UI→API conversion sends LOAD_3D viewport_state as "" where the frontend sends the viewport camera

未关闭
#1,126 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
53/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
python
领域
cli

调研方向

Start with convert_ui_to_api and trace how it serializes the LOAD_3D viewport_state widget; compare the result with the frontend’s graphToPrompt output described in the issue. Check the bundled templates 3d_pixal3d_multi_views and 3d_pixal3d_trellis2_image_to_model, then verify that conversion preserves the viewport state needed by execute_save_3d_advanced in comfy_extras/nodes_save_3d.py. Done means the converted prompt provides the expected camera information, including for downstream wiring.

由索引模型根据 Issue 内容生成。

描述

UI→API conversion sends viewport_state as the saved "", while the frontend's graphToPrompt sends the live 3D viewport state (camera, model info). Found by converting the bundled 3D templates with convert_ui_to_api (the comfy run --workflow path) and comparing each prompt with the frontend's app.graphToPrompt().output for the same graph.

Versions: comfy-cli 54732c6 (main), ComfyUI v0.44.2 on CPU, comfyui-frontend-package 1.57.6, comfyui-workflow-templates 0.11.103.

Where: Preview3DAdvanced and Save3DAdvanced (input viewport_state, type LOAD_3D, required). Templates 3d_pixal3d_multi_views (nodes 246, 323, 372) and 3d_pixal3d_trellis2_image_to_model (nodes 246, 322, 323). The saved node has widgets_values: ["", 1024, 1024].

  • comfy-cli: "viewport_state": ""
  • frontend: "viewport_state": {"image": "", "mask": "", "normal": "", "camera_info": {"position": {"x": 10, "y": 10, "z": 10.000000000000002}, "target": {"x": 0, "y": 0, "z": 0}, "zoom": 1, "cameraType": "perspective", ...}, "recording": "", "model_3d_info": []}

Effect: execute_save_3d_advanced (comfy_extras/nodes_save_3d.py) turns a non-dict viewport_state into {}, so the node's camera_info output, and the camera in its Saved3DModels UI result, are None instead of the viewport camera. In these two templates nothing is wired to those outputs, so the generated model is unaffected; a graph that wires camera_info downstream would get None.

The frontend builds this value at serialisation time from the live viewport rather than from widgets_values, so matching it exactly may need a default camera on the converter side, or the server treating "" as "default camera". Related: #893 (widget types without a slot), #1077 (converter convergence).

主要语言
Python
星标
1k
派生
176
平均合并
1 天 11 小时
30 天内合并 PR
87

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

Comfy-Org/comfy-cli 的其他 Issue

查看 Comfy-Org/comfy-cli 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。