From 7cceea2578ac409a5b8d3746541eca5eac0a3758 Mon Sep 17 00:00:00 2001 From: Christian Byrne Date: Thu, 6 Aug 2026 15:31:06 -0700 Subject: [PATCH] docs(qwen-image-layered): document the layers widget and the layers + 1 output The tutorial had no section on the `layers` widget and said nothing about what a generation actually produces, so a reader has no way to know that the first decoded image is not a layer. Add a Layers and output section covering: - the model regenerates the full image as well as the layers, so a generation decodes to layers + 1 images - the first image is that full image and should be discarded when only the layers are wanted; layers=0 is valid and yields it on its own - compositing all the layers recreates the full image, which makes a usable correctness check - the layers ride the temporal axis of the 5D latent, LatentCutToBatch with dim=t moves it into the batch dimension, and a layer's z-order is therefore just its batch index; there is no z-index field and no per-layer metadata - the layered VAE decodes to 4 channels, so layers carry alpha zh/ja/ko regenerated with the repo's translate pipeline, then hand checked and repunctuated to match the surrounding pages. --- .../image/qwen/qwen-image-layered.mdx | 27 ++++++++++++++++++- .../image/qwen/qwen-image-layered.mdx | 27 ++++++++++++++++++- tutorials/image/qwen/qwen-image-layered.mdx | 24 +++++++++++++++++ .../image/qwen/qwen-image-layered.mdx | 27 ++++++++++++++++++- 4 files changed, 102 insertions(+), 3 deletions(-) diff --git a/ja/tutorials/image/qwen/qwen-image-layered.mdx b/ja/tutorials/image/qwen/qwen-image-layered.mdx index b2c05499a..0b2751636 100644 --- a/ja/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ja/tutorials/image/qwen/qwen-image-layered.mdx @@ -2,13 +2,14 @@ title: "Qwen-Image-Layered ComfyUI ワークフロー例" description: "Qwen-Image-Layered は、画像を複数の RGBA レイヤーに分解できるモデルであり、レイヤー分解を通じて本質的な編集可能性を実現します。" sidebarTitle: "Qwen-Image-Layered" -translationSourceHash: bf2f2000 +translationSourceHash: dba6b09a translationFrom: tutorials/image/qwen/qwen-image-layered.mdx translationBlockHashes: "_intro": 19900234 "Qwen-Image-Layered workflow": ba937275 "Model links": 98d12555 "FP8 version": 6bffdd17 + "Layers and output": 2419fd13 "Workflow settings": b0f81aa2 --- @@ -75,6 +76,30 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' その後、[サブグラフ](/ja/interface/features/subgraph)内の **Load Diffusion model** ノードを更新し、このファイルを使用するように設定してください。 +## レイヤーと出力 + +レイヤーの数は、[サブグラフ](/ja/interface/features/subgraph)内の **Empty Qwen Image Layered Latent** ノードにある `layers` ウィジェットで設定します。 + +このモデルはレイヤーだけでなく完全な画像も生成するため、1 回の生成は常に `layers + 1` 枚の画像にデコードされます： + +| `layers` | デコードされる画像数 | 内容 | +| --- | --- | --- | +| `0` | 1 | 完全な画像のみ | +| `2`（テンプレートの既定値） | 3 | 完全な画像、続いて 2 枚のレイヤー | +| `5` | 6 | 完全な画像、続いて 5 枚のレイヤー | + +**最初の画像は完全な画像であり、レイヤーではありません。** 画像からレイヤーへのワークフローでは入力画像を再現したものになるため、レイヤーだけが必要な場合は破棄してください。`layers` には `0` を指定でき、その場合はこの画像だけが得られます。 + +**すべてのレイヤーを合成し直すと完全な画像が再現されます。** これは確認に使えます。レイヤーを重ねても最初の画像が再現されない場合、分解は意図したとおりに機能していません。 + +### レイヤーが個別の画像になる仕組み + +**Empty Qwen Image Layered Latent** ノードは `[batch_size, 16, layers + 1, height // 8, width // 8]` という形状の 5D Latent を確保します。レイヤーは 3 番目の軸に並び、これは映像モデルがフレームを置くのと同じ時間軸です。