mirror of
https://github.com/Wan-Video/Wan2.1.git
synced 2025-12-15 11:43:21 +00:00
Merge branch 'main' into better_Readme
This commit is contained in:
deepbeepmeep
GitHub
commit
263bef944e
14
README.md
14
README.md
@ -18,23 +18,23 @@ WanGP supports the Wan (and derived models), Hunyuan Video and LTV Video models
|
||||
**Discord Server to get Help from Other Users and show your Best Videos:** https://discord.gg/g7efUW9jGV
|
||||
|
||||
## 🔥 Latest Updates
|
||||
### May 28 2025: WanGP v5.31
|
||||
👋 Added Phantom 14B, a model that you can use to transfer objects / people in the video. My preference goes to Vace that remains the king of controlnets.
|
||||
VACE improvements: Better sliding window transitions, image mask support in Matanyone, new Extend Video feature, and enhanced background removal options.
|
||||
|
||||
### May 26, 2025: Wan 2.1GP v5.3
|
||||
### May 26, 2025: WanGP v5.3
|
||||
👋 Settings management revolution! Now you can:
|
||||
- Select any generated video and click *Use Selected Video Settings* to instantly reuse its configuration
|
||||
- Drag & drop videos to automatically extract their settings metadata
|
||||
- Export/import settings as JSON files for easy sharing and backup
|
||||
|
||||
### May 23, 2025: Wan 2.1GP v5.21
|
||||
👋 VACE improvements: Better sliding window transitions, image mask support in Matanyone, new Extend Video feature, and enhanced background removal options.
|
||||
|
||||
### May 20, 2025: Wan 2.1GP v5.2
|
||||
### May 20, 2025: WanGP v5.2
|
||||
👋 **CausVid support** - Generate videos in just 4-12 steps with the new distilled Wan model! Also added experimental MoviiGen for 1080p generation (20GB+ VRAM required).
|
||||
|
||||
### May 18, 2025: Wan 2.1GP v5.1
|
||||
### May 18, 2025: WanGP v5.1
|
||||
👋 **LTX Video 13B Distilled** - Generate high-quality videos in less than one minute!
|
||||
|
||||
### May 17, 2025: Wan 2.1GP v5.0
|
||||
### May 17, 2025: WanGP v5.0
|
||||
👋 **One App to Rule Them All!** Added Hunyuan Video and LTX Video support, plus Vace 14B and integrated prompt enhancer.
|
||||
|
||||
See full changelog: **[Changelog](docs/CHANGELOG.md)**
|
||||
|
||||
@ -114,6 +114,6 @@ def temporal_interpolation(model_path, frames, exp, device ="cuda"):
|
||||
model.to(device=device)
|
||||
|
||||
with torch.no_grad():
|
||||
output = process_frames(model, device, frames, exp)
|
||||
output = process_frames(model, device, frames.float(), exp)
|
||||
|
||||
return output
|
||||
|
||||
@ -82,9 +82,9 @@ class WanT2V:
|
||||
from mmgp import offload
|
||||
# model_filename = "c:/temp/vace1.3/diffusion_pytorch_model.safetensors"
|
||||
# model_filename = "vace14B_quanto_bf16_int8.safetensors"
|
||||
# model_filename = "c:/temp/movii/diffusion_pytorch_model-00001-of-00007.safetensors"
|
||||
# config_filename= "c:/temp/movii/config.json"
|
||||
self.model = offload.fast_load_transformers_model(model_filename, modelClass=WanModel,do_quantize= quantizeTransformer, writable_tensors= False) # , forcedConfigPath= config_filename)
|
||||
# model_filename = "c:/temp/phantom/Phantom_Wan_14B-00001-of-00006.safetensors"
|
||||
# config_filename= "c:/temp/phantom/config.json"
|
||||
self.model = offload.fast_load_transformers_model(model_filename, modelClass=WanModel,do_quantize= quantizeTransformer, writable_tensors= False)#, forcedConfigPath= config_filename)
