请问此处是否是依赖引入，还是文件引入（代码项目中却没有对应文件夹）

// App.tsx
import * as m from './paraglide/messages'
import {
  languageTag,
  onSetLanguageTag,
  setLanguageTag,
} from './paraglide/runtime'

The size of the brush is incorrect after inpainting once.

刷子在修复一个图像后，大小不对，期待有人修复。

chrome 浏览器

mac inter芯片

在文件 superResolution.ts 中 264行

    model = await ort.InferenceSession.create(modelBuffer, {
      executionProviders: [capabilities.webgpu ? 'webgpu' : 'wasm'],
    })

Error: Can't create a session. ERROR_CODE: 7, ERROR_MESSAGE: Failed to load model because protobuf parsing failed.

创建会话失败，onnx模型文件是正常的

file:///B:/AI/inpaint-web/node_modules/.store/@Inlang[email protected]/node_modules/@inlang/sdk/dist/loadProject.js:8
import { ProjectSettings, Message } from "./versionedInterfaces.js";
^^^^^^^
SyntaxError: The requested module './versionedInterfaces.js' contains conflicting star exports for name 'Message'
at ModuleJob._instantiate (node:internal/modules/esm/module_job:171:21)
at async ModuleJob.run (node:internal/modules/esm/module_job:254:5)
at async onImport.tracePromise.proto (node:internal/modules/esm/loader:485:26)
at async asyncRunEntryPointWithESMLoader (node:internal/modules/run_main:109:5)

上传图片后点击4倍高清黑屏了，效果如图

WebGpuBackend: Failed to get GPU adapter.

mac book pro intel 2019.
os version: the latest mac os

操作时提示如下

是我的chrome有什么问题么？

Hi, I found a few problems when viewing this app from China.

1. models download from huggingface is always failed, I tried to change the host of the download links of models to hf-mirror.com, but is still have some CORS problems. Maybe need to fallback to use the self-hosted models which could download using huggingface-cli.
2. the onnxruntime-web libs from jsdelivr could not download sometimes, I also tried unpkg, both of them are not stable. Maybe use self-hosted instead of CDN.

And I tried to run the app from github codespace, and I found some problems when running npm run start, the typescript version conflicts, some type errors related to ora (onnxruntime-web) and so on ...

您这个是否支持docker离线部署，docker build .有些问题，启动后产生了三个镜像，不支持离线，我看到模型有下载的

你的已经死了 无法下载

多次尝试始终无法跑完进度条：

as title

无论是本地端还是网页端，在执行和 demo 相同操作时，均无法实现 demo 的效果，返回的图都是纯黑色

使用刷子时报错如图。

When inpainting some region, then trying to edit another region, the region which was previously inpainted also gets modified. This behavior is likely undesirable in most cases and can be considered a bug. In other words, with every edit, mask should be composed from scratch and the image should be the latest edit, while the current approach seems to unite the currently created mask with all previous masks before passing it to the model.

下载下来的图片和原图一模一样

能否解决下这个问题，顺便兼容下手机浏览器？

请问是用的什么模型？是用onnx构建的嘛？我想自己试着调试下看能不能解决

add a progress bar for the image progress

在README文件里, Frontend code are modified from cleanup.pictures cleanup.picures的开源库没了

自从老哥放出这东西第一天开始，我就没能加载成功过那个30MB的模型......

如题。

部署完成后,发现4x-upscaling生成速度很慢,与demo演示站差距大

请问demo演示站服务器配置情况?

建议什么配置的服务器部署?

手动用笔刷去除字幕，操作起来需要小心翼翼，有点繁琐，希望能添加一键去除水印、字幕的功能。

最近修复后的图片没有任何改进，是项目关闭了吗？

CPU:AMD5500
显卡:AMD6650XT
系统:Win11
浏览器:Chrome
下载了30MB+70MB两个文件，然后上传了本地的“pic.jpg”，进度从0走到100%以后，没有进行处理，跟demo表现不一致。demo会在屏幕上方有一个timeline，方便回滚处理，但是，我这边网页上timeline是空的，屏幕上的图片看起来也没有任何处理

原图 3528*2232

修改后 1024*647