或云端服务提供。Needs Android 10 or later plus Haminn. Models can come from your own computer or a cloud service.
HamDrawLIVE
黄昏的海湾，远处有雾A bay at dusk, fog in the distance实时槽位 · 4 步Live slot · 4 steps
界面示意：涂鸦层与成图层共用一块画布，成图层随笔触实时重绘Interface sketch: the stroke layer and result layer share one canvas; the result redraws as you draw
实时是认真的Real-time, literally
一秒内眨眼出图， 不是几十秒一张。A blink per frame, not a minute per image.
实时不是把大模型加速，而是把工作拆成三个任务槽位：一个够快的小模型专门跟着你的笔走，一个局部重绘槽位只改你用「局部」标出来的那一块，一个高质量模型负责把你满意的画面渲染成 1024 大图。你一直在画，画面一直在跟。Real-time does not mean speeding up a big model. It means splitting the job into three task slots: a fast model that follows your pencil, a local redraw slot that rewrites only the region you marked, and a high-quality model that renders the frame you are happy with at 1024. You keep drawing, the picture keeps up.
实时槽位固定 512×512、2/4/8 步，用 DreamShaper8 LCM，CFG 固定为 2。The live slot stays at 512×512 with 2, 4 or 8 steps on DreamShaper8 LCM and a fixed CFG of 2.
服务端先把画布做轻微模糊副本再与原图混合，你的笔迹因此不会变成刺噪点。The server blends a lightly blurred copy with your canvas so strokes do not turn into noise.
画布原样提交：颜色、透明度、构图都不改写。你看到的就是模型看到的。The canvas is submitted as it looks: no rewriting of colour, opacity or composition. What you see is what the model sees.
成图层和元素层的透明度各自可调，画的时候一直看得见自己的笔；生成过程中还可以继续画，快速操作会自动排队合并。Result and element layers fade independently, so your own strokes stay visible while you work. You can keep drawing during a generation; rapid actions queue and coalesce.
三个槽位，各干各的Three slots, three jobs
实时Live
DreamShaper8 LCM
512×512 · 2/4/8 步 · CFG 2。跟着笔走的那个，几乎任何独立显卡都能跑。512×512, 2/4/8 steps, CFG 2. The one that follows your pencil, comfortably within reach of almost any discrete GPU.
局部Local
局部重绘Local redraw
用「局部」涂红要改的区域，写下这块要改成什么，只提交这一句描述和那一块画面。改一处，不用重画整张。Mark the area red, write what it should become, and only that sentence and that region are submitted. Fix one part without redrawing everything.
渲染Render
Flux.2 Klein 4B
distilled 版 2/4 步，8 步换 base。原生 1024×1024，四步就能出可用结果。Distilled for 2 or 4 steps, base at 8. Native 1024×1024 and usable output in four steps.
高保真Fidelity
QwenImage 2512
牺牲一些速度换细节与构图服从度，适合显存充足的台式机。Trades some speed for detail and instruction following; best on a desktop with headroom.
三个槽位完全独立：快速生图、局部重绘、高清渲染分别保存自己的模型设置，三者共用同一套插件地址与密码；改一个不会动另一个。All three task slots are independent: quick draw, local redraw and render keep their own model settings while sharing one plugin address and password. Changing one never rewrites the others.
涂鸦工具Drawing tools
按钮就在那儿，一按就会。The buttons are right there. Press one.
下面每个图标，都是应用里真实存在的按钮。AI 只是画布的一部分——笔、选区、图层、蒙版不顺手的工具，实时也救不了。Every icon below is a real button in the app. AI is only part of the canvas: if the pen, selection, layers and masks are clumsy, real-time cannot save it.
笔画与画布Strokes and canvas
选择Select
点选或框选，移动、缩放、组合画布对象。Tap or marquee to select, then move, resize or group objects.
铅笔Pencil
细线勾轮廓，粗细 1–96、透明度随时调。Sketch outlines, with size from 1 to 96 and live opacity.
涂色Brush
铺色块，颜色、粗细、透明度各自设置。Paint colour areas with their own colour, size and opacity.
擦除Eraser
擦掉笔画或图片上的内容，不碰 AI 成图。Erase strokes or image content; the AI result is untouched.
局部Mask
只把选中的一块交给模型重绘，不用重画整张。Hand only that region to the model instead of redrawing everything.
图片Image
导入一张图片，当作可移动、可缩放的画布元素。Import an image as a movable, resizable canvas element.