|
||||
# offload.load_model_data(self.model, "e:/vace.safetensors")
|
||||
# offload.load_model_data(self.model, "c:/temp/Phantom-Wan-1.3B.pth")
|
||||
# self.model.to(torch.bfloat16)
|
||||
@ -92,8 +92,8 @@ class WanT2V:
|
||||
self.model.lock_layers_dtypes(torch.float32 if mixed_precision_transformer else dtype)
|
||||
# dtype = torch.bfloat16
|
||||
offload.change_dtype(self.model, dtype, True)
|
||||
# offload.save_model(self.model, "wan2.1_moviigen_14B_mbf16.safetensors", config_file_path=config_filename)
|
||||
# offload.save_model(self.model, "wan2.1_moviigen_14B_quanto_fp16_int8.safetensors", do_quantize= True, config_file_path=config_filename)
|
||||
# offload.save_model(self.model, "wan2.1_phantom_14B_mbf16.safetensors", config_file_path=config_filename)
|
||||
# offload.save_model(self.model, "wan2.1_phantom_14B_quanto_fp16_int8.safetensors", do_quantize= True, config_file_path=config_filename)
|
||||
self.model.eval().requires_grad_(False)
|
||||
|
||||
|
||||
@ -477,7 +477,7 @@ class WanT2V:
|
||||
pass
|
||||
overlap_noise_factor = overlap_noise / 1000
|
||||
latents[:, conditioning_latents_size + ref_images_count:] = latents[:, conditioning_latents_size + ref_images_count:] * (1.0 - overlap_noise_factor) + torch.randn_like(latents[:, conditioning_latents_size + ref_images_count:]) * overlap_noise_factor
|
||||
timestep = [torch.tensor([t.item()] * (conditioning_latents_size + ref_images_count) + [t.item() - overlap_noise]*(len(timesteps) - conditioning_latents_size - ref_images_count))]
|
||||
#timestep = [torch.tensor([t.item()] * (conditioning_latents_size + ref_images_count) + [t.item() - overlap_noise]*(len(timesteps) - conditioning_latents_size - ref_images_count))]
|
||||
|
||||
if target_camera != None:
|
||||
latent_model_input = torch.cat([latents, source_latents], dim=1)
|
||||
@ -598,4 +598,4 @@ class WanT2V:
|
||||
setattr(target, "vace", module )
|
||||
delattr(model, "vace_blocks")
|
||||
|
||||
|
||||
|
||||
|
||||
28
wgp.py
28
wgp.py
@ -43,7 +43,7 @@ AUTOSAVE_FILENAME = "queue.zip"
|
||||
PROMPT_VARS_MAX = 10
|
||||
|
||||
target_mmgp_version = "3.4.7"
|
||||
WanGP_version = "5.3"
|
||||
WanGP_version = "5.31"
|
||||
prompt_enhancer_image_caption_model, prompt_enhancer_image_caption_processor, prompt_enhancer_llm_model, prompt_enhancer_llm_tokenizer = None, None, None, None
|
||||
|
||||
from importlib.metadata import version
|
||||
@ -1525,11 +1525,12 @@ for path in ["wan2.1_Vace_1.3B_preview_bf16.safetensors", "sky_reels2_diffusion
|
||||
os.remove( os.path.join("ckpts" , path))
|
||||
|
||||
|
||||
wan_choices_t2v=["ckpts/wan2.1_text2video_1.3B_bf16.safetensors", "ckpts/wan2.1_text2video_14B_bf16.safetensors", "ckpts/wan2.1_text2video_14B_quanto_int8.safetensors", "ckpts/wan2.1_Vace_1.3B_mbf16.safetensors",
|
||||
wan_choices_t2v=["ckpts/wan2.1_text2video_1.3B_bf16.safetensors", "ckpts/wan2.1_text2video_14B_bf16.safetensors", "ckpts/wan2.1_text2video_14B_quanto_int8.safetensors",
|
||||
"ckpts/wan2.1_recammaster_1.3B_bf16.safetensors", "ckpts/sky_reels2_diffusion_forcing_1.3B_mbf16.safetensors", "ckpts/sky_reels2_diffusion_forcing_14B_bf16.safetensors",
|
||||
"ckpts/sky_reels2_diffusion_forcing_14B_quanto_int8.safetensors", "ckpts/sky_reels2_diffusion_forcing_720p_14B_mbf16.safetensors","ckpts/sky_reels2_diffusion_forcing_720p_14B_quanto_mbf16_int8.safetensors",
