[PR #185] [MERGED] docs: add GetLatentSize, GetAudioLatentSize, Flux3LayoutPrompt, Krea3ImageNode, MOSS-Image Layout Prompt (+ CLIPLoader flux3/moss_image, Quiver Arrow 1.x removal) #189

New issue

Closed

opened 2026-10-26 22:10:27 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 22:10:27 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/embedded-docs/pull/185
Author: @lin-bot23
Created: 10/15/2026
Status: ✅ Merged
Merged: 10/17/2026
Merged by: @comfyui-wiki

Base: main ← Head: daily-docs-update-1016

📝 Commits (5)

6cc501f docs: add GetLatentSize, GetAudioLatentSize, Flux3LayoutPrompt and Krea3ImageNode
d093051 docs: add 14-language translations for the four new nodes
8ae39e5 docs: cover both Krea 3 workflows in the ja summary and lengthen the zh Layout Prompt summary
f54d961 docs: add MOSS-Image Layout Prompt, sync Arrow 1.x removal from Quiver and MOSS/empty-latent changes
12c4107 docs: add 14-language translations and refresh localized columns

📊 Changes

165 files changed (+2320 additions, -355 deletions)

View changed files

📝 comfyui_embedded_docs/docs/ClipLoader/ar.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/de.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/en.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/es.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/fa.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/fr.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/he.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/it.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/ja.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/ko.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/pt-BR.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/ru.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/tr.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/zh-TW.md (+3 -2)
📝 comfyui_embedded_docs/docs/ClipLoader/zh.md (+3 -2)
➕ comfyui_embedded_docs/docs/Flux3LayoutPrompt/ar.md (+21 -0)
➕ comfyui_embedded_docs/docs/Flux3LayoutPrompt/de.md (+21 -0)
➕ comfyui_embedded_docs/docs/Flux3LayoutPrompt/en.md (+21 -0)
➕ comfyui_embedded_docs/docs/Flux3LayoutPrompt/es.md (+21 -0)
➕ comfyui_embedded_docs/docs/Flux3LayoutPrompt/fa.md (+21 -0)

...and 80 more files

📄 Description

Summary

Daily built-in node documentation update (agent-written English, pipeline translations).

New nodes (4)

Node	Display name	Source
`GetLatentSize`	Get Image/Video Latent Size	ComfyUI #16093
`GetAudioLatentSize`	Get Audio Latent Size	ComfyUI #16093
`Flux3LayoutPrompt`	FLUX 3 Image Layout Prompt	ComfyUI #17064
`Krea3ImageNode`	Krea 3 Image	ComfyUI #17067

Each ships en.md plus 14 translations (zh, zh-TW, es, fr, ja, ko, ru, ar, tr, pt-BR, fa, de, he, it).

Source-driven updates to existing pages

ClipLoader: the type combo gained flux3 (qwen3-vl 4B encoder), so the option was appended to the type range in all 15 language files.
ReferenceMultiLatent: the vae tooltip changed to "Reference images are ignored when no VAE is connected." The description no longer claims the node raises an error; the sentence was rewritten in all 15 language files.

Reviewed and intentionally skipped

The weekly scan flagged 71 changed nodes; the class-level diffs were reviewed one by one (review_changed_node_diffs.py plus manual git diff on each node class). Changes that do not reach the documentation were left alone:

Flux3ImageNode: display label renamed to "FLUX 3 Image" (ComfyUI #17075). The page title comes from the node directory, so this is invisible in the published page; prose is unchanged.
CreateBoundingBoxes, Preview3DAdvanced, PreviewGaussianSplat, PreviewPointCloud, SaveGLB, SaveImageAdvanced, ByteDanceSeedAudio: execution/UI-layer changes (asset previews, UI.SavedResult, canvas background image) with no parameter or output changes.
RecraftV4TextToVectorNode and the other Recraft nodes: price-badge updates only; no page in that family documents prices.
ModelAttentionBackend: search_aliases only.

Verification

fix_translation_preamble.py --check-only --changed-only: 0 leaked lines
check_doc_structure.py --changed-only: 0 defects
check_md_format.py --changed-only: 0 errors
check_md_links.py --changed-only: all links valid
check_new_node_translations.py / check_translation_language.py: clean (only a false positive on the Italian overview, which is Italian text)
Range/Required cells localized for every language, including unit words (0 到 8 张图像, 0~8개 이미지, от 0 до 8 изображений, ...)
Source fingerprint footers refreshed for both changed nodes and present on all four new pages

Files

comfyui_embedded_docs/docs/GetLatentSize/, GetAudioLatentSize/, Flux3LayoutPrompt/, Krea3ImageNode/ (new, 15 files each)
comfyui_embedded_docs/docs/ClipLoader/, ReferenceMultiLatent/ (15 files each, one line + footer per file)

Follow-up (2026-10-17)

Same branch carried into the next daily run, appending two commits (English first, then translations).

