From X · Motion Transfer
Motion Transfer Prompts
Curated motion control and motion transfer video prompts collected from public posts on X. Try them in Oraclip with attribution to the original creators.
Alibaba rethinks character animation with Wan-Animate-2 Anyone who has t
Alibaba rethinks character animation with Wan-Animate-2 Anyone who has tried animating a character from a reference video knows the trade-off: you either lose identity, lose fine motion detail, or the compute cost becomes unbearable. Tongyi Lab, Alibaba's research team, just published a paper that tries to tackle all three problems at once, with a name that's already a statement: Wan-Animate-2. 🎬 The team frames the existing landscape around three paradigms and their flaws. Methods based on explicit motion representations suffer from extraction errors and identity drift, implicit motion features lose fine-grained dynamics through compression, and in-context learning avoids intermediate representations but incurs prohibitive computational costs. Wan-Animate-2 removes the intermediate motion extractor entirely, letting the driving video feed directly into a redesigned Diffusion Transformer. 📐 Architecturally, the framework conditions video generation on a reference image and a reference video through a dual-branch DiT, introducing Time-Align RoPE for temporal alignment between denoising video tokens and reference tokens, and Sparse-Ref Attention to selectively attend to the most informative reference features. 🕹️ One of the most interesting capabilities is text-driven viewpoint control, which decouples the output camera perspective from the driving video, a capability the authors describe as rarely supported by prior character animation methods that rely on explicit motion representations. ⚡ For interactive use cases, the team introduces Wan-Animate-2-Lite, an efficient variant designed to push inference latency down to real-time thresholds through a three-stage training paradigm combining teacher forcing pretraining with an error buffer mechanism and Self-Forcing distillation with chunk-wise backpropagation. According to independent reporting, this lite version reportedly reaches real-time streaming at 24 fps, opening deployment scenarios like digital avatars and live-streaming hosts that were previously infeasible at this quality level. Why it matters: if it holds up outside the demo reel, this changes the equation for anyone working with digital avatars, live streaming, or low-budget video production. No complex 3D rig, no expensive motion capture setup, just a reference video and an image. For companies and independent creators, it means access to professional-grade animation tooling without the hardware and pipeline barriers of traditional workflows, especially now that the team says it will release the Wan-Animate-2-Base model weights publicly. As a Creative & AI Specialist, what strikes me most isn't the visual quality itself, it's the decision to rethink the whole pipeline with live use cases in mind from day one: too many animation models stay stuck in offline mode because they were built for quality alone, while here the real-time variant looks like a design priority rather than an afterthought, and if the open weight release lives up to the paper, the real test will be how easily it slots into the workflows creators already rely on. 📱 Want AI updates on WhatsApp? DM me "AI" and I'll send you access. 🔔 Follow me for the last AI updates! #AInews #WanAnimate2 #Alibaba #CharacterAnimation #AIVideo
“Sube Conmigo (Classic House Mix)” takes the track into a classic house
“Sube Conmigo (Classic House Mix)” takes the track into a classic house direction, with lyrics written entirely by me. Visually, I imagined a futuristic world that, in many ways, we’re already living in. The visual concept was created with Midjourney 8.2 and animated with Kling AI Motion Control and Kling AI 3.0. Here’s a short glimpse from the video. Watch the full version on my YouTube and Instagram channels, and listen on Spotify, Apple Music and all major digital platforms. #ai #aiart #dance #musicvideo
Today I shipped motion control on PigenAI. Give it a character image + a
Today I shipped motion control on PigenAI. Give it a character image + a reference video, and it transfers the motion dance, gestures, movement - onto your character. Try it → https://t.co/AEEmbNI17U… #motion-control #aivideo #kling #pigenai https://t.co/L7lQE4oDgf
We’re almost there I took a photo of myself, found a dance video, and de
We’re almost there I took a photo of myself, found a dance video, and described the moves in a prompt. I attached my photo, and Seedance 2.5 generated this video of me dancing I also tried my photo and the reference video with Kling 3.0 Motion Control. It lost facial details and added lots of artifacts. Seedance gave me a much better result I regularly test the latest video models, and results like this make me think we’re almost there
Meet Lady Florentine 👑✨ An aristocrat with impeccable Rococo taste... an
Meet Lady Florentine 👑✨ An aristocrat with impeccable Rococo taste... and an undeniable obsession with filming herself in her lavish bedroom while dancing to the latest TikTok trends. Some things never go out of style. Made with @Kling_ai motion control https://t.co/NB4MdvymUp
The weirdest way to make AI copy a dance better? Take the dancer out of
