3D Pose Editor:
PoseEditor3D turns a captured view from a 3D pose editor into an image you can use elsewhere in an AI art workflow. It decodes a PNG or JPEG capture, scales it to the chosen output dimensions, and places it over an optional background using the capture’s transparency. It also creates a mask from that transparency, so you can identify the visible parts of the capture separately from the rest of the image. If no valid capture is available, the node can still produce a background image or a plain light-gray image, with an empty pose mask.
3D Pose Editor Input Parameters:
image_data
A string containing a base64-encoded PNG or JPEG capture. A data-URL-style prefix is accepted because the node removes everything through the first comma before decoding. The image’s visible pixels are composited over the background, while its alpha channel determines the masks. In Standard mode, a valid capture also determines the output dimensions. The default is an empty string; no length limit is specified. An empty or undecodable value produces no pose layer rather than stopping the node.
output_size_mode
Chooses how the final image dimensions are determined. The options are Standard, Background, and Custom; the default is Standard. Standard uses the capture’s dimensions when available, otherwise the background’s dimensions, otherwise 600 × 600 pixels. Background uses the background’s dimensions when one is connected; without a background, it falls back to the same sizing rule as Standard. Custom uses custom_width and custom_height. Images are resized to the selected dimensions, so choosing a different aspect ratio can stretch their appearance.
custom_width
Sets the final image width in pixels when output_size_mode is Custom. The default is 600, the minimum is 64, the maximum is 4096, and the interface step is 8 pixels. It does not control output width in the other modes. Larger values provide more pixels for downstream work but also create larger images and masks.
custom_height
Sets the final image height in pixels when output_size_mode is Custom. The default is 600, the minimum is 64, the maximum is 4096, and the interface step is 8 pixels. It does not control output height in the other modes. Choose it together with custom_width to preserve the capture’s proportions when stretching would be undesirable.
timer_interval
An integer setting with a default of 5, a minimum of 1, a maximum of 3600, and a step of 1. Although the interface exposes this setting, the supplied node method does not use it to time capture, change the exported image, or generate masks. No timing behavior or unit can be confirmed from the provided source.
lip_text
Optional text associated with the pose-editor workflow. It defaults to an empty string and supports multiple lines; no length limit is specified. The node does not interpret this text or use it to change the image or masks. It passes the text to the lip_text output, using an empty string if the supplied value is empty.
background_image
An optional IMAGE input placed beneath the capture. The node uses the first image in the supplied batch and resizes it to the final dimensions if necessary. It does not contribute to the pose masks: those come only from the capture’s alpha channel. No default image or size range is specified. When this input is absent, the node starts with a solid light-gray background.
3D Pose Editor Output Parameters:
image
The final RGB image, containing the resized background or light-gray base with the available pose capture composited on top. Its dimensions follow output_size_mode. Because this output is RGB, it does not retain a separate transparency channel; use the mask outputs when you need the capture’s transparency for later processing. If the capture is missing or cannot be decoded, this output contains only the background or light-gray base.
mask
A grayscale mask derived from the resized capture’s alpha channel. Values range from 0 to 1: 0 means the capture is fully transparent at that pixel, 1 means fully opaque, and intermediate values represent partial transparency, such as soft edges. The background image does not affect it. If no valid capture is available, every value is 0.
inverted_mask
The inverse of mask, calculated as 1 minus the mask value at each pixel. It is 1 where the capture is fully transparent, 0 where it is fully opaque, and takes intermediate values along partially transparent edges. This is useful when a later step needs the area outside the captured pose instead of the pose area. If no valid capture is available, every value is 1.
lip_text
The supplied lip_text, passed through as a string, or an empty string if no text was supplied. It lets a later node receive the text alongside the visual outputs; this node does not generate speech, lip movement, or an interpretation of the text.
3D Pose Editor Usage Tips:
- Use
Standardwhen you want to retain a valid capture’s pixel dimensions. UseBackgroundwhen the final image must match a connected background, andCustomwhen a later step requires a specific width and height. - Keep
custom_widthandcustom_heightat the same aspect ratio as the capture or background when you want to avoid stretching. The node resizes images to fit; it does not add padding to preserve their proportions. - Supply a capture with transparency when you want the background to show around the pose and the masks to distinguish the pose from its surroundings. An opaque capture produces an all-ones
maskand an all-zerosinverted_mask. - Check the
maskoutput before relying on it for downstream editing. If it is entirely zero, the capture may be missing, empty, or undecodable; the presence of a background inimagedoes not mean a pose was decoded. - Use moderate
Customdimensions while testing, then increase them if the final workflow needs more detail. Higher dimensions increase the number of pixels in all three visual outputs. - Treat
timer_intervalas an interface setting whose effect is not established by this node source. Changing it does not change the calculations shown here.
3D Pose Editor Common Errors and Solutions:
[ 3D Pose Editor] 画像デコードエラー: Invalid image magic number — expected PNG or JPEG
- Explanation: After decoding
image_data, the node did not find the expected PNG or JPEG signature at the start of the file data. The pose is omitted, although the node can still return a background or light-gray image. - Solution: Provide a base64-encoded PNG or JPEG capture. If the string includes a prefix, use a data-URL-style prefix followed by a comma and then the encoded image data.
[ 3D Pose Editor] 画像デコードエラー: ...
- Explanation: This message prefix is printed when decoding or opening
image_dataraises an exception; the text after the colon depends on the underlying problem. For example, the encoded data may be malformed or may not contain an image that can be opened. - Solution: Regenerate the capture, check that the complete encoded string reaches
image_data, and confirm that its underlying file is a valid PNG or JPEG. Inspect the remainder of the printed message for a more specific clue.
No decode message, but the pose is missing
- Explanation: An empty or whitespace-only
image_datavalue is skipped without a decode error. The output then contains only the background or light-gray base, with an all-zeromask. - Solution: Check that the pose editor has supplied a nonempty capture to
image_data, then inspectmaskto confirm that capture pixels are present.
The pose or background looks stretched
- Explanation: The node resizes each available image to the selected output width and height, even when that changes its original aspect ratio.
- Solution: Use
Standardfor the capture’s dimensions, useBackgroundto match a connected background, or setCustomdimensions with an appropriate aspect ratio.
mask is fully white or inverted_mask is fully black
- Explanation: The decoded capture is opaque across the frame. The masks describe the capture’s alpha channel, not the shape of the 3D figure within an opaque image.
- Solution: Use a capture with transparency around the figure if you need a mask that separates the figure from its surroundings.
Changing timer_interval has no visible effect
- Explanation: The supplied node method accepts
timer_intervalbut does not use it when producing its outputs. - Solution: Do not rely on this parameter to change the exported image or masks. Check the pose-editor component responsible for sending the capture if you are troubleshooting capture timing.
An updated background does not appear in a cached result
- Explanation: The node’s
IS_CHANGEDkey includes the capture, size settings, and lip text, but notbackground_image. Depending on how the workflow uses that key, a background-only change may not prompt this node to rerun. - Solution: Force the node or workflow to execute again after changing only the background, and confirm the refreshed
imageoutput.
