正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

We just tested what might be the most accurate motion transfer workflow we've seen so far. It's a little more involved than a standard workflow, but the results are pretty impressive. Here's the basic workflow: 1. Start with your reference footage. Create character and environment references. 2. Convert the... reference footage into a depth map (with Depth-Anything tool on Fal: 3. (Optional) Layer on additional guidance like colorized character masks (with Meta's Segment Anything: and or skeleton poses (with 4. Feed the depth map into Seedance 2.0 Omni as your motion reference along with your character and environment references. 5. Generate your final video. After testing several variations, the depth map workflow consistently produced the cleanest and most accurate motion transfer. If you're trying to recreate specific performances or complex movement, this is one of the closest AI workflows we've seen to achieving production-quality results.show more

Curious Refuge

22,578 subscribers

30,770 次观看 • 1 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

9 条评论

NuoNuo YSN 的头像

NuoNuo YSN1 个月前

Why not just upload the original video as a reference instead of processing it first

BongBong 的头像

BongBong1 个月前

So shouldn't the generative services be creating a depth map at the same time as the generated video? Might be something they should offer.

DENJIN 的头像

DENJIN1 个月前

The problem with the skeleton workflow (in addition to the occasional glitchiness) was that it was overlayed on top of the depth map footage and then being fed to the AI, which is why you were getting clones/duplicates in the output.

W. S. Waldo 的头像

W. S. Waldo1 个月前

Excellent video. Detailed. Thorough. And I like that you didn't show everything magically working. The failures are more informative than the successes. Thanks for this!

AI Mastery Guide 的头像

AI Mastery Guide1 个月前

Depth maps making motion transfer this clean is genuinely impressive work 🎥

Moez Zhioua 的头像

Moez Zhioua1 个月前

Using depth maps as the motion cue feels right, I noticed far fewer artifacts than pure skeleton feeds. The extra conversion step adds overhead but the consistency is worth it. Rule: let depth drive the transfer

Armaan | AI Systems 的头像

Armaan | AI Systems1 个月前

the reason depth-first beats skeleton-only tracking: depth maps carry occlusion and body volume, not just joint positions, so the generator has real spatial info instead of guessing what's behind an arm or how a character fills out a frame. stacking depth + segmentation + pose is multi-signal conditioning — removing ambiguity before generation even starts, not just throwing more tools at it.

Dir-BillSun 的头像

Dir-BillSun1 个月前

The depth pass preserves something skeleton guidance keeps losing here: the actor’s weight through turns and occlusions. I’d test the face separately—the body can be accurate while the intention of the performance disappears.

Pranav Joshi 的头像

Pranav Joshi1 个月前

The annoying part is that the workflow is still pretty involved, but the result makes it look worth it. Motion transfer is getting scary good.

相关视频

I've been playing more with this depth map motion transfer workflow, and if you're trying to recreate really specific body movements from a reference video, this works surprisingly well. Here's the workflow: 1. Grab your reference video, then create your character and environment references. 2. Run the video through Depth Anything on fal to generate a depth map: 3. Drop that depth map into Seedance Omni as your motion reference along with your character and environment references. You don't even need the original video. I also tested this against using the original video as the motion reference in Seedance, and there's definitely a noticeable difference. One thing I found, though, is that this works much better for body movement than facial performance since the depth map can't really capture subtle facial motion. I'm running more tests to see if there's a good way to control both body movement and facial performance. If you have any workflows I should try, do share in the comments!

I've been playing more with this depth map motion transfer workflow, and if you're trying to recreate really specific body movements from a reference video, this works surprisingly well. Here's the workflow: 1. Grab your reference video, then create your character and environment references. 2. Run the video through Depth Anything on fal to generate a depth map: 3. Drop that depth map into Seedance Omni as your motion reference along with your character and environment references. You don't even need the original video. I also tested this against using the original video as the motion reference in Seedance, and there's definitely a noticeable difference. One thing I found, though, is that this works much better for body movement than facial performance since the depth map can't really capture subtle facial motion. I'm running more tests to see if there's a good way to control both body movement and facial performance. If you have any workflows I should try, do share in the comments!

Curious Refuge

16,487 次观看 • 1 个月前

I wanted to experiment with the image-to-video workflow on Pixmax using Seedance 2.5, and the results genuinely impressed me🔥 I created this short Lay’s commercial from just a couple of reference images, building a playful family story around them. What I really like about the workflow is the control over the environment, product and storytelling while still getting natural-looking motion.

