Add same-day MiniMax H3 workflow imports

This commit is contained in:
Morpheus
2026-08-28 18:06:33 +00:00
parent 42a8755830
commit 2b88a96c7a
4 changed files with 104 additions and 0 deletions
@@ -0,0 +1,51 @@
# MiniMax H3 Steffalkok I2V PDD-Acc
## Summary
Same-day `MiniMax H3` image-to-video workflow from StefanFalkok, using the `PDD-Acc` turbo node path.
This is the companion import to the new ref2v workflow.
## Model Family
- `minimax`
## Status
- `imported-only`
## Source
- Model card: <https://huggingface.co/StefanFalkok/Minimax_H3_Workflows>
- Workflow file: <https://huggingface.co/StefanFalkok/Minimax_H3_Workflows/resolve/main/Minimax%20H3/Minimax%20H3%20I2V%20%28with%20PDD-Acc%208-Steps%20Turbo%20Nodes%29.json>
- Original publisher: `StefanFalkok`
- Date imported: `2026-08-28`
## Developer notes
- The source pack positions these as practical ComfyUI workflows for `MiniMax H3`.
- The same repo also includes a plain I2V workflow and a longer-video beta pack.
- The author credits `Comfy-Org/MiniMax-H3` for the model files.
## Our notes
- Useful as a lower-friction community I2V variant if you want a turbo-node route rather than the official baseline.
- Still likely tight on a `3070`, but the workflow is interesting enough to keep as a reference point.
- This is the companion import to the ref2v workflow above, not a separate new lane.
## Required custom nodes
- `ComfyUI-MiniMax-H3-Turbo`
- `ComfyUI-MiniMax-H3-PDD-Acc`
- `ComfyUI-MiniMax-H3-LongMedia`
## Required models
- `Comfy-Org/MiniMax-H3`
- matching `MiniMax H3` text encoder
- matching `MiniMax H3` VAE
- optional turbo LoRA files used by the graph
## Notes
- Best new same-day MiniMax H3 import for image-to-video.
- Good companion to the repo's existing official and INT8 MiniMax workflows.
File diff suppressed because one or more lines are too long
@@ -0,0 +1,51 @@
# MiniMax H3 Steffalkok Ref2V PDD-Acc
## Summary
Same-day `MiniMax H3` reference-to-video workflow pack from StefanFalkok, using the `PDD-Acc` turbo node path.
This is the more interesting of the new imports for reference control and identity continuity.
## Model Family
- `minimax`
## Status
- `imported-only`
## Source
- Model card: <https://huggingface.co/StefanFalkok/Minimax_H3_Workflows>
- Workflow file: <https://huggingface.co/StefanFalkok/Minimax_H3_Workflows/resolve/main/Minimax%20H3/Minimax%20H3%20Ref2V%20%28with%20PDD-Acc%208-Steps%20Turbo%20Nodes%29.json>
- Original publisher: `StefanFalkok`
- Date imported: `2026-08-28`
## Developer notes
- The source pack explicitly calls out `FL2VA` and `Ref2Video` workflows with turbo LoRA nodes.
- The author says all models come from `Comfy-Org/MiniMax-H3`.
- The pack also points to companion prompt-generator workflows and a longer-video beta variant.
## Our notes
- This is the better MiniMax H3 import for your current taste because it is directly about reference-driven video, not just generic generation.
- Still likely heavy on a `3070`, but the `PDD-Acc` turbo path makes it more worth testing than a brute-force baseline.
- Good candidate if you want a practical reference-control starting point that can sit beside the existing MiniMax imports.
## Required custom nodes
- `ComfyUI-MiniMax-H3-Turbo`
- `ComfyUI-MiniMax-H3-PDD-Acc`
- `ComfyUI-MiniMax-H3-LongMedia`
## Required models
- `Comfy-Org/MiniMax-H3`
- matching `MiniMax H3` text encoder
- matching `MiniMax H3` VAE
- optional turbo LoRA files used by the graph
## Notes
- Best new same-day MiniMax H3 import for reference control.
- Use this when you want a community workflow source rather than the official baseline.
File diff suppressed because one or more lines are too long