💀💋 DaSiWa-WAN 2.2 I2V 14B BoundBite v10 | Lightspeed | GGUF 💋💀
GGUF Version of the model.
This is a WAN 2.2 Model: You will need one pair of High + Low.
Version overview: https://civarchive.com/articles/23495/dasiwa-model-versions-and-timeline
🔮 Key Features:
🔥 LoRA-Free Generations
Generate high-quality videos without stacking Wan 2.2 LoRAs (unless you want adding spacial styles/concepts).☄️Fast: 4 step generation
💫Quality motions (less slowdowns, no pixelated hyper-motion)
🔞 NSFW and SFW + Extreme versatile (more build in concepts):
Enhanced anatomy + poses + framing
Better understanding of sexual concepts
🪄 Better prompt responsiveness
👘 Better understanding of anime/manga style composition
🪡 Q8 (FP16 base) precision
🚫 Do not use any extra speed-up (low step) LoRAs, this is baked in already
✅ Optimizations
❤️🔥 New level of prompt adherence
🧠 Enabling intuitive reasoning over motion, interaction, and causality (VBVR)
🧪 Best distillation (mixed distillation experts)
💫 Stable and fast motions - without sacrificing quality
🎥 Enhanced camera and motion control
🪄 Best in-class details even on low resolution
💎 Consistent details over the full video (almost never detail shift)
🎊 Better special effects from prompts
👁️🗨️ Better eye details
🎬 Highly dynamic video composition
👘 Strengthened visual consistency/understanding for anime
🧩 Even more stable LoRa compatibility
🫦 Better NSFW understanding, without raising bias too far
😵💫 Reduced some hallucinations
🌩️ Less flashy frames
🫨 No camera drift
🚫Excluded CLIP
🍒Workflow
Make sure to checkout my easy to use Workflows!
🍄LoRA's
Try first without additional LoRAs!
But: This checkpoint is not meant to replace all LoRAs, it is meant to:
Perform better overall at his own
As easy as possible to use
With LoRAs to be absolutely awesome
🪧Announcement
⚠️ Read the corresponding announcement.
📢 Make sure to check it out for in-depth information and a complex comparison!
🆕 New to WAN 2.2 I2V? - Check out my guide.
🛠️ Recommended Settings
Steps: 4
CFG: 1
Sampler/Scheduler: Euler/Simple, UniPC_BH2/Simple
Resolution up to 720p (native quality).
Add other LoRAs with 0.3 - 0.6 at first
16 or 24 fps, 81 or 97 frames ~ 5s
Dependencies
🩻 Known issues
🫦 Bites!
Tell me 🫵🫢
Approximate expected quality from quantization
This are my tests compared to a full fp16 safetensor checkpoint taking prompt- and visual satisfaction into account on my DaSiWa checkpoints.
Quantization - Quality estimation
Q8 ⭐⭐⭐⭐⭐ ~ like FP16/FP8+, excellent results
Q6 ⭐⭐⭐⭐ ~ like FP8, very good results
Q5 ⭐⭐⭐ ~ good to very good results
Q4 ⭐⭐ ~ medium-good results
Q2 ⭐ ~ poor results, only use if you have to
⚠️ Do not compare this with the unofficial quants of my checkpoint made by others, they are based on FP8 and not FP16 like my quants.
🩺 Fixes & Feedback
If you use LoRAs, try to respect the LoRA training triggers and try some versatile descriptions, most LoRAs will work with 0.3-0.6 (start with 0.3)
Raise LoRAs in little steps +0.1
Do not mass add LoRAs, just add 1 or 2 (x2 High+Low)
Negative prompting do not work with cfg 1, that's a limitation of speed-ups with cfg 1 (except you use NAG)
Low resolution (e.g.384x576) are for fast samples and will blur fine details, do a higher resolution if you want clear details
🪧❗ Test your comfyui-backend with this absolute basic test-workflow before asking about errors.
🖤 Why I Made This
I was tired of using all these massive list of LoRAs, just to get a remotely good result after 10 generations, consuming hours of time.
So I can just make my videos with 1 or 2 concept LoRAs without pushing 6 till 10 LoRAs (Low/High) into a generation.
This checkpoint is also my personal playground.
Closing words
🤩 I want to thank all the fantastic other creators who made super nice LoRAs and concepts to play with! Support that awesome creators by using their LoRAs and post to their gallery and share the meta-data!
⚠️ I made all this with permissions or open-source resources (the time it is incorporated).
I share as much insights as I can without compromising my work. I'm doing this for fun as my hobby and just do not want my hobby to be destroyed.
More details can be obtained in the corresponding announcements!
If you would like to contribute in my awesome (😉) checkpoint or willing to share resources I'll gladly give credit! Just contact me!
✅ All credits / resources are mentioned inside the announcements! - Since different versions may have different resources.
YOU are responsible for outputs as always! If you make ToS violating content and I get aware I WILL report this.
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
FAQ
Comments (53)
this checkpoint has been amazing so far. everything I've tried to make with it has had the most fluid movement i've gotten from any checkpoint yet, and doesn't degrade the stylization of my base images over time. works really well paired with DR34ML4Y lora as well, though on its own it's more than powerful enough to give great results. excited to test at some higher resolutions when i have time
It's been the best motion and understanding of all the models, however for me the final video seems choppy vs previous versions?
This is just amazing. The comparison post was all facts. Style - unchanged, quality - high, prompt adherence - high. One of the best existing WAN-based models so far.
any chance of T2V version of your models?