I wanted to experiment with the image-to-video workflow on Pixmax using Seedance 2.5, and the results genuinely impressed me🔥 I created this short Lay’s commercial from just a couple of reference images, building a playful family story around them. What I really like about the workflow is the control over the environment, product and storytelling while still getting natural-looking motion.

Isabella

47,320 次观看 • 18 天前

ACTORS ARE COOKED WITH GEN AI.... NO! ACTORS WILL COOK WITH GEN AI! Kling Just rolled out Kling 3.0 with Omni and motion control models. The new motion control gives you really detailed performance capture so you can transfer a genuine nuanced human performance onto your AI characters. You can also blend that performance in with AI generated events , using Kling's start and end frame workflow. So your character can interact with the AI generated environment, or fly or transform into something else. Actors now it the time to get involved. Kling AI P.S Sorry for my G.C.S.E level acting, for reference only 😂

ACTORS ARE COOKED WITH GEN AI.... NO! ACTORS WILL COOK WITH GEN AI! Kling Just rolled out Kling 3.0 with Omni and motion control models. The new motion control gives you really detailed performance capture so you can transfer a genuine nuanced human performance onto your AI characters. You can also blend that performance in with AI generated events , using Kling's start and end frame workflow. So your character can interact with the AI generated environment, or fly or transform into something else. Actors now it the time to get involved. Kling AI P.S Sorry for my G.C.S.E level acting, for reference only 😂

Uncanny Harry AI

25,521 次观看 • 6 个月前

This is getting ridiculous. You need to try this AI filmmaking workflow. My Blender reference is embarrassingly simple. You can drive a precise multi-character shot with nothing but basic shapes. 1. Generate a start frame in Midjourney 2. Block out the scene in Blender with boxes, animate the camera 3. Feed both to Seedance This feels like a real unlock. Your 3D reference doesn't need to look good.

This is getting ridiculous. You need to try this AI filmmaking workflow. My Blender reference is embarrassingly simple. You can drive a precise multi-character shot with nothing but basic shapes. 1. Generate a start frame in Midjourney 2. Block out the scene in Blender with boxes, animate the camera 3. Feed both to Seedance This feels like a real unlock. Your 3D reference doesn't need to look good.

Reid Hannaford

246,646 次观看 • 2 个月前

Depth maps just made motion transfer insane 😳🔥 Here's the prompts I used: Try it out on Higgsfield: Depth maps give AI a 3D read of your subject, so movement transfers with accurate volume, spacing, and occlusion instead of flat guesswork. That means character animation, dance transfer, and camera moves that hold their shape frame to frame. A serious upgrade for filmmakers, animators, and creators building AI video. #DepthMaps #MotionTransfer #AIVideo #CreativeAI #GenerativeAI

Depth maps just made motion transfer insane 😳🔥 Here's the prompts I used: Try it out on Higgsfield: Depth maps give AI a 3D read of your subject, so movement transfers with accurate volume, spacing, and occlusion instead of flat guesswork. That means character animation, dance transfer, and camera moves that hold their shape frame to frame. A serious upgrade for filmmakers, animators, and creators building AI video. #DepthMaps #MotionTransfer #AIVideo #CreativeAI #GenerativeAI

Rourke Heath

19,166 次观看 • 1 个月前

In this video I demonstrate #UnrealEngine5.4's brand new workflow to automatically retarget animations between common skeleton types. This tool makes it super easy to bring your animation into Unreal from outside sources. With just a few clicks, you can transfer animation two/from most industry standard skeleton types (Mannequins, Mixamo, Motionbuilder, Xsens etc...) with support for root motion and IK bones. If you want to make simple edits to the resulting motion on your character, you can optionally auto-generate an IK Retargeter to tune as desired.

Kiaran Ritchie

80,793 次观看 • 2 年前

A Chinese creator found a way to sell $3,000-$5,000 brand videos with no team, no camera crew, and no agency. no film crew. no editor. no storyboard artist. just one ComfyUI canvas turning product photos into finished AI ads. > GPT-image2 builds the director board: shots, camera moves, scenes, sound design > Seedance 2.0 turns the board into motion > product reference locks the object > character reference locks the face, outfit, and style > 16:9. 4K. one visual plan. > final prompt: generate the video from this storyboard old workflow: brand → agency → director → editor → revisions → invoice new workflow: product image → storyboard → reference sheet → video the expensive part was never rendering.