The weirdest way to make AI copy a dance better? Take the dancer out of the video. Sounds backwards. But this simple trick can dramatically change how motion transfer works in Seedance 2.0. When you give a video of someone dancing directly to a video model, you’re not giving it “motion.” You’re giving it everything: The person. Their face. Their clothes. The room. The lighting. The camera. The background. And somewhere inside all that noise — the choreography. Now ask the model to put that movement onto a completely different character. You’re essentially asking it to solve ten problems at once. And then we wonder why the choreography drifts. The trick is to separate two things that shouldn’t be mixed: WHO is moving and HOW they are moving. Here’s the workflow: 1. Lock the character first. Create the character reference in GPT Image 2 — face, body, clothing, overall identity. The goal is simple: the character’s identity should exist independently from the movement. 2. Strip the dancer out of the source video. Convert the dance footage into a depth map. Instead of giving the model the original performer, you give it spatial information — essentially a representation of where the body is in 3D space. No face. No clothes. No background. No visual identity. Just movement. 3. Give the model two separate references. Character sheet = identity Depth map = motion Now each input has one job. And that’s the important part. 4. Don’t let the references compete. The depth map shouldn’t tell the model who is dancing. The character reference shouldn’t tell it how to dance. You’re removing the ambiguity instead of asking the model to resolve it. That’s why this works. Raw footage contains an enormous amount of information, and the model has to decide what matters. A depth map dramatically narrows the problem. Instead of: “What am I supposed to copy from this video?” the model gets something closer to: “Here is the movement. Put this character inside it.” And that distinction matters. Because better AI video isn’t always about giving the model more information. Sometimes it’s about giving it less — but making every piece of information intentional. This opens up some interesting workflows: → Transfer choreography onto completely original characters → Create motion-capture-style results without a mocap suit → Preserve specific movements more accurately → Build character showcases from existing performance footage → Separate identity, motion and environment instead of letting them bleed into each other The character sheet answers: “Who is dancing?” The depth map answers: “How are they moving?” And keeping those two questions separate may be one of the most useful tricks for getting more reliable motion transfer out of AI video models. The model doesn’t necessarily need to become smarter. Sometimes you just need to stop giving it the wrong problem.
Reading this u'd think Hao made an ai video mocking Indian culture But a
Reading this u'd think Hao made an ai video mocking Indian culture But all he did was be disappointed that the silly dancing cow he wanted to see projected on a ferris wheel looked and danced different The cow changes outfits periodically, he has no control over it https://t.co/YmWh61ulWq https://t.co/TDDxNoPeAi
AI video gets expensive when you let the model guess everything. I teste
AI video gets expensive when you let the model guess everything. I tested a different Seedance workflow: Motion reference → open-source depth/control maps → structured generation → Seedance. Result: • fewer rerolls • better motion control • more consistent output • less credit wasted The interesting part isn’t the dancing cat. It’s that better control before generation can be cheaper than brute-forcing generations until one works. AI video shouldn’t be a slot machine. Give the model structure first.
Saw everyone making these 4×4 dance sheets and had to try it myself. Pip
Saw everyone making these 4×4 dance sheets and had to try it myself. Pipeline: GPT Image 2.5 → 16-pose sheet Grok Imagine (image-to-video) → 8s dance clip Prompts in the replies 👇 GPT Image prompt: Create a 4×4 dance pose sheet featuring the similar character in Genshin https://t.co/qwClcGcDfb
Playing with my son’s toy cars and magnetic tiles for some Seedance 2.5
Playing with my son’s toy cars and magnetic tiles for some Seedance 2.5 testing. The idea was simple: use real toy movement as a motion reference, then see how Seedance translates it into a cinematic real-world scene. A traffic jam made from toy cars became a full-scale road scene. A toy excavator clearing the jam became a massive construction vehicle pushing through real cars. A magnetic-tile house became an actual building attempt. What’s interesting is that Seedance doesn’t just copy the action. It picks up on camera angle, object dominance, timing, contact points, weight, direction, and even accidental scale cues. In one clip, the toy excavator looked huge because it was close to the camera and cropped at the edge of frame. Seedance translated that almost literally, making the real excavator tower over the cars. That’s the fascinating bit. The reference video becomes more than choreography. It becomes a kind of physical prompt: perspective, momentum, gravity, friction, hierarchy, and staging all baked into a few seconds of footage. Basically: tabletop previs -> real-world cinematic scene. Still messy. Still unpredictable. But honestly, that’s what makes it fun. Feels less like prompting now, and more like directing motion with whatever is lying around the house. Share this with someone who still thinks AI video is just about writing better prompts.
