ComfyUI Node: CLIP Positive-Negative (WLSH)

CLIP Positive-Negative (WLSH):

The CLIP Positive-Negative (WLSH) node is designed to facilitate the encoding of positive and negative textual prompts using the CLIP model. This node is particularly useful for AI artists who want to leverage the power of CLIP to condition their models with specific textual descriptions. By providing both positive and negative text inputs, the node encodes these texts into a format that can be used to influence the generation process, allowing for more controlled and nuanced outputs. This capability is essential for tasks that require fine-tuned conditioning, such as generating images that adhere closely to a given description while avoiding certain unwanted features.

CLIP Positive-Negative (WLSH) Input Parameters:

clip

This parameter expects a CLIP model instance. The CLIP model is responsible for encoding the provided textual inputs into a format that can be used for conditioning. The model should be pre-loaded and ready to use.

positive_text

This parameter takes a string input that represents the positive textual prompt. The text provided here will be encoded by the CLIP model to generate a positive conditioning vector. This vector will guide the model towards generating outputs that align with the positive description. The default value is an empty string, and it supports multiline input for more complex descriptions.

negative_text

This parameter takes a string input that represents the negative textual prompt. Similar to the positive_text parameter, the text provided here will be encoded by the CLIP model to generate a negative conditioning vector. This vector will help the model avoid generating outputs that contain features described in the negative prompt. The default value is an empty string, and it supports multiline input for more detailed descriptions.

CLIP Positive-Negative (WLSH) Output Parameters:

positive

This output provides the encoded positive conditioning vector. It is a tuple containing the encoded representation of the positive text and an empty dictionary. This vector is used to influence the model towards generating outputs that match the positive description.

negative

This output provides the encoded negative conditioning vector. Similar to the positive output, it is a tuple containing the encoded representation of the negative text and an empty dictionary. This vector helps the model avoid generating outputs that include features described in the negative prompt.

CLIP Positive-Negative (WLSH) Usage Tips:

• To achieve the best results, provide clear and concise descriptions in both the positive and negative text inputs. This helps the CLIP model generate more accurate conditioning vectors.
• Experiment with different combinations of positive and negative prompts to fine-tune the output. For example, if you want to generate an image of a cat without any background, you can use "a cat" as the positive text and "background" as the negative text.
• Utilize multiline input for more complex descriptions. This can help in scenarios where a single line of text is not sufficient to capture the desired features or avoid certain elements.

CLIP Positive-Negative (WLSH) Common Errors and Solutions:

"CLIP model not provided"

• Explanation: This error occurs when the CLIP model instance is not provided as an input.
• Solution: Ensure that you have loaded and passed a valid CLIP model instance to the clip parameter.

"Text input is empty"

• Explanation: This error occurs when the positive_text or negative_text input is left empty.
• Solution: Provide a valid string input for both the positive_text and negative_text parameters. Even if you do not have a specific negative prompt, consider using a generic term to avoid unwanted features.

"Encoding failed"

• Explanation: This error occurs when the CLIP model fails to encode the provided text inputs.
• Solution: Verify that the CLIP model is functioning correctly and that the text inputs are valid strings. If the problem persists, try reloading the CLIP model or using different text inputs.