MiniMax H3 Decode AV:
MiniMaxH3Decode turns MiniMax-H3’s combined audiovisual representation into media you can use in a workflow. For a standard encoded result, it decodes the video, arranges the frames in sequence, and converts their values to the usual image range. It also decodes the sound when an audio decoder is available. If the result is already in the node’s rolling-output form, the node passes its video and audio through without decoding them again. This gives you one place to prepare generated visuals and sound while keeping their order together.
MiniMax H3 Decode AV Input Parameters:
latent
The required MINIMAX_H3_LATENT value containing the audiovisual result to process. For a standard latent, the node reads its video and audio components and decodes them; for a rolling output, it uses the video and audio already present. Connect a compatible MiniMax-H3 result, since this input determines the content of both returned outputs. The source specifies no default value, minimum, maximum, or selectable options.
av_encoder
The required MINIMAX_H3_AV_ENCODER value that supplies the video and audio decoder paths for a standard latent. The node uses those paths to load the decoder pack, so the selected encoder affects whether the encoded media can be decoded. The rolling-output branch does not load that pack, although the socket remains required. The source specifies no default value, minimum, maximum, or selectable options.
MiniMax H3 Decode AV Output Parameters:
video
The IMAGE output containing video frames in sequence. For a standard latent, the node decodes the video, combines its batch and time dimensions into a frame sequence, converts frame values to the 0–1 image range, and moves the result to the CPU. For a rolling output, it returns the video already stored in that result. Use this output wherever your workflow accepts image frames; the source does not specify a frame rate for playback.
audio
The AUDIO output containing a waveform and a sample_rate of 32000 Hz. For a standard latent, the waveform is decoded audio moved to the CPU when an audio decoder is available. If the audio decoder is unavailable, or if a rolling output has no audio, the node returns a zero-length, two-channel waveform instead. An audio output socket can therefore be present even when there is no audible content.
MiniMax H3 Decode AV Usage Tips:
- Connect both required MiniMax-H3 sockets. For a standard latent, use an
av_encoderwith valid video and audio decoder paths; this node exposes no quality, frame-rate, or performance settings to tune. - Connect
videoto a frame-compatible preview or export step andaudioto an audio-compatible step. Set playback or export timing downstream, since this node does not provide a video frame rate. - Check that the audio waveform has samples before expecting sound. A missing audio decoder or absent rolling-output audio produces an empty waveform, not an error or recorded silence of the video’s duration.
- When the incoming value is a rolling output, the node returns its existing media directly. The video normalization and CPU transfer described for standard latents are not performed in that branch.
MiniMax H3 Decode AV Common Errors and Solutions:
Required input is not connected
- Explanation:
latentandav_encoderare both required sockets. The workflow may refuse to run if either connection is missing. - Solution: Connect a compatible
MINIMAX_H3_LATENTsource tolatentand aMINIMAX_H3_AV_ENCODERsource toav_encoder.
FileNotFoundError or a decoder-loading error
- Explanation: When processing a standard latent, the node loads a decoder pack from
av_encoder.video_pathandav_encoder.audio_path. The exact error depends on the decoder loader, whose implementation is not included here. - Solution: Check that the encoder configuration points to the expected decoder files and that those files are accessible. The node does not load a decoder pack for a rolling output.
AttributeError: an expected latent or encoder attribute is missing
- Explanation: A standard latent must provide
videoandaudio, and the encoder must providevideo_pathandaudio_path. An incompatible object may fail when the node accesses one of these attributes. - Solution: Use nodes that produce the socket types declared by this node, and check that the connected objects come from compatible MiniMax-H3 components.
RuntimeError: a video tensor has the wrong number of dimensions
- Explanation: After decoding a standard latent, the node rearranges video data assuming five dimensions in the order batch, color channels, time, height, and width. A differently shaped decoder result can fail at
permute. - Solution: Check that the latent and decoder pack are compatible. Avoid supplying an image tensor or another latent format as the MiniMax-H3 latent.
AttributeError or a shape error when audio is unavailable
- Explanation: If no audio decoder is available, the node creates an empty waveform using
latent.audio.shape[0]. A standard latent without a usable audio component can fail at that step; the exact message depends on its value. - Solution: Supply a compatible standard audiovisual latent with a valid audio component, even if you do not plan to use the audio output.
Audio output is present but has no sound
- Explanation: This can be expected behavior rather than an exception. The node returns a zero-length waveform when the standard decoder pack has no audio decoder or when a rolling output’s audio is
None. - Solution: Check for a valid audio decoder in the standard-latent path, or check whether the rolling output actually contains audio. Inspect waveform length before sending it to an audio export step.