New node

Node	Display name	Source
`MOSSImageLayoutPrompt`	MOSS-Image Layout Prompt	ComfyUI #17095 (MOSS-image support)

Builds the JSON layout prompt that MOSS-Image reads from the boxes drawn in Create Bounding Boxes: one entry per box with its label (object / character / text), caption, optional palette and rectangle converted from the 0-1000 normalized coordinates into the pixel grid set by width / height. Both sizes step by 16 and default to 1024; character_boxes lists the box numbers to label as characters. With no connected elements the prompt text is passed through unchanged. en.md plus 14 translations.

Source-driven updates to existing pages

QuiverTextToSVGNode, QuiverImageToSVGNode, QuiverTextToSVGNodeV2, QuiverImageToSVGNodeV2 (ComfyUI #17102): the retired Arrow 1.x models were removed from the source, so "arrow-1.1", "arrow-1.1-max" and "arrow-preview" were dropped from the model range in all 60 files and the stale "retires on October 16, 2026" note was removed (the two legacy pages keep the "marked as deprecated in the source code" sentence). QuiverTextToSVGNodeV2 also caps reference_images at 14 for every remaining model, so the second range value and the note were reduced, including the overview sentence that still claimed the cap depends on the selected model.
ClipLoader (ComfyUI #17095): the type combo also gained moss_image, and the missing flux3 | qwen3-vl 4B row was added to the type-to-encoder mapping table in all 15 files.
ReferenceMultiLatent (ComfyUI #17053): empty_latent is no longer None when nothing is encoded. The output description and the closing note now say that without a connected VAE (or without images) a square empty latent is returned, 1024 x 1024 or the size derived from megapixels when resolution is "resize", in all 15 files.
update_param_translations.py also localized parameter names that had drifted from the frontend term table inside the touched node directories (for example prompt -> Prompt, reference_images -> Referenzbilder in the German Quiver pages, image -> immagine in Italian). Its library-wide footprint (1676 files across de/he/it/es, mostly the upstream batch translation debt) was reverted and stays out of this PR.

Reviewed and intentionally skipped (this round)

Flux3ImageNode / Flux3LayoutPrompt: display labels renamed to "FLUX 3 Image" (ComfyUI #17075). The published page title comes from the node directory and the frontend still shows "Flux 3", so the headings were left alone.
SaveImageAdvanced (ComfyUI #16867): EXR is converted to scene-linear before encoding and asset previews are written for the asset manager, both execution-layer changes; the documented parameters, ranges and behaviour notes are unchanged.
CreateBoundingBoxes, Preview3DAdvanced, PreviewGaussianSplat, PreviewPointCloud, SaveGLB, ByteDanceSeedAudio, ModelAttentionBackend, RecraftV4TextToVectorNode: unchanged from the analysis above.

Verification (follow-up commits)

fix_translation_preamble.py --check-only --changed-only --base origin/main: 0 files, 0 leaked lines
check_doc_structure.py --changed-only: 0 defects
check_md_format.py --changed-only: 0 errors (150 files)
check_md_links.py --changed-only: all local resource links valid
check_new_node_translations.py MOSSImageLayoutPrompt: clean; check_translation_language.py: 0 problems
Table row counts for MOSSImageLayoutPrompt match en.md in all 15 files; Range cells localized (16 到 16384（步长 16）, 16〜16384 （ステップ 16）, 16 bis 16384 (Schritt: 16), ...)
Remaining arrow-1.1 / arrow-preview / Arrow 1 mentions: 0 across the four Quiver pages
Source fingerprints recomputed with lib.node_source_extract for ClipLoader (3abcb39d) and the four Quiver nodes (9c85e3be, 196b0693, a7618459, 3531975f); ReferenceMultiLatent already carried the current hash (06f98e7c)

