Hello,
I use Linux and I want to have an integrated environment where I can use all the modules. For example, Fooocus is only for producing photos, and if I want to produce videos, I have to install ComfyUI separately. Is there an environment where I can have all these features in one place?
Thank you.
While it’s not impossible to link multiple workflows together, why not just use software that supports both T2V and I2V?
Hello,
Thank you so much for your reply.
Do I need a desktop environment to run this program? My Linux server does not have a graphical environment.
(post deleted by author)
If a GUI isn’t necessary, I think using Diffusers for both T2V/I2V is the simplest approach.
Sorry, what I meant was, does the program itself require a graphical desktop environment to run?
does the program itself require a graphical desktop environment to run?
I think so. From the names of the software mentioned, I thought it was about GUIs…
Sure, but does it just run as a service? If yes, then there is no need for a graphical environment and it can be connected remotely.
Hmm… I don’t think SD.Next is suited for that kind of use. It’s maybe purely a desktop GUI. If you want to run T2V or I2V as a service, it’s probably easier to use ComfyUI CLI or Diffusers with shell scripts or Python. Using MCP is an option, but it would be overkill.
I will test it.
Hello,
Unfortunately, when downloading the WAN modules, I get the following error:
2025-10-11T09:25:10.187925Z WARN Reqwest(reqwest::Error { kind: Request, url: "https://transfer.xethub.hf.co/xorbs/default/8e191b93cc43f7a2c3ebcd8333d35cd90c140ef72cd8d564ae2dd5fe1e2c7894?X-Xet-Signed-Range=bytes%3D0-63159569&X-Xet-Session-Id=01K7997CS5PS8H1WXPQXVRRJAX&Expires=1760178120&Policy=eyJTdGF0ZW1lbnQiOlt7IlJlc291cmNlIjoiaHR0cHM6Ly90cmFuc2Zlci54ZXRodWIuaGYuY28veG9yYnMvZGVmYXVsdC84ZTE5MWI5M2NjNDNmN2EyYzNlYmNkODMzM2QzNWNkOTBjMTQwZWY3MmNkOGQ1NjRhZTJkZDVmZTFlMmM3ODk0P1gtWGV0LVNpZ25lZC1SYW5nZT1ieXRlcyUzRDAtNjMxNTk1NjkmWC1YZXQtU2Vzc2lvbi1JZD0wMUs3OTk3Q1M1UFM4SDFXWFBRWFZSUkpBWCIsIkNvbmRpdGlvbiI6eyJEYXRlTGVzc1RoYW4iOnsiQVdTOkVwb2NoVGltZSI6MTc2MDE3ODEyMH19fV19&Signature=dLBAzRwV19fmL3K6AXTyosRpPWPVPV4BkGU3CDc4tIk-OqPWa1eCOXt8MOyEAChN0vU4S2Pb9w13C286scqR9Mg-38p13C~IeCtvCn8A~y1Y1YQn7iEYW~06RVpjZCwnOg7oEQsp9bynM8zqI~6skGUJ~4Z4XYDVR76UMQ2fYbS~4yNl2Vd~TMFCijFU2Km-8xV1r3Xs5l5LHqJwJ41UZSUzZcLagYz4bHAfxNRSpMElmkEFreSn8pUHBscRN9m8AOGNwHRpHKbNvTiVM5JSDDEaboLkBunw-VV9QSH~bFVIlLIMH1EcExpYwWT14q4bRP8FNoZTPK77~u9sUnBX0A__&Key-Pair-Id=K2L8F4GPSG1IFC", source: hyper_util::client::legacy::Error(SendRequest, hyper::Error(IncompleteMessage)) }). Retrying...
I have two questions:
1- Can I download the module manually?
2- What other module can do the WAN job?
Thank you.
Avoiding the issue itself is easy, but Rust is throwing errors…?
I also encountered errors when using an old version of Candle before, so this might be a similar case. Just in case, I’ll ping @jsulz
Yes to both.
Use the Hugging Face tools. Bypass the flaky Xet path if needed. As of 2025-10-11 these methods are current.
pip install -U "huggingface_hub[cli]" hf_transfer
# Option 1: bypass Xet entirely
export HF_HUB_DISABLE_XET=1
# Option 2: try Rust downloader (faster on good links)
# export HF_HUB_ENABLE_HF_TRANSFER=1
# TI2V-5B (720p) essentials
hf download Wan-AI/Wan2.2-TI2V-5B-Diffusers \
--include "diffusion_pytorch_model-*.safetensors" \
"diffusion_pytorch_model.safetensors.index.json" \
"Wan2.2_VAE.pth" "models_t5_umt5-xxl-enc-bf16.pth" \
"config.json" "configuration.json" \
--local-dir ./Wan2.2-TI2V-5B
CLI supports --include/--exclude, --local-dir, and works with the env vars above. (Hugging Face)
snapshot_download) with patternsimport os
os.environ["HF_HUB_DISABLE_XET"]="1" # or try HF_HUB_ENABLE_HF_TRANSFER=1
from huggingface_hub import snapshot_download
snapshot_download(
repo_id="Wan-AI/Wan2.2-T2V-A14B-Diffusers",
local_dir="./Wan2.2-T2V-A14B-Diffusers",
allow_patterns=[
"transformer/diffusion_pytorch_model-*.safetensors",
"transformer/diffusion_pytorch_model.safetensors.index.json",
"vae/*","text_encoder/*","tokenizer/*","scheduler/*","model_index.json",
],
)
allow_patterns avoids downloading the whole repo; env vars control the transport backend. (Hugging Face)
The TI2V-5B and A14B cards expose ModelScope mirrors; use their commands when HF is rate-limited or Xet is blocked. (Hugging Face)
Pick by task first: text-to-video vs image-to-video. Then match VRAM and toolchain.
CogVideoX (Diffusers)
# docs: https://huggingface.co/docs/diffusers/en/api/pipelines/cogvideox
from diffusers import CogVideoXPipeline
import torch
pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-5b", torch_dtype=torch.bfloat16).to("cuda")
video = pipe("A tabby cat astronaut slowly walks on the moon, cinematic, 24 fps", num_frames=49, height=480, width=720).frames
Mochi-1 preview (Diffusers)
# docs: https://huggingface.co/docs/diffusers/en/api/pipelines/mochi
from diffusers import MochiPipeline
pipe = MochiPipeline.from_pretrained("genmo/mochi-1-preview", torch_dtype=torch.bfloat16).to("cuda")
out = pipe("Aerial drone shot over snowy mountains at golden hour", num_frames=49, height=480, width=832)
SVD i2v
# model card: https://huggingface.co/stabilityai/stable-video-diffusion-img2vid
from diffusers import StableVideoDiffusionPipeline
pipe = StableVideoDiffusionPipeline.from_pretrained(
"stabilityai/stable-video-diffusion-img2vid", torch_dtype=torch.float16
).to("cuda")
Hello,
Thanks again.
I have a question about manual download. Let’s say I want to download Wan-AI/Wan2.2-T2V-A14B-Diffusers. First, I create a directory in the sdnext/models directory. Now, do I need to download all the files and directories located at https://huggingface.co/Wan-AI/Wan2.2-T2V-A14B-Diffusers/tree/main?
Hey there!
These print statements are just warnings and shouldn’t impede your ability to finish a download. If you were unable to complete a download, was there an error statement or stack trace provided?
Hello,
Please forget that error. If I want to download the files manually, which files should I download?