ComfyUI Node: Fisher Qwen2.1 自由姿势

Fisher Qwen2.1 自由姿势:

FisherQwenFreePose is a versatile node designed to facilitate the creation of free-form poses for characters in digital art workflows. It leverages the VNCCS Pose Studio LoRA convention, specifically tailored for Qwen Image 2.1, to enable artists to manipulate character poses with precision and creativity. This node allows you to input a reference image and a pose description in JSON format, which it uses to generate a mannequin render that serves as a guide for reposing characters. The primary goal of FisherQwenFreePose is to provide a seamless integration into any image editing workflow, allowing for the preservation of character appearance and scene context while offering the flexibility to adjust poses as needed. By using this node, you can achieve a high level of detail and accuracy in character positioning, making it an invaluable tool for artists seeking to enhance their creative projects.

Fisher Qwen2.1 自由姿势 Input Parameters:

clip

The clip parameter is a required input that represents the CLIP model used for encoding the text prompt. It plays a crucial role in interpreting the pose instructions and generating the corresponding visual output. This parameter does not have a default value and must be provided to ensure the node functions correctly.

vae

The vae parameter is another required input that stands for the Variational Autoencoder model. It is essential for decoding the latent space representations into images, ensuring that the generated poses are visually coherent and aligned with the input instructions. Like the clip parameter, it does not have a default value and must be specified.

reference_image

The reference_image parameter is a required input that provides the initial visual context for the pose generation. This image serves as a baseline for the node to understand the character's appearance and scene, allowing it to accurately adjust the pose while maintaining visual consistency. There is no default value for this parameter, and it must be supplied by the user.

width

The width parameter specifies the width of the output image in pixels. It is an integer value that determines the horizontal resolution of the generated pose image. The default value is not explicitly mentioned, but it is crucial to set this parameter according to the desired output size to ensure the image fits the intended use case.

height

The height parameter defines the height of the output image in pixels. Similar to the width parameter, it is an integer value that affects the vertical resolution of the generated image. Setting this parameter appropriately is important for achieving the desired aspect ratio and visual quality.

pose_json

The pose_json parameter is a required input that contains the pose description in JSON format. This parameter is critical as it provides the detailed instructions for how the character should be posed, including information about joint positions and orientations. The default value is an empty JSON object ({}), but it should be populated with specific pose data to achieve the desired results.

Fisher Qwen2.1 自由姿势 Output Parameters:

image_1

The image_1 output parameter represents the mannequin render generated based on the input pose description. This image serves as a visual guide for reposing the character and is crucial for ensuring that the final pose aligns with the artist's intentions.

image_2

The image_2 output parameter corresponds to the reference image provided as input. It is included in the output to maintain a connection between the original visual context and the generated pose, allowing for a seamless integration into the overall workflow.

image_3

The image_3 output parameter is optional and only present when working with two separate reference images. It provides the second reference image, ensuring that both characters in a two-person scene are accurately represented and posed according to the input instructions.

prompt

The prompt output parameter is a string that contains the text prompt generated based on the input pose description and additional instructions. This prompt is essential for guiding the CLIP model in understanding the desired pose and generating the corresponding visual output.

latent

The latent output parameter is a dictionary containing the encoded latent space representation of the generated pose. This data is crucial for further processing and manipulation within the image editing workflow, allowing for advanced adjustments and refinements.

pose_json

The pose_json output parameter is the same JSON object provided as input, included in the output to maintain consistency and allow for easy reference and modification if needed.

Fisher Qwen2.1 自由姿势 Usage Tips:

• Ensure that the pose_json parameter is accurately populated with detailed pose instructions to achieve the desired character positioning.
• Use high-quality reference images to provide a clear visual context for the node, enhancing the accuracy and coherence of the generated poses.
• Adjust the width and height parameters to match the intended output size, ensuring that the generated images fit seamlessly into your workflow.

Fisher Qwen2.1 自由姿势 Common Errors and Solutions:

合照模式只需要 reference_image；请断开 reference_image_2，或在编辑器切换为两张单人照。

• Explanation: This error occurs when the node is set to group mode, but a second reference image is provided, which is unnecessary in this mode.
• Solution: Disconnect the reference_image_2 input or switch the editor to use two separate single-person images.

编辑器里有 2 个人：请把人物 2 的照片接到 reference_image_2（对应 image3）。

• Explanation: This error indicates that the editor is set to separate mode for two people, but the second reference image is missing.
• Solution: Connect the second person's photo to the reference_image_2 input to ensure both characters are represented.

接了 reference_image_2，但编辑器里只有 1 个人：请在编辑器里点「＋ 第二个人」，或断开 reference_image_2。

• Explanation: This error arises when a second reference image is connected, but the editor is set for a single person.
• Solution: Either add a second person in the editor or disconnect the reference_image_2 input to resolve the mismatch.