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/Comfy-Org/embedded-docs/pull/185 **Author:** [@lin-bot23](https://github.com/lin-bot23) **Created:** 10/15/2026 **Status:** ✅ Merged **Merged:** 10/17/2026 **Merged by:** [@comfyui-wiki](https://github.com/comfyui-wiki) **Base:** `main` ← **Head:** `daily-docs-update-1016` --- ### 📝 Commits (5) - [`6cc501f`](https://github.com/Comfy-Org/embedded-docs/commit/6cc501f9c1efc27d67bdb33abd3c96523a1bed5a) docs: add GetLatentSize, GetAudioLatentSize, Flux3LayoutPrompt and Krea3ImageNode - [`d093051`](https://github.com/Comfy-Org/embedded-docs/commit/d0930513cd64a2bc20a0cffcf0c2731db099c280) docs: add 14-language translations for the four new nodes - [`8ae39e5`](https://github.com/Comfy-Org/embedded-docs/commit/8ae39e584507ae748ea682beffc0ef2eb1460b5c) docs: cover both Krea 3 workflows in the ja summary and lengthen the zh Layout Prompt summary - [`f54d961`](https://github.com/Comfy-Org/embedded-docs/commit/f54d961f4e7d1c1598bff5d8c69c19d45b8b9dea) docs: add MOSS-Image Layout Prompt, sync Arrow 1.x removal from Quiver and MOSS/empty-latent changes - [`12c4107`](https://github.com/Comfy-Org/embedded-docs/commit/12c41073bb37fe4b3485d9877a2c927b635e3f39) docs: add 14-language translations and refresh localized columns ### 📊 Changes **165 files changed** (+2320 additions, -355 deletions) <details> <summary>View changed files</summary> 📝 `comfyui_embedded_docs/docs/ClipLoader/ar.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/de.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/en.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/es.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/fa.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/fr.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/he.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/it.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/ja.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/ko.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/pt-BR.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/ru.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/tr.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/zh-TW.md` (+3 -2) 📝 `comfyui_embedded_docs/docs/ClipLoader/zh.md` (+3 -2) ➕ `comfyui_embedded_docs/docs/Flux3LayoutPrompt/ar.md` (+21 -0) ➕ `comfyui_embedded_docs/docs/Flux3LayoutPrompt/de.md` (+21 -0) ➕ `comfyui_embedded_docs/docs/Flux3LayoutPrompt/en.md` (+21 -0) ➕ `comfyui_embedded_docs/docs/Flux3LayoutPrompt/es.md` (+21 -0) ➕ `comfyui_embedded_docs/docs/Flux3LayoutPrompt/fa.md` (+21 -0) _...and 80 more files_ </details> ### 📄 Description ## Summary Daily built-in node documentation update (agent-written English, pipeline translations). ### New nodes (4) | Node | Display name | Source | | --- | --- | --- | | `GetLatentSize` | Get Image/Video Latent Size | ComfyUI #16093 | | `GetAudioLatentSize` | Get Audio Latent Size | ComfyUI #16093 | | `Flux3LayoutPrompt` | FLUX 3 Image Layout Prompt | ComfyUI #17064 | | `Krea3ImageNode` | Krea 3 Image | ComfyUI #17067 | Each ships `en.md` plus 14 translations (`zh`, `zh-TW`, `es`, `fr`, `ja`, `ko`, `ru`, `ar`, `tr`, `pt-BR`, `fa`, `de`, `he`, `it`). ### Source-driven updates to existing pages - `ClipLoader`: the `type` combo gained `flux3` (qwen3-vl 4B encoder), so the option was appended to the `type` range in all 15 language files. - `ReferenceMultiLatent`: the `vae` tooltip changed to "Reference images are ignored when no VAE is connected." The description no longer claims the node raises an error; the sentence was rewritten in all 15 language files. ### Reviewed and intentionally skipped The weekly scan flagged 71 changed nodes; the class-level diffs were reviewed one by one (`review_changed_node_diffs.py` plus manual `git diff` on each node class). Changes that do not reach the documentation were left alone: - `Flux3ImageNode`: display label renamed to "FLUX 3 Image" (ComfyUI #17075). The page title comes from the node directory, so this is invisible in the published page; prose is unchanged. - `CreateBoundingBoxes`, `Preview3DAdvanced`, `PreviewGaussianSplat`, `PreviewPointCloud`, `SaveGLB`, `SaveImageAdvanced`, `ByteDanceSeedAudio`: execution/UI-layer changes (asset previews, `UI.SavedResult`, canvas background image) with no parameter or output changes. - `RecraftV4TextToVectorNode` and the other Recraft nodes: price-badge updates only; no page in that family documents prices. - `ModelAttentionBackend`: `search_aliases` only. ### Verification - `fix_translation_preamble.py --check-only --changed-only`: 0 leaked lines - `check_doc_structure.py --changed-only`: 0 defects - `check_md_format.py --changed-only`: 0 errors - `check_md_links.py --changed-only`: all links valid - `check_new_node_translations.py` / `check_translation_language.py`: