Text-to-Video
MiniMax H3
Safetensors
lora
adapter
comfyui
reference-to-video
audio-video
synchronized-audio
few-step
turbo
accelerated-inference
hyperflow
lightx2v
dynamic-rank
svd
bfloat16
pruned-model
curve-form
Instructions to use drbaph/MiniMax-H3-Turbo-Lora-ComfyUI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Inference
- Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -37,6 +37,12 @@ This repository contains third-party ComfyUI compatibility conversions of the or
|
|
| 37 |
|
| 38 |
These conversions are intended for the **pruned/curve-form MiniMax-H3 checkpoint used by ComfyUI**.
|
| 39 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
## Original Project
|
| 41 |
|
| 42 |
The original LoRA weights, training work, four-step distillation method, dual video/audio sampling implementation, documentation and source files were created and released by **larryvrh**:
|
|
@@ -54,6 +60,9 @@ Full credit for the original MiniMax-H3 Turbo LoRA belongs to **larryvrh**.
|
|
| 54 |
|---|---|
|
| 55 |
| `minimax_h3_turbo_4step_pruned_comfyui.safetensors` | Partial ComfyUI compatibility conversion of the original trained non-EMA weights for the pruned/curve-form MiniMax-H3 checkpoint. The original author describes these weights as crisper and better at holding fast motion. |
|
| 56 |
| `minimax_h3_turbo_4step_ema_pruned_comfyui.safetensors` | Partial ComfyUI compatibility conversion of the time-averaged EMA weights for the pruned/curve-form MiniMax-H3 checkpoint. The original author describes this early EMA checkpoint as smoother but softer and not yet fully matured. |
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
The EMA and non-EMA files remain separate and contain different tensor values.
|
| 59 |
|
|
|
|
| 37 |
|
| 38 |
These conversions are intended for the **pruned/curve-form MiniMax-H3 checkpoint used by ComfyUI**.
|
| 39 |
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
[Download workflow](https://huggingface.co/drbaph/MiniMax-H3-Turbo-Lora-ComfyUI/resolve/main/fl_minimax_h3_turbo_lora_example_workflow.json?download=true)
|
| 44 |
+

|
| 45 |
+
|
| 46 |
## Original Project
|
| 47 |
|
| 48 |
The original LoRA weights, training work, four-step distillation method, dual video/audio sampling implementation, documentation and source files were created and released by **larryvrh**:
|
|
|
|
| 60 |
|---|---|
|
| 61 |
| `minimax_h3_turbo_4step_pruned_comfyui.safetensors` | Partial ComfyUI compatibility conversion of the original trained non-EMA weights for the pruned/curve-form MiniMax-H3 checkpoint. The original author describes these weights as crisper and better at holding fast motion. |
|
| 62 |
| `minimax_h3_turbo_4step_ema_pruned_comfyui.safetensors` | Partial ComfyUI compatibility conversion of the time-averaged EMA weights for the pruned/curve-form MiniMax-H3 checkpoint. The original author describes this early EMA checkpoint as smoother but softer and not yet fully matured. |
|
| 63 |
+
| `fl_minimax_h3_turbo_lora_example_workflow.json` | first & last frame comfyui example workflow using turbo lora + optional optimisations |
|
| 64 |
+
|
| 65 |
+
|
| 66 |
|
| 67 |
The EMA and non-EMA files remain separate and contain different tensor values.
|
| 68 |
|