|
||||
"ckpts/wan2_1_phantom_1.3B_mbf16.safetensors", "ckpts/wan2.1_Vace_14B_mbf16.safetensors", "ckpts/wan2.1_Vace_14B_quanto_mbf16_int8.safetensors",
|
||||
"ckpts/wan2.1_Vace_1.3B_mbf16.safetensors", "ckpts/wan2.1_Vace_14B_mbf16.safetensors", "ckpts/wan2.1_Vace_14B_quanto_mbf16_int8.safetensors",
|
||||
"ckpts/wan2.1_moviigen1.1_14B_mbf16.safetensors", "ckpts/wan2.1_moviigen1.1_14B_quanto_mbf16_int8.safetensors",
|
||||
"ckpts/wan2_1_phantom_1.3B_mbf16.safetensors", "ckpts/wan2.1_phantom_14B_mbf16.safetensors", "ckpts/wan2.1_phantom_14B_quanto_mbf16_int8.safetensors",
|
||||
]
|
||||
wan_choices_i2v=["ckpts/wan2.1_image2video_480p_14B_mbf16.safetensors", "ckpts/wan2.1_image2video_480p_14B_quanto_mbf16_int8.safetensors", "ckpts/wan2.1_image2video_720p_14B_mbf16.safetensors",
|
||||
"ckpts/wan2.1_image2video_720p_14B_quanto_mbf16_int8.safetensors", "ckpts/wan2.1_Fun_InP_1.3B_bf16.safetensors", "ckpts/wan2.1_Fun_InP_14B_bf16.safetensors",
|
||||
@ -1548,12 +1549,12 @@ def get_dependent_models(model_filename, quantization, dtype_policy ):
|
||||
return [get_model_filename("ltxv_13B", quantization, dtype_policy)]
|
||||
else:
|
||||
return []
|
||||
model_types = [ "t2v_1.3B", "t2v", "i2v", "i2v_720p", "flf2v_720p", "vace_1.3B","vace_14B","moviigen", "phantom_1.3B", "fantasy", "fun_inp_1.3B", "fun_inp", "recam_1.3B", "sky_df_1.3B", "sky_df_14B", "sky_df_720p_14B", "ltxv_13B", "ltxv_13B_distilled", "hunyuan", "hunyuan_i2v", "hunyuan_custom"]
|
||||
model_types = [ "t2v_1.3B", "t2v", "i2v", "i2v_720p", "flf2v_720p", "vace_1.3B","vace_14B","moviigen", "phantom_1.3B", "phantom_14B", "fantasy", "fun_inp_1.3B", "fun_inp", "recam_1.3B", "sky_df_1.3B", "sky_df_14B", "sky_df_720p_14B", "ltxv_13B", "ltxv_13B_distilled", "hunyuan", "hunyuan_i2v", "hunyuan_custom"]
|
||||
model_signatures = {"t2v": "text2video_14B", "t2v_1.3B" : "text2video_1.3B", "fun_inp_1.3B" : "Fun_InP_1.3B", "fun_inp" : "Fun_InP_14B",
|
||||
"i2v" : "image2video_480p", "i2v_720p" : "image2video_720p" , "vace_1.3B" : "Vace_1.3B", "vace_14B" : "Vace_14B","recam_1.3B": "recammaster_1.3B",
|
||||
"flf2v_720p" : "FLF2V_720p", "sky_df_1.3B" : "sky_reels2_diffusion_forcing_1.3B", "sky_df_14B" : "sky_reels2_diffusion_forcing_14B",
|
||||
"sky_df_720p_14B" : "sky_reels2_diffusion_forcing_720p_14B", "moviigen" :"moviigen",
|
||||
"phantom_1.3B" : "phantom_1.3B", "fantasy" : "fantasy", "ltxv_13B" : "ltxv_0.9.7_13B_dev", "ltxv_13B_distilled" : "ltxv_0.9.7_13B_distilled", "hunyuan" : "hunyuan_video_720", "hunyuan_i2v" : "hunyuan_video_i2v_720", "hunyuan_custom" : "hunyuan_video_custom" }
|
||||
"phantom_1.3B" : "phantom_1.3B", "phantom_14B" : "phantom_14B", "fantasy" : "fantasy", "ltxv_13B" : "ltxv_0.9.7_13B_dev", "ltxv_13B_distilled" : "ltxv_0.9.7_13B_distilled", "hunyuan" : "hunyuan_video_720", "hunyuan_i2v" : "hunyuan_video_i2v_720", "hunyuan_custom" : "hunyuan_video_custom" }
|
||||
|
||||
|
||||
def get_model_type(model_filename):
|
||||
@ -1611,8 +1612,12 @@ def get_model_name(model_filename, description_container = [""]):
|
||||
description = "The SkyReels 2 Diffusion Forcing model has been designed to generate very long videos that exceeds the usual 5s limit. You can also use this model to extend any existing video."