clean (only a false positive on the Italian overview, which is Italian text) - Range/Required cells localized for every language, including unit words (`0 到 8 张图像`, `0~8개 이미지`, `от 0 до 8 изображений`, ...) - Source fingerprint footers refreshed for both changed nodes and present on all four new pages ## Files - `comfyui_embedded_docs/docs/GetLatentSize/`, `GetAudioLatentSize/`, `Flux3LayoutPrompt/`, `Krea3ImageNode/` (new, 15 files each) - `comfyui_embedded_docs/docs/ClipLoader/`, `ReferenceMultiLatent/` (15 files each, one line + footer per file) ## Follow-up (2026-10-17) Same branch carried into the next daily run, appending two commits (English first, then translations). ### New node | Node | Display name | Source | | --- | --- | --- | | `MOSSImageLayoutPrompt` | MOSS-Image Layout Prompt | ComfyUI #17095 (MOSS-image support) | Builds the JSON layout prompt that MOSS-Image reads from the boxes drawn in Create Bounding Boxes: one entry per box with its label (object / character / text), caption, optional palette and rectangle converted from the 0-1000 normalized coordinates into the pixel grid set by `width` / `height`. Both sizes step by 16 and default to 1024; `character_boxes` lists the box numbers to label as characters. With no connected elements the prompt text is passed through unchanged. `en.md` plus 14 translations. ### Source-driven updates to existing pages - `QuiverTextToSVGNode`, `QuiverImageToSVGNode`, `QuiverTextToSVGNodeV2`, `QuiverImageToSVGNodeV2` (ComfyUI #17102): the retired Arrow 1.x models were removed from the source, so `"arrow-1.1"`, `"arrow-1.1-max"` and `"arrow-preview"` were dropped from the `model` range in all 60 files and the stale "retires on October 16, 2026" note was removed (the two legacy pages keep the "marked as deprecated in the source code" sentence). `QuiverTextToSVGNodeV2` also caps `reference_images` at 14 for every remaining model, so the second range value and the note were reduced, including the overview sentence that still claimed the cap depends on the selected model. - `ClipLoader` (ComfyUI #17095): the `type` combo also gained `moss_image`, and the missing `flux3 | qwen3-vl 4B` row was added to the type-to-encoder mapping table in all 15 files. - `ReferenceMultiLatent` (ComfyUI #17053): `empty_latent` is no longer `None` when nothing is encoded. The output description and the closing note now say that without a connected VAE (or without images) a square empty latent is returned, 1024 x 1024 or the size derived from `megapixels` when `resolution` is `"resize"`, in all 15 files. - `update_param_translations.py` also localized parameter names that had drifted from the frontend term table inside the touched node directories (for example `prompt` -> `Prompt`, `reference_images` -> `Referenzbilder` in the German Quiver pages, `image` -> `immagine` in Italian). Its library-wide footprint (1676 files across de/he/it/es, mostly the upstream batch translation debt) was reverted and stays out of this PR. ### Reviewed and intentionally skipped (this round) - `Flux3ImageNode` / `Flux3LayoutPrompt`: display labels renamed to "FLUX 3 Image" (ComfyUI #17075). The published page title comes from the node directory and the frontend still shows "Flux 3", so the headings were left alone. - `SaveImageAdvanced` (ComfyUI #16867): EXR is converted to scene-linear before encoding and asset previews are written for the asset manager, both execution-layer changes; the documented parameters, ranges and behaviour notes are unchanged. - `CreateBoundingBoxes`, `Preview3DAdvanced`, `PreviewGaussianSplat`, `PreviewPointCloud`, `SaveGLB`, `ByteDanceSeedAudio`, `ModelAttentionBackend`, `RecraftV4TextToVectorNode`: unchanged from the analysis above. ### Verification (follow-up commits) - `fix_translation_preamble.py --check-only --changed-only --base origin/main`: 0 files, 0 leaked lines - `check_doc_structure.py --changed-only`: 0 defects - `check_md_format.py --changed-only`: 0 errors (150 files) - `check_md_links.py --changed-only`: all local resource links valid - `check_new_node_translations.py MOSSImageLayoutPrompt`: clean; `check_translation_language.py`: 0 problems - Table row counts for `MOSSImageLayoutPrompt` match `en.md` in all 15 files; Range cells localized (`16 到 16384（步长 16）`, `16〜16384 （ステップ 16）`, `16 bis 16384 (Schritt: 16)`, ...) - Remaining `arrow-1.1` / `arrow-preview` / `Arrow 1` mentions: 0 across the four Quiver pages - Source fingerprints recomputed with `lib.node_source_extract` for `ClipLoader` (`3abcb39d`) and the four Quiver nodes (`9c85e3be`, `196b0693`, `a7618459`, `3531975f`); `ReferenceMultiLatent` already carried the current hash (`06f98e7c`) --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 22:10:27 +00:00