@darksidewalker
Absolutely impressed. The quality, movement and prompt adherence is fantastic compared to what I was using, and smoother transitions with frame to frame. I don't often post comments but just had to for this model, great work!
Thank you :)
The camera and motion control is much much better! Amazing work!
Outstanding update. Fixed a lot of problems from the previous update. The eyes and pupils don't shift anymore; better, more natural movement. This is hands down the peak of what wan 2.2 model can do.
Thank you for the kind words!
I already mentioned this in another post but I'll reiterate here: There is simply no better WAN2.2 model out there. Your previous version was already unquestionably the best but you just nuked any reason to use another model with this new release.
Magnificent job and I look forward to all your future models (WAN or otherwise).
Thank you very much!
Idk if it's just me but my outputs look like those micro movement live 2D animations. What should I do to make the characters be more animated and lifelike?
same here, see if anyone could help on this.
use 16fps or 24fps in video combine output, if still too slow, use a frame interpolation(rife 47 - 49) node, search on yt for more details on that
@magoeditz Got it. Will try this and report back on my results
Thanks for the quants! Got the Q6 running on my 9070xt
It can run on AMD card? :O
Thank you for the GGUF
I have been using your models from midnightflirt. Now trying synth seduction & this. Even gguf has great quality. No need to use the base models anymore. Can we have some model for Wan2.2Animate ? Is it possible to use these for them? I m still learning. But nevertheless, great work indeed.
There will be no WAN-Animate. Sorry. If you can run q8 I would suggest fp8 from me instead, faster and slightly better.
@darksidewalker I can run them but Im currently running Q4, cause the quality decrease is hardly visible with your models plus the quick loading. Perhaps after awaited upgrade on my config, I ll get back to using FP8 as I used to with midnightflirt. Really great work though.
this model is realy great!
after first test i immediately discarded the original haha
i also just saw the s2v model you posted definitly going to try that one! was always interested in this
but the time limit of wan is still a problem so i wonder i f you ever thought about adding vace to your models for better merging/extending the individual videos to a longer?
lol just saw i use the fp8 safetensors version not the gguf <.<
but the questione still remains haha
Since there are already more modern things than VACE, it will not be added on the list.
@darksidewalker can you recomend something else then vace?
just a name i can search for it myself if i know what is good
@Shenzi SVI Pro 2.0 or LTX23 video extending is much better than using VACE.
Both are possible with workflows from me.
@darksidewalker thanks alot!
i will check out ^^
Looks good but Tastysin has better understanding of the prompt,camera handling and best poses, idk why since i use the same workflow, both Q8 models and text encoder Q5 KM
Nahhh, not. 😏 You just did not read the announcement and using it wrong.
oh i see why, then i will keep using tastysin then
@rashield710 If it is what gives you best result. Valid.👍
I really love the quality of this model! However, I'm struggling with fluid placement during intimate scenes. The white liquid often generates around the character's mouth, even when using negative prompts.
I want the leaking effect to be strictly localized to the lower body/genitals only, while keeping the face perfectly clean. Does anyone have a short prompt structure or a specific technique to 'lock' the fluid to the bottom area? Thanks for any tips!"
As the announcement states there are no "liquid altering" techniques inside this checkpoint this time. This is basic WAN 2.2 behavior. You want to use a lora for this.
just wondering why u dont put fp16 models and post fp8 only?
coz gguf's r from fp16 right?
All my models are from fp16 base at some point. And I answered the question more than once why I do not make the fp16 available.
I've tested this just a little so far but so far I like it. I'd been using Tasty Sin v8. Does this model pretty much replace it?
Thank you so much, it took me a little while but I've finally learned how to use your "workfl0w", especially the "switchs" to use files ending in GGUF
Oh! Thank you so much!
💀💋 DaSiWa-WAN 2.2 I2V 14B SynthSeduction v9 | Lightspeed | GGUF 💋💀 (((q8 - fp16))) 💋
is the best thx
"I made a 6-second video, and I noticed the start and end frames are always identical. Does anyone know why?
change the seeds to randomize.
"I made a 6-second video, and I noticed the start and end frames are always identical. Does anyone know why?
6s with 24fps would be a loop for wan22
Wan 2.2 almost always loops videos longer than 5s or just anything beyond the number ~100 in the "length" setting.
a lot of things are in literal slowmotion, no matter what. apart from that it looks really good and consistent. does anybody have a workaround the slow motion?
Yeah i wish i knew answer to that question too, both BoundBite and SnatchKiss High models are like that. SynthSeduction meanwhile do not have that issue.
I had not yet tried tweaking with model shift but it did seemed to be responsible for speed of movement (among the things), perhaps BB and SK need different value than 5, compared to SS.
Or perhaps BB and SK just can't do stuff in 12 frames. You can always use High model from SS and BB or SK Low model
*my english is outright terrible.
Holy moly this model is fast, 3 minutes for a 720p 5 second clip at 16 fps. Using a Rtx 5080😊
I've been using TastySin for the past few days coz I didn't realize it was outdated compared to your newer models. I'm trying BoundBite now but it doesn't seem to adhere to the same prompts as well as TastySin does. I'm testing a nice cum shot scene that I already got good results from TastySin but the result was totally different and the cum turned into a yellow custard. Both of these using the Q5 models without loras.
Fan of your work though. I just started last week and TastySin was far better than the FP8 Remix model.
You weren't lying. TastySin has way better prompt understanding and doesn't give that blurry gta3 trail effect that BoundBite gives.
Are there any prompts that keep the facial expression static? Or ones that make the mouth only keep moaning?