ComfyUI Node: MiniMax H3 Video Reference (Plan v2)

MiniMax H3 Video Reference (Plan v2):

The MiniMaxH3PlanV2VideoReference node is designed to facilitate the generation of video content by leveraging reference images, videos, and audio inputs. This node is part of the MiniMax H3 model suite, which is known for its ability to condition video outputs based on a variety of media references. The primary goal of this node is to enable users to create videos that are influenced by the characteristics and content of the provided reference media. By integrating multiple types of media, the node offers a comprehensive approach to video generation, allowing for nuanced and contextually rich outputs. This capability is particularly beneficial for AI artists looking to produce videos that maintain a coherent style or theme across different media types.

MiniMax H3 Video Reference (Plan v2) Input Parameters:

h3_plan

This parameter represents the plan setup for the MiniMax H3 model. It is crucial for defining the overall configuration and execution strategy of the node. The plan ensures that the node operates within the specified guidelines and constraints, allowing for a structured approach to video generation. The tooltip suggests connecting this to a preceding Plan v2 setup node, indicating its role in the broader workflow.

video_frames

This parameter consists of the decoded reference-video frames. It is essential for providing the visual content that the node will process and analyze. The source_fps describes the incoming batch of frames, and the node returns a batch aligned to the native 24-FPS. This ensures that the video output is consistent with the expected frame rate, which is crucial for maintaining video quality and synchronization.

video_use

This parameter allows you to specify the relationship between the whole video or its visible content. It provides options to choose how the video content is utilized within the node's processing. The default setting is UNASSIGNED_VIDEO_USE, but users can adjust this based on their specific needs, ensuring flexibility in how the video content is interpreted and applied.

reference_name

This parameter is a human-readable name for the source video, which is useful for reports and documentation. It allows users to easily identify and reference the video content within their projects. The default is an empty string, but users are encouraged to provide descriptive names to enhance clarity and organization.

MiniMax H3 Video Reference (Plan v2) Output Parameters:

resampled

This output contains the resampled video frames that have been processed by the node. It reflects the adjustments made to align the frames with the native 24-FPS, ensuring that the video output is consistent and meets the expected standards for playback and further processing.

source_duration

This output provides the duration of the source video, which is important for understanding the length and scope of the video content being processed. It helps users gauge the extent of the video and plan accordingly for any additional editing or integration tasks.

resampled_count

This output indicates the number of frames in the resampled video batch. It provides insight into the volume of data being handled by the node, which can be useful for performance considerations and resource management.

native_count

This output specifies the number of frames that fit within the native frame grid of the MiniMax H3 model. It ensures that the video content is compatible with the model's requirements, facilitating seamless integration and processing within the broader MiniMax H3 framework.

MiniMax H3 Video Reference (Plan v2) Usage Tips:

• Ensure that the reference media files do not exceed the specified limits: 9 images, 3 videos, and 3 audio clips, with a total of 12 mixed media files. This will prevent errors related to media file limits.
• When setting the video_frames parameter, ensure that the source video is properly decoded and matches the expected frame rate to avoid synchronization issues.
• Use descriptive names for the reference_name parameter to maintain clarity and organization in your projects, especially when dealing with multiple media sources.

MiniMax H3 Video Reference (Plan v2) Common Errors and Solutions:

"analysis_long_edge must be from 256 through 1536 pixels."

• Explanation: This error occurs when the specified long edge of the analysis is outside the acceptable range.
• Solution: Adjust the analysis_long_edge parameter to be within the range of 256 to 1536 pixels.

"video_analysis_fps must be above 0 and at most 2 FPS."

• Explanation: The specified frames per second for video analysis is either too low or too high.
• Solution: Set the video_analysis_fps parameter to a value greater than 0 and no more than 2 FPS.

"max_analysis_frames must be from 2 through 32."

• Explanation: The number of frames specified for analysis is outside the allowed range.
• Solution: Ensure that the max_analysis_frames parameter is set between 2 and 32.

"MiniMax H3 accepts at most 12 mixed reference media files."

• Explanation: The total number of reference media files exceeds the allowed limit.
• Solution: Reduce the number of reference media files to a maximum of 12, considering the limits for each media type.

"The prepared reference video is too short for H3's native frame grid."

• Explanation: The reference video does not have enough frames to fit the native frame grid of the MiniMax H3 model.
• Solution: Ensure that the reference video has a sufficient number of frames to meet the model's requirements.