๐ DaSiWa WAN 2.2 14B S2V - Lightspeed ๐
๐ฎ Key Features:
๐ฅ LoRA-Free Generations
โ๏ธFast: 4/6 step generation
๐ชก FP8 precision
โ ๏ธ Read "About this version" details for the version you are using for more information!
๐ซ Do not use any extra speed-up (low step) LoRAs, this is baked in already
๐กHint for best results: Make sure your audio length is as close as possible to the video length (frames/fps=seconds). Longer Audio is OK or cut the video at the end.
โ ๏ธ The first frame burn/over-saturation can not be patched. It is from the VAE. Only a workflow can counter this.
๐Workflow
Make sure to checkout my easy to use C-S2V Workflow with anti-burn, upscaling and โพ๏ธ true infinite length!
โ ๏ธ Read the corresponding announcement.
๐ข Make sure to check it out for in-depth information!
๐ ๏ธ Recommended Settings
Steps: 4-6
CFG: 1
Sigma Shift: 10
Sampler/Scheduler: Euler/Simple
Resolution up to 720p (native quality).
๐๏ธ Dependencies:
๐คHonourable Mentions ๐ค
For "LittleDemon"
Awesome initial AI voice and music was provided from: ๐ฆ YT channel Omega-VIII
The inspiration of the sample/front-character was made with Lyara ๐ซถ
Spacial thanks to S1LV3RC01N for help with quanting this model and with the repo! โ๏ธ
Disclaimer
This models are shared without warranties and with the condition that it is used in a lawful and responsible way. I do not support or take responsibility for illegal, harmful, or harassing uses. By downloading or using it, you accept that you are solely responsible for how it is used.
Custom License Addendum: Distribution Restriction
Notice: Notwithstanding the base license selected for this model, the following restrictive terms apply:
No Redistribution: You are not permitted to host, mirror, or redistribute this model (checkpoint, LoRA, or Safetensors files) on any other platform, website, or service (including but not limited to Hugging Face, Tensor.art, or SeaArt) without explicit written permission from the creator.
Attribution & Source: This model is officially maintained only on Civitai or other platforms where I explicitly own the repository. To ensure users receive the correct version, updates, and safety metadata, please point users to the original URL.
Usage: All other rights regarding the use of the model for image generation remain as per the terms and the restrictions provided per model.
Description
โ Optimisations
๐ CFGZeroStar patch (better results and prompt adherence)
๐ฐ Baked Latest Distillation (r64-1022)
๐ Additional concept optimisation (SFW/NSFW)
๐ฌ Reward Attention
๐ Bright and light optimisations
โจ Control optimisation
๐ Refined motions
๐ฐ Less tries for good results
๐ Zero prompt results capable
๐ซExcluded CLIP
๐๏ธ Dependencies:
๐ฉป Known issues
๐ She will try to get your treasures!
๐พ Sometimes can introduce noise-artefacts on low resolution samples
โ ๏ธ The first frame burn/over-saturation can not be patched. It is from the VAE. Only a workflow can counter this.
โ ๏ธ Requirements:
Comfyui โฅ 0.4.0
FAQ
Comments (22)
I'm having issues actually getting any results at all, even with shift on 50. I get small amounts of movement, but nothing much else. Any ideas what it could be?
Maybe you set no sound file?
is there any gguf model??
Not made.
Model Link: QuantStack/Wan2.2-S2V-14B-GGUF
I gotta upload a vid. Much improved furry cartoon mouths from the last version. Love it! โค๏ธ
ltx-2 video model whennnn legend?
Sounds great, right?! - But is is 1 day old :) Preparing a finetune needs time and testing. I will definitely look into it!
Also thanks for the heads up :)
@darksidewalkerย thanks for your work๏ผ
LTX2 is incredible. Even with my low-end specs, I can run 1024x1024 at 150 frames without running out of memory.
Well.. I tested a bunch of things with LTX-2 and ...
I'm not impressed, here is why:
Poor visual quality, extreme morphing, low sound quality, low temporal consistency.
I doubt it will be anyhow good soon.
I could not generate any good long video nor did I see any from the community.
Maybe I'm missing something at the moment, but for now it is not as good as tuned WAN 2.2 i2v or SVI.
@darksidewalkerย probly just have to wait for them to come out with somthing not so buggy as they usally do within the first week
@PastellPastellPastellIย We have to wait and see how much the community and dev collaborate on LTX-2, or if someone will just repurpose LTX-2 as a sound generator for the Wan model.
Hi master, the new version is not working, results are black screen videos, the previous version is working fine. I'm using your workflow. could it be issue of my settings?
Please update comfyui
any chance of getting a gguf version? this isn't gonna fit in my 12GB VRAM :(
When i run S2V+CozyVoice function,show error:
AudioEncoderEncode
Input type (float) and bias type (struct c10::Half) should be the same
Does anyone has the same issue?
Sorry,i solve by asking gemini3.I add --lowvram to run_nvidia_gpu.bat, then it can works.
is there a way to make it work with SLV workflow??
What do you mean by SLV?
Hello i use your workflow, what model is better to use MMAudio or Hunyuan Foley?
Totally depends what you want to achieve. mmaudio is good for nsfw sounds, foley for other ambient sounds. S2V for music and lipsync.