Instructions to use MATLOWAI/MiniMax-H3-Motion-Adapter with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use MATLOWAI/MiniMax-H3-Motion-Adapter with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image, export_to_video # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("MiniMaxAI/MiniMax-H3", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("MATLOWAI/MiniMax-H3-Motion-Adapter") prompt = "A man with short gray hair plays a red electric guitar." input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/guitar-man.png") image = pipe(image=input_image, prompt=prompt).frames[0] export_to_video(output, "output.mp4") - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
model card: colour muting is about strong colour and particle effects, seen on both checkpoints
Browse files
README.md
CHANGED
|
@@ -51,10 +51,13 @@ Where it is not great (also measured):
|
|
| 51 |
- it costs about 1 dB of anchor fidelity on native keyframes and it
|
| 52 |
over-corrects calm chains (it smooths things that did not need
|
| 53 |
smoothing); keep it off on calm content
|
| 54 |
-
-
|
| 55 |
-
as a plain calico
|
| 56 |
-
|
| 57 |
-
|
|
|
|
|
|
|
|
|
|
| 58 |
- it is anime-heavy in effect: on very fast anime it wins by a mile at
|
| 59 |
every setting; on live-action character work it needs the lower inject
|
| 60 |
values below
|
|
|
|
| 51 |
- it costs about 1 dB of anchor fidelity on native keyframes and it
|
| 52 |
over-corrects calm chains (it smooths things that did not need
|
| 53 |
smoothing); keep it off on calm content
|
| 54 |
+
- it mutes strong colour and thins particle effects. A prismatic creature
|
| 55 |
+
came out as a plain calico on the Ref2VA checkpoint (mean saturation 46
|
| 56 |
+
to 38 at denoise 0.40, 54 to 40 at 0.70) and again on the fl2va
|
| 57 |
+
checkpoint when the whole clip was de-roped at 3x. Worst on Ref2VA and
|
| 58 |
+
on full-clip passes; a windowed de-rope on the same creature kept its
|
| 59 |
+
colour. If the subject is the colour, lower the strength or accept the
|
| 60 |
+
trade
|
| 61 |
- it is anime-heavy in effect: on very fast anime it wins by a mile at
|
| 62 |
every setting; on live-action character work it needs the lower inject
|
| 63 |
values below
|