|
||||
elif "phantom" in model_filename:
|
||||
model_name = "Wan2.1 Phantom"
|
||||
model_name += " 14B" if "14B" in model_filename else " 1.3B"
|
||||
description = "The Phantom model is specialized to transfer people or objects of your choice into a generated Video. It produces very nices results when used at 720p."
|
||||
if "14B" in model_filename:
|
||||
model_name += " 14B"
|
||||
description = "The Phantom model is specialized to transfer people or objects of your choice into a generated Video. It seems to produce better results if you keep the original background of the Image Referendes."
|
||||
else:
|
||||
model_name += " 1.3B"
|
||||
description = "The Phantom model is specialized to transfer people or objects of your choice into a generated Video. It produces very nice results when used at 720p."
|
||||
elif "fantasy" in model_filename:
|
||||
model_name = "Wan2.1 Fantasy Speaking 720p"
|
||||
model_name += " 14B" if "14B" in model_filename else " 1.3B"
|
||||
@ -1741,11 +1746,12 @@ def get_default_settings(filename):
|
||||
})
|
||||
|
||||
|
||||
if get_model_type(filename) in ("phantom_1.3B"):
|
||||
if get_model_type(filename) in ("phantom_1.3B", "phantom_14B"):
|
||||
ui_defaults.update({
|
||||
"guidance_scale": 7.5,
|
||||
"flow_shift": 5,
|
||||
"resolution": "1280x720"
|
||||
"remove_background_images_ref": 0,
|
||||
# "resolution": "1280x720"
|
||||
})
|
||||
|
||||
elif get_model_type(filename) in ("hunyuan_custom"):
|
||||
@ -4116,6 +4122,7 @@ def load_settings_from_file(state, file_path):
|
||||
return gr.update(), gr.update(), None
|
||||
|
||||
configs = None
|
||||
tags = None
|
||||
if file_path.endswith(".json"):
|
||||
try:
|
||||
with open(file_path, 'r', encoding='utf-8') as f:
|
||||
@ -4124,7 +4131,6 @@ def load_settings_from_file(state, file_path):
|
||||
pass
|
||||
else:
|
||||
from mutagen.mp4 import MP4
|
||||
tags = None
|
||||
try:
|
||||
file = MP4(file_path)
|
||||
tags = file.tags['©cmt'][0]
|
||||
@ -4138,7 +4144,7 @@ def load_settings_from_file(state, file_path):
|
||||
|
||||
prompt = configs.get("prompt", "")
|
||||
current_model_filename = state["model_filename"]
|
||||
model_filename = configs["model_filename"]
|
||||
model_filename = configs.get("model_filename", current_model_filename)
|
||||
model_type = get_model_type(model_filename)
|
||||
defaults = state.get(model_type, None)
|
||||
defaults = get_default_settings(model_filename) if defaults == None else defaults
|
||||
|
||||
Loading…
Reference in New Issue
Block a user