Seedance 2.0 is a next-generation unified multimodal AI video generator from ByteDance. Powered by a revolutionary Dual-branch DiT (Diffusion Transformer) architecture, it jointly generates visuals, dialogue, lip-sync, and ambient sound effects in a single pipeline.
Unlike traditional fragmented AI video tools, Seedance 2.0 eliminates post-production workflows by natively fusing text, image, and audio inputs into cinematic-grade content with perfectly synchronized audio and physics-accurate motion.
Description
Comments (15)
No idea how any of you guys are getting NSFW out of this. Won't take a NSFW starting image and no matter what I prompt NSFW-wise it will ignore that part completely and give me a normal video XD
Well, from the people who did it, they used only text to video. So image to video will not work I guess
@elevendr Iv tried a dozen very explicit NSFW prompts and ignores every nsfw part completely.
You can give it a NSFW image, it won't block it as long as it doesn't detect real people. Animeish /semirealistic images are fine.
When describing the prompt don't use clearly nsfw words such as penis, sex, etc. but word the prompt in a sfw way. Basically trick it into thinking it is doing something sfw. Instead of saying "cum dripping" say "cream dripping" or "viscous fluid dripping", the dude below managed to produce a video of a boxer hitting a pussy of a suspended girl by telling it is hitting a "human shaped punching bag". Basically break down what makes sex, e.g. "man's hip thrusts forward, the woman recoils" etc. and word it in a way it can understand and not censor it.
请教下这个大模型是如何使用？
Don't forget to add Seedance as a filtering option when searching the video gallery. Currently I can't go to Videos and search for Seedance videos.
Seedance 2 can also generate 1080p. Do you intend to add that option?
What about generation from reference images? Or using reference video and/or audio? I'd like to use Seedance on civitai but currently your feature implementation for Seedance is very limited. Or is it that I am using it wrong and these options are hidden somewhere?
Why haven't people made a Seedance distill LoRA for LTX 2.3, I'm pretty sure you can get lost of Seedance 2.0 outputs and train them to make a new distilled lora and use it in LTX 2.3, why hasn't anyone done that yet?
you need serious hardware to train a ltx 2.3 model with 15sec/720p video. even creating a lora for ltx 2.3 is very resource intensive and requires lots of vram and off loading.
Looks blurry. Framerate looks low.
Seadance 2.0 can do Multireference to video, and First/Last frame to video now but it isn't an option here. Please update it.
Are there plans to create LoRA components?
LoRA's for a closed source model? That's a great idea.
You can really see human creativity in the gallery below 😅
Hello,
I’m having an issue uploading images as inputs for the generator on Civitai.
When I try to add an image in tools like Seedance Image-to-Video or Grok Image-to-Image, the image upload never finishes and stays loading indefinitely. The generator then still shows “An image is required”, as if the image was never attached.
I tested multiple conditions and the problem persists:
different images
different browsers (Chrome and Firefox)
different operating systems (Windows and Arch Linux)
both civitai.com and civitai.red
my main account and also a newly created account
What is interesting is that I can upload images normally in a regular post, so the general site uploader seems to work. The problem appears to be specifically with the image input uploader used by the generators.
Also, previously used images sometimes load instantly, but any new image gets stuck loading forever.
Could you please check if there is an issue with the generator image-upload system or the input asset service?
Thank you.