H3 Contact Sheet (five views):
The H3ContactSheet node is designed to generate a contact sheet consisting of five coordinated views derived from a single reference image. This node is part of the ComfyUI-MAINodes collection and is specifically tailored for use with the MiniMax-H3 model. It leverages a unique diffusion process that packs five standalone image latents onto the model's time axis, allowing them to be jointly denoised and independently decoded. This approach is particularly beneficial for creating a comprehensive visual representation of a subject from multiple angles, making it ideal for tasks that require a detailed examination of an object or scene. By using a five-view LoRA (Low-Rank Adaptation) model, users can achieve high-quality, consistent outputs across all views. The node is compatible with various resolutions, providing flexibility in output quality and processing speed.
H3 Contact Sheet (five views) Input Parameters:
clip
This parameter accepts a CLIP (Contrastive Language–Image Pretraining) model input, which is used to process and understand the textual and visual information provided. It plays a crucial role in aligning the generated images with the given prompt.
vae
The VAE (Variational Autoencoder) input is essential for encoding and decoding image data. It helps in transforming the latent representations into actual images, ensuring that the output is visually coherent and matches the intended style and content.
prompt
This is a multiline, dynamic text input that guides the generation process. The prompt should describe the desired scene or object, and it can include specific tags like <Picture 1> to reference the input image. The default prompt suggests a camera orbit around the subject, which helps in creating a rotational view across the generated images.
ref_image
An optional image input that serves as the reference for generating the contact sheet. If left unconnected, the node will rely solely on the textual prompt to create the images, resulting in a text-to-sheet generation process.
size
This integer input determines the square size of each generated view, with options ranging from 512 to 2048. The default value is 1024, which is considered the sweet spot for balancing quality and processing time. A size of 512 corresponds to the LoRA training size and offers the fastest processing, while 2048 provides the highest resolution, with more pronounced rotational effects in later views.
H3 Contact Sheet (five views) Output Parameters:
positive
This output provides the conditioning data that aligns with the positive aspects of the prompt and reference image. It is used to guide the generation process, ensuring that the output images adhere to the desired characteristics and style.
latent
The latent output contains the encoded representations of the generated views. These latents are crucial for further processing and decoding into actual images, allowing for independent manipulation and refinement of each view.
H3 Contact Sheet (five views) Usage Tips:
- Ensure that the reference image is of high quality and relevant to the prompt to achieve the best results in the generated contact sheet.
- Experiment with different prompt descriptions to explore various angles and perspectives of the subject, enhancing the diversity of the output views.
- Adjust the size parameter based on your processing power and desired output quality. Lower sizes will render faster, while higher sizes will provide more detail.
H3 Contact Sheet (five views) Common Errors and Solutions:
Error: "Invalid reference image format"
- Explanation: This error occurs when the provided reference image is not in a supported format.
- Solution: Ensure that the reference image is in a standard format such as JPEG or PNG before uploading it to the node.
Error: "Prompt tag <Picture 1> not recognized"
- Explanation: The prompt tag
<Picture 1>is missing or incorrectly formatted in the prompt text. - Solution: Double-check the prompt to ensure that the tag
<Picture 1>is included and correctly formatted to reference the input image.
Error: "Size parameter out of range"
- Explanation: The size parameter is set outside the allowed range of 512 to 2048.
- Solution: Adjust the size parameter to fall within the specified range to ensure proper processing and output generation.
