Anima 提示词分级审查(V4):
AnimaV4PromptGrader reviews a text prompt and assigns an NSFW content level using the same classification rules as AnimaV4RandomPrompt. It also identifies matching adult-content categories, helping you understand why a prompt received its classification rather than relying on a level alone. This makes it useful for reviewing manually written prompts, checking generated tag combinations, and maintaining consistent content classifications throughout your workflow. Its assessment is based on comma-separated prompt tags and the plugin’s classification table, not on an analysis of a generated image. The node is a reviewer rather than an editor: it does not remove adult-content tags, repair conflicting descriptions, or rewrite your prompt.
Anima 提示词分级审查(V4) Input Parameters:
text
A required STRING containing the prompt you want to review. The default is an empty string, and the interface provides a multiline text field; no minimum or maximum text length is specified. For classification, the node splits the text at ordinary commas, trims whitespace around each resulting tag, and ignores empty entries. For example, 1girl, blue hair, winter coat is processed as three tags. Line breaks alone do not separate tags, so comma-separated wording is important for predictable tag parsing. The original text is preserved separately and returned unchanged; trimming and removal of empty entries apply only to the list used for assessment.
detailed
An optional BOOLEAN that controls whether the report includes an additional list of matching tags. The default is True, and the available values are True and False; numeric minimum and maximum values do not apply. When enabled, the node appends a 命中标签: line if the classifier finds matching tags, showing at most the first 20 matches. When disabled, that extra line is omitted, but the report still includes the assigned level, the classification reason, the parsed tag count, and any adult-content category summaries. This setting changes report detail only; it does not change the classification or modify your prompt.
Anima 提示词分级审查(V4) Output Parameters:
text
A STRING containing the original prompt text, allowing you to pass your prompt onward while also reviewing its classification. The node does not reconstruct this output from its parsed tags, so original spacing, commas, and line breaks remain intact. An empty input produces an empty string here. This output is useful when you want to place the grader in a prompt-processing workflow without having it rewrite the text.
num
An INT representing the numeric NSFW level assigned by the plugin’s classifier. You can use this value to compare assessments or connect it to compatible downstream numeric logic. The provided node code does not define the complete level scale or its minimum and maximum, so those boundaries should not be assumed. The report displays this number alongside the classifier’s reported maximum and classification explanation, giving you context for interpreting it.
level
A STRING containing the classifier’s level identifier, taken directly from the classification result. It is distinct from the numeric num output and from the human-readable label shown in the report. This is useful when you need a textual classification value for a compatible downstream node or for recording results. The supplied code does not enumerate the possible identifiers, so you should inspect an actual result rather than assume specific names or options.
report
A STRING containing a multiline, Chinese-language explanation of the assessment. It starts with 档位 for the assigned numeric level, maximum, readable label, and level identifier; 基线 for the classifier’s reason; and 标签 for the number of nonempty comma-separated tags processed. The 成人向内容 section lists detected adult-content categories, with each category’s item count and up to six matching entries; an ellipsis indicates that a category contains additional entries. If no categories are returned, it displays 成人向内容:无. With detailed enabled and matching tags available, the report also includes 命中标签: followed by up to 20 matches. These display limits shorten the explanation; they do not mean the remaining tags were excluded from classification.
Anima 提示词分级审查(V4) Usage Tips:
- Separate your tags with ordinary commas, even when writing a multiline prompt. For example, use
1girl, blue hair, winter coatrather than placing each phrase on a separate line without commas. - Keep
detailedset toTruewhen investigating an unexpected classification. Compare the classification reason, category entries, and matching-tag line to understand which parts of your prompt were recognized. - Set
detailedtoFalsewhen you want a shorter report for routine review. Category summaries remain visible, and the classification stays the same. - Connect AnimaV4RandomPrompt’s
textoutput to this node’stextinput to review generated prompts using the same classification rules used by the generator. - Send
reportto a compatible text display or text-saving node. The supplied standalone workflow demonstrates prompt generation and review without requiring an image model or LoRA. - Grade the final assembled prompt if you add handwritten tags after generation. Reviewing only the original generated text will not assess additions made later in your workflow.
- Use AnimaV4PromptCheck when you need diagnosis and automatic repair rather than classification alone. AnimaV4PromptGrader does not remove over-level tags or resolve conflicting prompt content.
- Treat the assessment as a classification of recognized prompt tags, not a guarantee about the content of a future image. The node receives text, not image data.
Anima 提示词分级审查(V4) Common Errors and Solutions:
Missing node: Anima 提示词分级审查(V4)
- Explanation: If ComfyUI reports this node type as missing, the custom-node package may not be installed, may not have loaded successfully, or may be an older version. The exact warning text depends on your ComfyUI version.
- Solution: Check that Comfyui-Anima-Wardrobe-V4 is installed in your ComfyUI custom nodes directory, restart ComfyUI, and inspect the startup console for package-loading errors. The supplied code also registers
AnimaV3PromptGraderas a compatibility alias, so an older workflow using that name should be supported when this package loads correctly.
Report shows fewer tags than expected
- Explanation: This is a parsing symptom, not a built-in error message. The node separates tags only at ordinary commas and ignores empty entries. A list separated solely by line breaks, semicolons, or full-width commas will not be split as you might expect; repeated commas do not create additional nonempty tags.
- Solution: Replace your separators with ordinary commas and review the
标签count again. Ensure each intended tag is a separate, nonempty comma-separated entry.
Report displays 成人向内容:无 despite concerning wording
- Explanation: This message means the classifier returned no adult-content categories for the parsed tags. It does not establish that every possible interpretation of the prompt, or any image generated from it, is free of adult content. The assessment depends on the plugin’s classification rules and the supplied wording.
- Solution: Check your comma separation, review the complete level and baseline explanation, and enable
detailedto inspect any available matches. Review ambiguous wording yourself rather than treating an empty category list as a universal safety guarantee.
The 命中标签: line is missing or incomplete
- Explanation: The matching-tag line appears only when
detailedisTrueand the classifier returns matches. It shows at most 20 matches. Separately, each category summary shows at most six entries, followed by an ellipsis when more exist. - Solution: Enable
detailedfirst. If the line remains absent, no matching-tag list was returned for display. If the list is shortened, review the category counts and test smaller portions of the prompt when you need to investigate specific tags; the node has no setting to increase these display limits.
The prompt remains unchanged after review
- Explanation: This is expected behavior, not an execution error. The grader returns the original prompt and reports its classification; it does not rewrite text, remove adult-content tags, or correct conflicting descriptions.
- Solution: Edit the prompt yourself or use AnimaV4PromptCheck for its supported diagnosis and repair features. Run the revised text through the grader again to check the resulting classification.
Changing detailed does not change num or level
- Explanation: This is expected because
detailedcontrols only the additional matching-tag line in the report. It does not alter the tags submitted to the classifier or the classification rules. - Solution: Use
detailedto adjust report verbosity. To reassess different content, change the prompt itself and run the node again.