0xTria

15,847 次观看 • 2 个月前

A single 2D image → a 110,745-triangle 3D monster with 27 animations, running live in the browser. This is the Abyss Monster, built with img2threejs: The workflow: Image → 3D Reconstruction → Reference GLB + Specs → Rig → Animation Loops → VFX → Interactive Three.js 🔥 Inside this demo: 1/ 110,745 triangles reconstructed from the reference image 2/ 27 animation clips - Spin Flurry, Leap & Stomp, Charge, Collapse & more 3/ Reference-driven rig & animation workflow 4/ Dynamic lighting 5/ Real-time floating ember VFX 6/ Interactive WebGL character What I find most interesting is the workflow itself. The character and scene are built through a multi-prompt workflow, using visual references, GLB specs and animation loops to progressively construct the Three.js experience. Then, for the VFX, I used a single focused prompt to create the floating ember effect and integrate it into the scene. The final result is Three.js + TypeScript code - turning a visual reference into a programmable, interactive web experience. I’ll share the VFX prompt in the comments. 👇 #ThreeJS #WebGL #GameDev #TypeScript #Web3D #img2threejs

A single 2D image → a 110,745-triangle 3D monster with 27 animations, running live in the browser. This is the Abyss Monster, built with img2threejs: The workflow: Image → 3D Reconstruction → Reference GLB + Specs → Rig → Animation Loops → VFX → Interactive Three.js 🔥 Inside this demo: 1/ 110,745 triangles reconstructed from the reference image 2/ 27 animation clips - Spin Flurry, Leap & Stomp, Charge, Collapse & more 3/ Reference-driven rig & animation workflow 4/ Dynamic lighting 5/ Real-time floating ember VFX 6/ Interactive WebGL character What I find most interesting is the workflow itself. The character and scene are built through a multi-prompt workflow, using visual references, GLB specs and animation loops to progressively construct the Three.js experience. Then, for the VFX, I used a single focused prompt to create the floating ember effect and integrate it into the scene. The final result is Three.js + TypeScript code - turning a visual reference into a programmable, interactive web experience. I’ll share the VFX prompt in the comments. 👇 #ThreeJS #WebGL #GameDev #TypeScript #Web3D #img2threejs

Nick

11,302 次观看 • 11 天前

I just tested Motion Sync on Yapper and I like the cinematic Video made in minutes. Let me show you how you can do it 👇 Step 1: Open Step 2: Click on video and select Motion Sync. Then, ✅️Select base video Pick a video with the movement you want — this will drive the animation. ✅️Add reference image Upload your character or image you want to bring to life. ✅️Generate video Hit generate and watch your idea turn into a smooth motion sync clip. Simple, fast, and actually fun to experiment with.

I just tested Motion Sync on Yapper and I like the cinematic Video made in minutes. Let me show you how you can do it 👇 Step 1: Open Step 2: Click on video and select Motion Sync. Then, ✅️Select base video Pick a video with the movement you want — this will drive the animation. ✅️Add reference image Upload your character or image you want to bring to life. ✅️Generate video Hit generate and watch your idea turn into a smooth motion sync clip. Simple, fast, and actually fun to experiment with.

Leonardo

50,931 次观看 • 4 个月前

Ok, this is absurd. You can choreograph a complex action scene in Blender with basic shapes, then let Seedance make it real. You need to try this AI filmmaking workflow: 1. Generate a start frame in Midjourney 2. Block out the action in Blender 3. Feed both to Seedance My Blender reference was just rough timing, camera shake, and spatial choreography, and Seedance tracked the speed, motion, and action way better than I expected. This is the difference between describing a shot and directing one.

Ok, this is absurd. You can choreograph a complex action scene in Blender with basic shapes, then let Seedance make it real. You need to try this AI filmmaking workflow: 1. Generate a start frame in Midjourney 2. Block out the action in Blender 3. Feed both to Seedance My Blender reference was just rough timing, camera shake, and spatial choreography, and Seedance tracked the speed, motion, and action way better than I expected. This is the difference between describing a shot and directing one.

