Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > ComfyUI-MiniMax-H3-MotionCache > MiniMax H3 Fast VAE Decode

ComfyUI Node: MiniMax H3 Fast VAE Decode

Class Name

MiniMaxH3FastVAEDecode

Category
model/latent/minimax
Author
starsfriday (Account age: 0days)
Extension
ComfyUI-MiniMax-H3-MotionCache
Latest Updated
2026-11-11
Github Stars
0.01K

How to Install ComfyUI-MiniMax-H3-MotionCache

Install this extension via the ComfyUI Manager by searching for ComfyUI-MiniMax-H3-MotionCache
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter ComfyUI-MiniMax-H3-MotionCache in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

MiniMax H3 Fast VAE Decode Description

Decodes MiniMax H3 video latents into image frames using batched GPU tiles with memory fallback.

MiniMax H3 Fast VAE Decode:

MiniMax H3 Fast VAE Decode converts MiniMax H3 video latents into viewable images using an experimental batched-tile decoding method. It processes multiple tiles together on the GPU, which may improve decoding speed in some setups but can require more video memory. Because larger batches are not always faster, compare its performance with the standard VAEDecode node on your own GPU. If a tile batch runs out of memory, the node automatically retries with one tile per batch.

MiniMax H3 Fast VAE Decode Input Parameters:

samples

The LATENT data to decode. Connect a MiniMax H3 video latent or joint video/audio latent from your workflow. The node reads the latent samples and converts their video content into images; it does not provide a separate audio output. This input has no numeric minimum, maximum, or default value.

vae

The VAE model used to turn the latent into images. This must be a MiniMax H3 video VAE: the node checks the model type and rejects other VAEs. Connect the matching VAE for your MiniMax H3 workflow rather than a general-purpose image VAE. This input has no numeric minimum, maximum, or default value.

tile_batch_size

The number of decoding tiles processed together in one GPU batch. Its default is 4, with a minimum of 1, a maximum of 8, and increments of 1. Higher values use more video memory and are not guaranteed to decode faster; lower values can help when memory is limited. If decoding runs out of memory at the selected value, the node clears its cache and retries with a batch size of 1.

MiniMax H3 Fast VAE Decode Output Parameters:

images

The decoded IMAGE data, ready to connect to image-processing or viewing nodes. For video latents, the result represents decoded video frames; when decoding produces a five-dimensional result, the node reshapes it into an image batch suitable for the IMAGE output. This output does not include a separate audio stream.

MiniMax H3 Fast VAE Decode Usage Tips:

  • Start with the default tile_batch_size of 4, then compare decoding time and GPU memory use against values such as 1 and 2. Keep the setting that performs best on your GPU rather than assuming 8 will be fastest.
  • If you have limited GPU memory, set tile_batch_size to 1 before running. This avoids first attempting a larger batch that may run out of memory and require a retry.
  • Use a MiniMax H3 video VAE with MiniMax H3 latents. Other VAE types are rejected by this node.
  • Benchmark this experimental decoder against standard VAEDecode using the same workflow and GPU. A completion-time message is written to the application log to help you compare runs.

MiniMax H3 Fast VAE Decode Common Errors and Solutions:

MiniMax H3 Fast VAE Decode requires the MiniMax H3 video VAE

  • Explanation: The connected VAE is not a MiniMax H3 video VAE, so the node stops before decoding.
  • Solution: Connect the MiniMax H3 video VAE intended for your workflow, then run the node again.

MiniMax H3 Fast VAE Decode ran out of memory at tile batch N; retrying with 1

  • Explanation: This is a warning rather than necessarily a final failure. The selected tile batch exceeded available memory, and the node is automatically retrying with a batch size of 1.
  • Solution: If the retry succeeds, you can use the resulting images. To avoid the warning and retry on future runs, lower tile_batch_size, preferably to 1 if memory is tight.

Out-of-memory error during the retry with tile batch 1

  • Explanation: The automatic retry can also fail if there is not enough available GPU memory to decode even one tile at a time.
  • Solution: Free GPU memory used by other models or applications and try again. If your workflow allows it, reduce the amount of video content being decoded per run.

Missing or invalid samples data

  • Explanation: The node expects a LATENT connection containing latent samples. If the connection is missing or the data is not a compatible latent, decoding cannot proceed; the exact error text depends on the failure.
  • Solution: Check that samples is connected to a MiniMax H3 video or joint video/audio latent produced by the appropriate workflow.

Invalid VAE error

  • Explanation: Before decoding, the node asks the connected VAE to check whether it is valid. A VAE that fails this check stops execution; the exact error text depends on the VAE.
  • Solution: Reload or replace the VAE and confirm that your workflow supplies a working MiniMax H3 video VAE.

MiniMax H3 Fast VAE Decode Related Nodes

Go back to the extension to check out more related nodes.
ComfyUI-MiniMax-H3-MotionCache
RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

MiniMax H3 Fast VAE Decode