@techhalla Was curious about why use Motion control with Kling when you
@techhalla Was curious about why use Motion control with Kling when you can just use Omni on seed dance but... I see the wisdom 😜 and love it https://t.co/5TscZZougx
Unitree R1 dancing outdoors with smooth balance and clean motion control
Unitree R1 dancing outdoors with smooth balance and clean motion control. Humanoid robotics is getting more exciting every day. #UnitreeR1 #HumanoidRobot #RobotDance #Robotics #AIrobot https://t.co/kBCwYKqrMa
I tried generating a long dance video with scaiL-2: 30 FPS, 301 frames,
I tried generating a long dance video with scaiL-2: 30 FPS, 301 frames, around 10 seconds, 896×1600 resolution — basically a 9:16 vertical video. I also used Uni3C’s camera-motion module to get the automatic camera movement from the reference video. But honestly… this workflow https://t.co/VH1ad9nAbO
20 SEC SCENES - BETRAYAL 🎬 Created with @thankyouai_hq . Every shot was
20 SEC SCENES - BETRAYAL 🎬 Created with @thankyouai_hq . Every shot was planned with intention—from claustrophobic handheld close-ups to ultra-wide cinematic reveals, using fast-cut pacing, match-on-action transitions, and grounded choreography to build tension before the first https://t.co/SN9BDhXjlG
Whistle dance, 100% AI generated 🤯 Clean studio background, perfect for
Whistle dance, 100% AI generated 🤯 Clean studio background, perfect for e-commerce product showcase. One reference image is all you need. Full prompt in bio 👇 🛠️ https://t.co/PMtLq9hxhT #AIGC #AIVideo #Fashion #Ecommerce #WhistleDance #MangoAI https://t.co/BTRL02oT2U
OKAY SO WHILE EVERYONE IS BUILDING CLAUDE + OBSIDIAN SECOND BRAINS — MY
OKAY SO WHILE EVERYONE IS BUILDING CLAUDE + OBSIDIAN SECOND BRAINS — MY AI RE-DRESSING PIPELINE IS QUIETLY PRINTING $16,400 A MONTH Been deep in productivity workflows and local LLM notes lately, but the second machine running in the background is doing all the heavy lifting. Took static photos of real girls, used AI to completely re-dress them, and ran a motion-transfer script to make them dance in 4K sync. Sounds like a gimmick until you check the Stripe dashboard. $500 in GPU compute. $16,400 out. Two months. Subscriptions, clothing brand promos, and custom video requests — all stacking on top of each other while people spend 4 hours tweaking note-taking apps. The wild part? Dancing is just 1% of this. Once you master the photo-to-video re-dressing pipeline, you can put these models in any outfit, any location, doing literally anything. Infinite content leverage off a single PC setup. Curious if you want the step-by-step breakdown of this image-to-video workflow, or should I stick to Obsidian setups? Drop a comment below.
75 million views in 5 days. The dancer on that stage isn't real. This cl
75 million views in 5 days. The dancer on that stage isn't real. This clip was shared as an example of what AI character pages built with Kling motion control can pull. The person who posted it says they were making $100 a day a few months ago running a page like this and selling a low-ticket guide on Whop. That's the most honest income claim I found in the whole AI character trend. A lot of the money in this boom comes from teaching it.
Luxury skincare, entirely imagined with AI. ✨🧴 Built this skincare comme
Luxury skincare, entirely imagined with AI. ✨🧴 Built this skincare commercial with MidJourney, Nano Banana Pro, and Seedance. I used MidJourney and Nano Banana Pro to create a consistent visual identity for every environment, product shot, and beauty scene, then brought https://t.co/lxF2d7fyGd
Four grey mannequins holding a choreography. Same timing, same weight, n
Four grey mannequins holding a choreography. Same timing, same weight, now with faces, fabric and a European square around them. Seedance 2.5 on SeaArt, GPT-6 Astra. Motion transfer is the hard part. This holds it. #Ad #SeaArtAstra #gpt6astra @SeaArt_Ai @SeaArtCoLab https://t.co/aLJDIjENcX
Same moves. Same camera. My face. 🕺 Day 1/100 exploring AI: putting your
Same moves. Same camera. My face. 🕺 Day 1/100 exploring AI: putting yourself into the viral motion-transfer trend with Seedance 2.5. The prompt + full breakdown 👇 https://t.co/BUNm2VTxO0 https://t.co/DNUCtssKuk
Making viral dance videos is ridiculously easy now 🤯 Upload an image + a
Making viral dance videos is ridiculously easy now 🤯 Upload an image + a reference video into Higgsfield's Kling Motion Control and let AI do the rest. 💃🔥 AI WITH HUZAIFA #aiwithhuzaifa #AI #HiggsfieldAI #Kling #AIVideo #Viral https://t.co/liLtjddH6n
#InimbaMzansi I am glad uQhawe, Tsaki noPortia are putting the pieces to
#InimbaMzansi I am glad uQhawe, Tsaki noPortia are putting the pieces together regarding the AI video. Even though it was so obvious after uYanga put uAviwe in his place during the Matric Dance https://t.co/ivXzF9WOlW
Want to create an AI dance like this? 🕷️👙🔥 Try it here 👇 https://t.co/OE
Want to create an AI dance like this? 🕷️👙🔥 Try it here 👇 https://t.co/OEkiFPlfGh #SpiderGirl #Spidey #AIDance #ViralDance #RunwayModel #AIVideo https://t.co/Nbfhi3BjHp
we're going to need a CAPTCHA for influencers 😭 this AI Wednesday-style
we're going to need a CAPTCHA for influencers 😭 this AI Wednesday-style account pulled ~83.3M views across three videos of the same dance. different outfits, different rooms, nearly identical moves. you can try that format with your own AI character: → make a character image https://t.co/skLhyhlhMp https://t.co/x6dX8svd46