abokalypsis

I created a pipeline for Seedance 2.0 that perfectly makes 1:1 animation with 3D Omni ref inputs. This method requires comfyUI for img2img and blender or similar programs to create the 3D input video. Check replies. #CapCutSeedance2 @capcutapp

@GasperChia45367 @venturetwins Thanku! Yeah the previz was done in blender

@venturetwins @fatboypink God this is amazing control. @fatboypink was this using blender for the 3d?

AI video is insanely powerful in the hands of creative pros. This creator (@fatboypink) created a motion animation in 3D - and then used Seedance 2 to generate the final stylized render. They used the reference video + frames to guide the generation. Crazy cool!

@jfischoff 100%. Excited to see where this will be in a few months. Soon Claude will handle the animation too

@fatboypink time to build more

this would've been useful 5days ago 💀

bought twitter blue for this. This prompt guided most of the generations to complete this. it's nothing special. By keeping it simple, Seedance leaned heavily on the video input (previz render) to guide the output. Prompt: @ design of "Rari" and her grappling gun/ hook. @ design for the "hellbike." @ @ @ @ for the key frames. @Video1 as the master for all camera movement, character, and prop movement. Visual Mapping & Evolution: The video must start with the exact camera pan of @ , move to the high-fidelity extreme low angle shot of the hellbike blsting froward, then into a POV shot of the grappling hook aiming forward as the buildings crumble apart. In the distance, the bike 'blinks' with an anime spark. The video ends with the extreme close up eye-shot from @ . Following the camera path and shot sequence of @Video1 Environment & Detail: Replace all low-poly shapes in @Video1 with professional-grade Japanese anime background art. Inspired by @Env1 , the city has a strong perspective with oppressive urban depth. Cold blue-cyan exterior grade with faint sodium-white street spill, subtle frost bloom on windows, and soft cryo haze in the distance. As the bike blasts forward, chunks of frozen road blast backward like shrapnel — fist-sized pieces of ice and asphalt tumbling end over end toward camera. The air behind the bike distorts with shockwaves and velocity. A trench is carved into the road where the treads were. . Style & Quality: > High-quality Japanese anime film with fluid-like animation, sharp cel-shading, 24fps. Technical Excellence: > Masterpiece, 8k resolution, cinematic composition, high-fidelity visual, exquisite textures, smoke and wind animation with subtle secondary VFX (bike wind bursts and rubble smoke). Professional production quality. The bike digs a trench as it drives forward. Audio: No music; SFX only #dreaminaCPP @dreamina_ai

Thank you all for taking interest in my work. Let’s keep pushing AI filmmaking together 🫀

@InnerRefle11312 @capcutapp Essentially yeah. Theres rough animation is the main driver. The other references are character sheets and a still that represents the final look for several key frames in the animation. If you have 2-3 stills, Seedance can fill in the rest pretty well.

@fatboypink @capcutapp Loving the Vid2Vid! I have noticed some inconsistent effects with some of my work - what conditioning are you using? Start frames, character references and your rough animation?

I created a pipeline for Seedance 2.0 that perfectly makes 1:1 animation with 3D Omni ref inputs. This method requires comfyUI for img2img and blender or similar programs to create the 3D input video. Check replies. #CapCutSeedance2 @capcutapp

@zabiisuto @capcutapp Agreed. Theres so much potential for multiple workflows here. Instead of fighting AI, if more people embraced it we could be seeing new levels of creativity unfold. At the very least we can boost the efficiency of current pipelines

@fatboypink @capcutapp This is what I'm the most excited to see. Imagine what a team of stuntmen and actors could now achieve with 2D animation. Instead of reacting to animation, it can be the other way around.

@VahnAeris @capcutapp Thank you! I’m putting together a video now, so it should be up this weekend. I’m down to answer any questions though

@fatboypink @capcutapp good work man, would like to see a bit your pipeline and how you do this in blender ( I m still a mid begginer in the world of blender, especially video/animation ) Would like to create my own movie set/shots too. Awesome job, following this closely!

@venturetwins Thank you for sharing this.

Control > chaos Built the shot in 3D, then translated it 1:1 into final animation with #capcutseedance2 @capcutapp #dreaminaCPP @dreamina_ai

@capcutapp My entry! I hope you like it #CapCutSeedance2 @capcutapp x.com/fatboypink/sta…

🎉 CapCut Dreamina Seedance 2.0 Video Challenge is LIVE! Unleash your creativity with Dreamina Seedance 2.0 and make the most viral, high-quality video! Show off your creativity and win big: 🏆 Prizes: $10,000 – Best Overall $3,000 – 2nd place $1,000 – 3rd place $100 each – For the 10 videos with the most likes. 📌 Get Started: Create your Dreamina Seedance 2.0 video Create: Find Dreamina Seedance 2.0 in AI Lab (App) and Video Studio (Web) Post your video on X Tag #CapCutSeedance2 and @capcutapp ⏰ Challenge Period: Apr 1 - Apr 22 Don’t miss your chance to go viral and win! 🚀

@Darksky_46275 @capcutapp In this fashion, you could even say Seedance is a better image editor than nano banana. If you undercook the references, you can take the output it gives and refine one of the "undercooked" frames. Boom. Now you have another reference and can try the gen again.

@Darksky_46275 @capcutapp Reference images along with the shot. If you have at least 1 reference per shot, seedance can fill in the blanks. The more references you have, the more blanks are filled. I reccomend 2 references for each shot. One for the start (mandatory) and one for the middle or end.

@Darksky_46275 Thanks homie! This means a lot fr

@fatboypink Good luck. I will follow you and check occasionally. -Darksky

My discord account got banned, unfortunate. Currently appealing. Follow me here for updates while I sort this out.

SDXL controlNet workflow to render rough stills as anime screen caps. Requires a sketched over 3D still (can be rough, but the more refined the better) and a depth map