Video-to-Video
Diffusers
Safetensors
English
JoyVideoEditPipeline
video
video-editing
reference-image-guided
autoregressive-diffusion
Instructions to use jdopensource/JoyAI-Video-Edit-Diffusers with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use jdopensource/JoyAI-Video-Edit-Diffusers with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("jdopensource/JoyAI-Video-Edit-Diffusers", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
scheduler: shift=5.159 to match the deployed denoise schedule
Browse filesThe serving stack constructs FlowMatchDiscreteScheduler(shift=5.159)
(deploy/xvideo/config.py), giving 2-step sigmas [1.0, 0.8376, 0.0].
With shift=1.0 the pipeline denoised along [1.0, 0.5, 0.0] — a schedule
the distilled checkpoint was never trained on.
Verified: with shift=5.159 the JoyVideoEditPipeline sigma/timestep
trajectory (raw linspace sigmas -> config-shift in set_timesteps)
matches the deployment scheduler exactly.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
scheduler/scheduler_config.json
CHANGED
|
@@ -7,7 +7,7 @@
|
|
| 7 |
"max_image_seq_len": 4096,
|
| 8 |
"max_shift": 1.15,
|
| 9 |
"num_train_timesteps": 1000,
|
| 10 |
-
"shift":
|
| 11 |
"shift_terminal": null,
|
| 12 |
"stochastic_sampling": false,
|
| 13 |
"time_shift_type": "exponential",
|
|
|
|
| 7 |
"max_image_seq_len": 4096,
|
| 8 |
"max_shift": 1.15,
|
| 9 |
"num_train_timesteps": 1000,
|
| 10 |
+
"shift": 5.159,
|
| 11 |
"shift_terminal": null,
|
| 12 |
"stochastic_sampling": false,
|
| 13 |
"time_shift_type": "exponential",
|