**LatentCutToBatch** ノードの `dim` を `t` に設定するとその軸がバッチ次元へ移されるため、**VAE Decode** の後は各レイヤーがバッチ内の個別の画像になります。 + +したがって、レイヤーの重なり順はバッチ内のインデックスそのものであり、インデックス 0 の完全な画像から数えます。z-index フィールドも、その順序を保持するレイヤーごとのメタデータも存在しないため、バッチを並べ替えたりフィルタリングしたりすることは、そのままレイヤーの並べ替えになります。 + +Qwen-Image-Layered の VAE は 4 チャンネルにデコードするため、デコードされたレイヤーはアルファチャンネルを持ちます。 + ## ワークフローの設定 ### サンプラー設定 diff --git a/ko/tutorials/image/qwen/qwen-image-layered.mdx b/ko/tutorials/image/qwen/qwen-image-layered.mdx index 51aae2791..4c2a8703b 100644 --- a/ko/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ko/tutorials/image/qwen/qwen-image-layered.mdx @@ -2,13 +2,14 @@ title: "Qwen-Image-Layered ComfyUI 워크플로우 예시" description: "Qwen-Image-Layered는 이미지를 여러 개의 RGBA 레이어로 분해할 수 있는 모델로, 레이어 분해를 통해 본질적인 편집 가능성을 제공합니다." sidebarTitle: "Qwen-Image-Layered" -translationSourceHash: bf2f2000 +translationSourceHash: dba6b09a translationFrom: tutorials/image/qwen/qwen-image-layered.mdx translationBlockHashes: "_intro": 19900234 "Qwen-Image-Layered workflow": ba937275 "Model links": 98d12555 "FP8 version": 6bffdd17 + "Layers and output": 2419fd13 "Workflow settings": b0f81aa2 --- @@ -75,6 +76,30 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 그런 다음 [서브그래프](/ko/interface/features/subgraph) 내의 **Load Diffusion model** 노드를 업데이트하여 이를 사용하세요. +## 레이어와 출력 + +레이어 수는 [서브그래프](/ko/interface/features/subgraph) 내의 **Empty Qwen Image Layered Latent** 노드에 있는 `layers` 위젯으로 설정합니다. + +이 모델은 레이어와 함께 전체 이미지도 다시 생성하므로, 한 번의 생성은 항상 `layers + 1`개의 이미지로 디코딩됩니다. + +| `layers` | 디코딩되는 이미지 수 | 내용 | +| --- | --- | --- | +| `0` | 1 | 전체 이미지만 | +| `2` (템플릿 기본값) | 3 | 전체 이미지, 그다음 레이어 2개 | +| `5` | 6 | 전체 이미지, 그다음 레이어 5개 | + +**첫 번째 이미지는 전체 이미지이며 레이어가 아닙니다.** 이미지를 레이어로 분해하는 워크플로우에서는 입력 이미지를 재현한 것이므로, 레이어만 필요하다면 이 이미지는 버리세요. `layers`에는 `0`을 지정할 수 있으며, 이 경우 해당 이미지 하나만 얻게 됩니다. + +**모든 레이어를 다시 합성하면 전체 이미지가 재현됩니다.** 이를 검증에 사용하세요. 레이어를 쌓아도 첫 번째 이미지가 재현되지 않는다면 분해가 의도대로 이루어지지 않은 것입니다. + +### 레이어가 개별 이미지가 되는 방식 + +**Empty Qwen Image Layered Latent** 노드는 `[batch_size, 16, layers + 1, height // 8, width // 8]` 형태의 5D Latent를 할당합니다. 레이어는 세 번째 축에 놓이며, 이는 비디오 모델이 프레임을 두는 것과 같은 시간 축입니다. **LatentCutToBatch** 노드의 `dim`을 `t`로 설정하면 그 축이 배치 차원으로 옮겨지므로, **VAE Decode** 이후 각 레이어는 배치 안의 개별 이미지가 됩니다. + +따라서 레이어의 쌓임 순서는 배치 안에서의 인덱스 그 자체이며, 인덱스 0의 전체 이미지부터 셉니다. z-index 필드도, 그 순서를 담는 레이어별 메타데이터도 없으므로 배치를 재정렬하거나 걸러내는 것은 곧 레이어를 재정렬하는 것입니다. + +Qwen-Image-Layered VAE는 4채널로 디코딩하므로 디코딩된 레이어는 알파 채널을 가집니다. + ## 워크플로우 설정 ### 샘플러 설정 diff --git a/tutorials/image/qwen/qwen-image-layered.mdx b/tutorials/image/qwen/qwen-image-layered.mdx index 05275ee64..c929bf46b 100644 --- a/tutorials/image/qwen/qwen-image-layered.mdx +++ b/tutorials/image/qwen/qwen-image-layered.mdx @@ -66,6 +66,30 @@ By default we are using bf16, which requires high VRAM. For lower VRAM usage, yo Then update the **Load Diffusion model** node inside the [Subgraph](/interface/features/subgraph) to use it. +## Layers and output + +The number of layers is set by the `layers` widget on the **Empty Qwen Image Layered Latent** node inside the [Subgraph](/interface/features/subgraph). + +This model regenerates the full image as well as the layers, so a generation always decodes to `layers + 1` images: + +| `layers` | Images decoded | Contents | +| --- | --- | --- | +| `0` | 1 | The full image only | +| `2` (template default) | 3 | The full image, then 2 layers | +| `5` | 6 | The full image, then 5 layers | + +**The first image is the full image, not a layer.