Repository navigation
docs: add PH3ImageConditioning, SoniloMusic15, TextGenerateV2; sync OpenAI and TextGenerate changes - #224
Merged
Conversation
…c OpenAI/TextGenerate schema changes New nodes (English): - PH3ImageConditioning (ComfyUI #17506): p-h3-image conditioning for a single still frame, autogrow image_1..image_9 references, VAE required with images. - SoniloMusic15 (ComfyUI #17475): full song with vocals, auto/custom duration. - TextGenerateV2 (ComfyUI #17492): growable images/videos/audios context stacks. Existing nodes (zero-AI injection, identifiers only): - OpenAIChatNode: model combo gains gpt-6.1-sol / gpt-6.1-terra and loses gpt-4.1-nano, o4-mini, o1-pro, o1 (#17493). - OpenAIChatConfig: reasoning-effort enumeration gains the two 6.1 models. - OpenAIGPTImage1 / OpenAIGPTImageNodeV2: gpt-image-1 retired (#17493). - TextGenerate: source marks it deprecated, localized note added.
|
Warning Review limit reached
This review includes 135 billable files and costs up to $33.75. View limit details
|
…g size rows The width/height widgets only carry step=32; the node does not round the value (latent dimensions are width//16), so the sentence was wrong in all 15 locales.
execute() raises when the aspect-preserving result exceeds 179 megapixels or 32000 pixels per side, so the width/height rows now say so in all 15 locales (CodeRabbit finding on Comfy-Org/docs#2235).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Daily node-docs update for 2026-11-04. Three new nodes documented by hand, five existing nodes updated by zero-AI injection (identifiers and one localized deprecation note), all 15 locales carried in the same PR.
120 files: 45 new (3 nodes × 15 locales) + 75 modified (5 nodes × 15 locales).
New nodes
PH3ImageConditioningfeat: Support PH3Imageimagesholdsimage_1…image_9whereimage_1is the edited image; VAE is required as soon as an image slot is connected. 2912×2912 default canvas when no image, size on a 32 px grid.SoniloMusic15feat(Sonilo): add Sonilo Music 1.5 nodedurationis a DYNAMIC_COMBO (auto/customwithseconds10–360), prompt capped at 2000 characters,seedis re-run control only (output is non-deterministic).TextGenerateV2Add TextGenerateV2 with Autogrow image inputsimages/videos/audioscontext stacks (50 / 100 / 100 slots), DYNAMIC_COMBO sampling profile,mtpspeculative decoding options,thinkingsplit into its own output.English bodies were written against the node source (
define_schema+execute) and saved throughsave_agent_doc.py, so title, disclaimer and fingerprint footer come from the pipeline. None of the three nodes exists in the frontendnodeDefsyet (verified withsync_frontend_translations.py --export), so their sourcedisplay_namewas injected intodata/node_translations.jsonfor all 15 locales before generation — all 15 files share the same title.Existing nodes (zero-AI injection)
No re-translation anywhere: every edit is a model identifier, a count word, or a note copied from an existing precedent page.
OpenAIChatNode—modelcombo followsSupportedOpenAIModelafter #17493:gpt-6.1-solandgpt-6.1-terraadded,gpt-4.1-nano,o4-mini,o1-pro,o1removed (16 options now).OpenAIChatConfig— the reasoning-effort enumeration gainsGPT-6.1 Sol low-max/GPT-6.1 Terra none-max. Still matchesSUPPORTED_REASONING_EFFORTS.OpenAIGPTImage1—gpt-image-1retired: option removed, intro counts two models, and the legacy size note now names onlygpt-image-1.5(article/verb agreement fixed per locale).OpenAIGPTImageNodeV2—gpt-image-1removed from the DYNAMIC_COMBO, intro counts four models, the shared section is nowgpt-image-1.5alone.TextGenerate— #17492 made it a deprecated subclass withdisplay_name="Generate Text (DEPRECATED)". Added the repo's standard deprecation note (wording copied byte-for-byte fromByteDanceSeedreamNode/ the canonical fr line, not re-translated). H1 staysGenerate Text: the published page title comes from the directory name, so H1 drift is invisible on the site.Footer fingerprints were refreshed for all five nodes with
lib.node_source_extract+ the report comparison; every hash prefix matches the scanner'snew_hash(TextGenerate→2ac2ffc6…exactly).Changed-but-skipped (79 scan entries)
scan_missing_nodes.pyreported 94 changed nodes; 89 were noise, reviewed withreview_changed_node_diffs.py(blacklist method) plus manual diffs:display_name/categoryrenames,search_aliases, transport-layer edits (asset_urls=True,PrunaPredictionRequestwrapping,UI.SavedResultrefactors,qem_simplifyrename, helper renames).RecraftTextToImageNodestill says $0.0572 and matches its schema).CreateVideochroma-subsampling validation (auto→Noneinstead of a string) — no doc-visible behaviour change.CreateBoundingBoxesbackground thumbnail saved intoui["background_images"]— thebackgroundinput and its aspect notes were already documented.Flux3ImageNodedescription wording ("FLUX 3" → "FLUX 3 Image") — cosmetic.ByteDanceFirstLastFrameNode/ByteDanceImageToVideoNode/ByteDanceTextToVideoNodegainedis_deprecated=Trueupstream (#17507). Their pages are already covered by the open docs: render the Seedance retirement callout as Markdown and finish all 15 locales #222 retirement callout, which tells users the model is being shut down and that saved workflows stop working — I did not stack a second generic note on the same three pages there.BatchImagesNode,PadImageNode,RotateMesh,SVD_img2vid_Conditioning: fingerprint baseline not resolvable. Parameter/output sets compared by hand (SVD uses the legacyINPUT_TYPESAPI so the extractor sees no inputs) — no drift.Verification
Ran after committing, so
--changed-onlyreally saw the 120 files:fix_translation_preamble.py --check-only --changed-only→ 0 files, 0 leaked linescheck_doc_structure.py --changed-only→ 0 defectscheck_md_format.py --changed-only→ 120 files, 0 errors, 0 warningscheck_md_links.py→ all local resource links validcheck_new_node_translations.py→ 1 finding, a known false positive: de0 bis 100 Videos("Videos" is the German word)check_translation_language.py→ 0 problems|row counts toen.md(12 / 11 / 26), no fenced blocks, overview present in the local language, no English Range/Required residue (esNoand itNoare correct localizations; fr "0 à 9 images" / es "0 a 100 videos" / it "0 a 100 video" are correct unit words)update_param_translations.pywas deliberately not run: none of the three new nodes has frontend terminology, so a full pass would only rewrite ~1700 unrelated files (de/he/it parameter-name debt), which belongs in its own batch PR.Not touched:
docsrepo. Node content lives in embedded-docs; a sync PR will pick it up.