全屏画布Fullscreen
一键全屏，换更大的绘制与选择空间。One tap for a much larger drawing and selection area.
调色Color
亮度、对比度、饱和度、色相、梦幻辉光、清晰度，实时调整成图。Brightness, contrast, saturation, hue, dream glow and sharpness, tuned live on the result.
撤销与重做Undo and redo
上一步改错了，退回来再试。Step back after a wrong move, then try again.
清空草稿Clear sketch
删掉全部画布元素，已经生成的成图保留。Remove every canvas element while keeping the generated result.
生成Generation
快速Fast
用实时模型按当前画面出一张预览，通常一秒之内。A preview from the live model, normally inside a second.
自动Auto
落笔停下就自动调用快速模型，边画边刷新。Generates automatically once you stop drawing, refreshing as you work.
随机Roll
换一个随机种子立刻再出一版，用来找灵感。Roll a new seed and generate again straight away, looking for ideas.
渲染Render
把当前画布交给高质量模型，出 1024×1024 大图。Sends the visible canvas to the high-quality model for a 1024×1024 image.
叠加生成Overlay
成图移到元素下方，并参与下一次参考。Moves the result below your elements and includes it in the next reference.
快照Snapshot
把当前显示画面压平成一张可编辑图片，于是可以直接在成图上继续画。Flattens what you see into one editable image, so you can keep drawing on top of the result.
下载Download
保存画布实际显示效果；渲染预览里另有一键单独保存大图。Saves the canvas exactly as shown; the render preview has its own button for the big image.
选中之后With a selection
成组与解组Group and ungroup
把几个对象绑在一起整体操作，需要时再解开。Bind several objects into one unit, then release them when needed.
放大与缩小Scale up and down
按比例缩放所选对象，也可以双指捏合和拖动。Scale the selection proportionally, or pinch and drag with two fingers.
上移与下移一层Layer order
调整单个对象的叠放次序。Reorder how a single object stacks against the rest.
复制与删除Duplicate and delete
复制所选对象并轻微错开位置，不要的随手删掉。Duplicate the selection with a small offset, or remove what you do not need.
参数都摆在明面上Every parameter in the open
尺寸Size256–2048
步数Steps1–150
绘制稿保留强度Sketch preservation0–100%
随机种子与锁定Seed and lock0–2^53
笔刷等待Brush wait400–5000 ms
请求超时Timeout5–300 s
成图与元素层透明度Layer opacity0–100%
调色Color adjustments6 项channels
模型返回的图片会校验真实尺寸，512 的结果不会被放大冒充高清。也可以先只接实时槽位，等画满意了再配渲染。Returned images are checked at their real dimensions, so a 512 result is never upscaled and passed off as high quality. You can start with only the live slot and add rendering later.
模型接入Connecting models
本地、云端，还是自己的协议。Local, cloud, or your own protocol.
HamDraw 不自带 API Key，也不绑定任何平台。它按你的选择提交请求，模型和工作流仍然在你手上。HamDraw ships no API key and locks you to no platform. It submits requests to whatever you configure, and the models and workflows stay yours.
本地 ComfyUI + HamDraw 插件节点Local ComfyUI with HamDraw nodes
装上 HamDraw 专用插件后，在任意工作流里放一个 HamDraw Input 和一个 HamDraw Output：提示词、参考图、蒙版、种子、强度、步数、尺寸接进你自己的采样或重绘节点，最终图片接回输出节点。应用把 API 格式的工作流和当前输入一起提交，再直接从 ComfyUI 读回结果图。With the dedicated plugin installed, drop one HamDraw Input and one HamDraw Output into any workflow: wire prompt, reference, mask, seed, strength, steps and size into your own sampler or inpainting nodes, and route the final image back. The app submits the API-format workflow with the current inputs and reads the result straight from ComfyUI.
不想在本地跑模型，就把实时和渲染都交给云端。手机只负责画和看，电脑可以关着。If you would rather not run models locally, hand both slots to the cloud. The phone only draws and displays; your computer can stay off.
Flux.2 4BQwenImage 2512DreamShaper8 LCM
局部重绘：只改你想改的那一块Local redraw: only the part you mark
先点「局部」把要改的区域涂红，再用键盘图标写下这块要改成什么，然后点「快速」。提交的是这张画布、那一块遮罩和那一句描述，画面其余部分原样保留。局部重绘有自己独立的描述框，和整幅画的提示词互不混用。Tap Mask and paint the region red, use the keyboard icon to write what it should become, then tap Fast. The canvas, that mask and that one sentence are what get submitted; everything else stays as it was. Local redraw keeps its own description field and never mixes with the whole-canvas prompt.