** In the image to layers workflow it reproduces the input image, so discard it when you only want the layers. `layers` accepts `0`, which gives you that image on its own. + +**Compositing all of the layers back together recreates the full image.** Use it as a check: if stacking the layers does not reproduce the first image, the decomposition did not do what you wanted. + +### How the layers become separate images + +The **Empty Qwen Image Layered Latent** node allocates a 5D latent shaped `[batch_size, 16, layers + 1, height // 8, width // 8]`. The layers sit on the third axis, the same temporal axis that video models use for frames. The **LatentCutToBatch** node with `dim` set to `t` moves that axis into the batch dimension, so after **VAE Decode** each layer is a separate image in the batch. + +A layer's position in the stack is therefore its index in the batch, counting from the full image at index 0. There is no z-index field and no per-layer metadata carrying that order, so reordering or filtering the batch reorders the layers. + +The Qwen-Image-Layered VAE decodes to 4 channels, so the decoded layers carry an alpha channel. + ## Workflow settings ### Sampler settings diff --git a/zh/tutorials/image/qwen/qwen-image-layered.mdx b/zh/tutorials/image/qwen/qwen-image-layered.mdx index 49f685ab0..50eb9e28a 100644 --- a/zh/tutorials/image/qwen/qwen-image-layered.mdx +++ b/zh/tutorials/image/qwen/qwen-image-layered.mdx @@ -2,13 +2,14 @@ title: "Qwen-Image-Layered ComfyUI 工作流示例" description: "Qwen-Image-Layered 是一个能够将图像分解为多个 RGBA 图层的模型，通过图层分解实现固有的可编辑性。" sidebarTitle: "Qwen-Image-Layered" -translationSourceHash: bf2f2000 +translationSourceHash: dba6b09a translationFrom: tutorials/image/qwen/qwen-image-layered.mdx translationBlockHashes: "_intro": 19900234 "Qwen-Image-Layered workflow": ba937275 "Model links": 98d12555 "FP8 version": 6bffdd17 + "Layers and output": 2419fd13 "Workflow settings": b0f81aa2 --- @@ -75,6 +76,30 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 然后更新[子图](/zh/interface/features/subgraph)中的 **Load Diffusion model** 节点来使用它。 +## 图层和输出 + +图层数量由[子图](/zh/interface/features/subgraph)内 **Empty Qwen Image Layered Latent** 节点上的 `layers` 组件设置。 + +该模型会在生成图层的同时重新生成完整图像，因此一次生成总是解码出 `layers + 1` 张图像： + +| `layers` | 解码出的图像数 | 内容 | +| --- | --- | --- | +| `0` | 1 | 仅完整图像 | +| `2`（模板默认值） | 3 | 完整图像，然后是 2 个图层 | +| `5` | 6 | 完整图像，然后是 5 个图层 | + +**第一张图像是完整图像，而不是图层。** 在图像转图层的工作流中，它重现的是输入图像，因此当你只需要图层时应将其丢弃。`layers` 可以设为 `0`，此时只会得到这一张图像。 + +**将所有图层合成回去可以重建完整图像。** 可以用它来检查：如果堆叠图层无法重现第一张图像，说明分解没有达到预期效果。 + +### 图层如何变成单独的图像 + +**Empty Qwen Image Layered Latent** 节点会分配一个形状为 `[batch_size, 16, layers + 1, height // 8, width // 8]` 的 5D Latent。图层位于第三个轴上，也就是视频模型用于存放帧的那个时间轴。将 **LatentCutToBatch** 节点的 `dim` 设置为 `t`，会把该轴移动到批次维度上，因此在 **VAE Decode** 之后，每个图层都是批次中一张单独的图像。 + +因此，图层在堆栈中的位置就是它在批次中的索引，从索引 0 处的完整图像开始计数。这里没有 z-index 字段，也没有承载该顺序的逐图层元数据，因此重新排序或过滤批次就等于重新排序图层。 + +Qwen-Image-Layered VAE 解码为 4 个通道，因此解码出的图层带有 Alpha 通道。 + ## 工作流设置 ### 采样器设置