涂红区域Mark region写这一句Describe it只重绘这一块Redraw that part
通用兼容层Compatible servers
OpenAI Images 兼容接口、Stable Diffusion WebUI / Forge、Stability AI v2beta 都能直接接。可信局域网允许明文 HTTP，公网服务必须 HTTPS。OpenAI Images compatible endpoints, Stable Diffusion WebUI and Forge, and Stability AI v2beta all connect directly. Plain HTTP is allowed on a trusted LAN; public services must use HTTPS.
实时跑 DreamShaper8 LCM，渲染用 Flux.2 Klein 4B distilled 的 2 步。速度和清晰度都在线。Run DreamShaper8 LCM live and Flux.2 Klein 4B distilled at 2 steps to render. Speed and clarity both hold up.
DreamShaper8 LCMFlux.2 Klein 4B
12 GB
实时同上，渲染可以上到 4 步，需要更细的细节时切 8 步的 base 版本。Same live slot, and rendering can go to 4 steps, switching to the base model at 8 steps when detail matters.
DreamShaper8 LCMFlux.2 Klein base
16 GB+
把 QwenImage 2512 接进渲染槽位，拿原生 1024×1024 的构图服从度和质感。Put QwenImage 2512 on the render slot for native 1024×1024 composition accuracy and texture.
QwenImage 2512Flux.2 4B
基于 Haminn，所以能改Built on Haminn, so it can be changed
整个应用，你都能换成自己的。Change anything. The whole app.
HamDraw 是跑在你自己的 Haminn 里的 MIT 开源应用。在 Haminn 的开发页开启开发模式，把地址和密码交给你信任的智能体，画笔参数、界面配色、工具排布、生成流程，都可以按你的习惯重做。HamDraw is an MIT-licensed app running inside your own Haminn. Turn on development mode from the Haminn development page, hand the address and password to an agent you trust, and brush parameters, colours, tool layout and generation flow can all be rebuilt around your habits.
WorkBuddy桌面智能体Desktop agent或orCodex命令行智能体CLI agent任何支持 MCP 的智能体都可以any MCP-capable agent works
第一步：装 HaminnStep one: install Haminn
HamDraw 跑在 Haminn 里，先把它装上。HamDraw runs inside Haminn. Install that first.
Haminn 是手机上装应用的地方，负责安装、数据、文件、权限和系统能力。装好 Haminn，再把 HamDraw 加进去。不接模型也能先画——笔、图层、蒙版和导出都能单独用。Haminn is where apps live on your phone: it owns installation, data, files, permissions and system capabilities. Install Haminn, then add HamDraw. It draws before any model is connected: pen, layers, masks and export all work on their own.
插件是 zip 直下，解压到 ComfyUI 的 custom_nodes/ 即可；习惯用 Git 的话也可以从仓库目录取。The plugin downloads as a zip: unzip it into ComfyUI's custom_nodes/. If you prefer Git, take the same folder from the repository.
同一位作者的上一个作品The previous one from the same workshop
画完了，找几个 AI 评一评。Done drawing? Let a room of AIs argue about it.
chataxi 是同一个 Haminn 里的多角色 AI 群聊：每个角色有自己的个性、模型和声音，可以让它们一起讨论你的设定、命名和文案。chataxi is the multi-role AI group chat for the same Haminn: each role has its own personality, model and voice, ready to argue over your settings, names and copy.
本地跑才有这个要求，而且实时槽位用的是 512×512 的小模型，门槛不高。不想装环境就用云端服务，手机照样画。Only for local models, and the live slot runs a small 512×512 model, so the bar is low. If you would rather not set anything up, a cloud service leaves the phone doing all the drawing.
出的图不是我想要的怎么办？What if the result is not what I meant?
先调“绘制稿保留强度”：调高更忠于你的线稿，调低更多交给模型发挥。再锁定种子复现某一版，或者用局部蒙版只改一块。颜色如果是靠细线表达的，模型很难判断那是表面色还是标注线，这时候把颜色写进提示词更稳。Start with sketch preservation: higher stays truer to your drawing, lower gives the model more room. Then lock a seed to reproduce a frame, or mask a region to change only that part. When colour is carried by thin lines the model cannot tell a surface from an annotation, so naming the colour in the prompt is more reliable.
本地 ComfyUI + HamDraw 插件节点Local ComfyUI with HamDraw nodes