Reid Hannaford

527,505 次观看 • 2 个月前

Trying out a multi-character commercial vibe with a consistent product 🍓🍫 It used to be super tricky to keep different faces and product details stable in one video, but Seedream 5.0 Pro (images) + Seedance 2.5 (animation) handled it pretty smoothly. Dropping the full workflow (1 Product Map, 3 Character Maps, and the Animation prompt) in the thread below 👇 Which girl’s look is your favorite? ✨

Trying out a multi-character commercial vibe with a consistent product 🍓🍫 It used to be super tricky to keep different faces and product details stable in one video, but Seedream 5.0 Pro (images) + Seedance 2.5 (animation) handled it pretty smoothly. Dropping the full workflow (1 Product Map, 3 Character Maps, and the Animation prompt) in the thread below 👇 Which girl’s look is your favorite? ✨

Ivanna | AI Art & Prompts

13,164 次观看 • 24 天前

The simplest way to create an AI influencer using workflow (3 steps): Nano Banana → Veo 3 → Arcads AI Workflow Step 1: Build your character with a JSON prompt (Nano Banana Workflow) → Use a structured JSON prompt instead of writing random text → Define the look clearly: character, outfit, lighting, camera style, background → Add negative constraints like: no logos, no text, no extra fingers, no CGI skin, no warped mirror This makes the output more stable and reusable. You’re not just prompting. You’re building a repeatable image system Step 2: Turn image → video (Veo 3 Workflow) → Connect the Nano Banana output directly into Veo 3 → Add a simple motion prompt: slight body movement, natural head turn, subtle phone movement → Keep it minimal so it feels more real Workflow removes the manual back-and-forth. Step 3: Add script + voice (Arcads AI Workflow) → Feed the generated video into Arcads AI → Add a short script, usually 10–20 seconds → Generate voice + lipsync that matches the character vibe Now the character becomes actual content. Why workflow matters: Prompt → Image → Video → Voice No messy manual process. No repeating the same setup every time. Just a clean pipeline you can scale. Nano Banana + Veo 3 + Arcads AI + workflow feature = simple AI influencer factory

The simplest way to create an AI influencer using workflow (3 steps): Nano Banana → Veo 3 → Arcads AI Workflow Step 1: Build your character with a JSON prompt (Nano Banana Workflow) → Use a structured JSON prompt instead of writing random text → Define the look clearly: character, outfit, lighting, camera style, background → Add negative constraints like: no logos, no text, no extra fingers, no CGI skin, no warped mirror This makes the output more stable and reusable. You’re not just prompting. You’re building a repeatable image system Step 2: Turn image → video (Veo 3 Workflow) → Connect the Nano Banana output directly into Veo 3 → Add a simple motion prompt: slight body movement, natural head turn, subtle phone movement → Keep it minimal so it feels more real Workflow removes the manual back-and-forth. Step 3: Add script + voice (Arcads AI Workflow) → Feed the generated video into Arcads AI → Add a short script, usually 10–20 seconds → Generate voice + lipsync that matches the character vibe Now the character becomes actual content. Why workflow matters: Prompt → Image → Video → Voice No messy manual process. No repeating the same setup every time. Just a clean pipeline you can scale. Nano Banana + Veo 3 + Arcads AI + workflow feature = simple AI influencer factory

Alex Nguyen

35,608 次观看 • 5 个月前

The previs workflow is still leveling up! My AI filmmaking process is starting to feel less like wrangling AI and more like animation. Still just cubes in Blender, but now I’m moving the character through space and doing a handheld camera follow. Same basic process: 1. Generate a start frame in Midjourney 2. Block out the scene in Blender, animate the camera 3. Feed both to Seedance Seedance nailed the gritty handheld feel and followed the character movement around the car way better than expected. You can get way more ambitious with your directing than I thought.

The previs workflow is still leveling up! My AI filmmaking process is starting to feel less like wrangling AI and more like animation. Still just cubes in Blender, but now I’m moving the character through space and doing a handheld camera follow. Same basic process: 1. Generate a start frame in Midjourney 2. Block out the scene in Blender, animate the camera 3. Feed both to Seedance Seedance nailed the gritty handheld feel and followed the character movement around the car way better than expected. You can get way more ambitious with your directing than I thought.

Reid Hannaford

98,114 次观看 • 2 个月前

Seedance 2.0 vs Gemini Omni, tested under the same conditions 👀 I was working on a new storyboard. After generating it with Seedance 2.0, I gave the exact same prompt + storyboard reference + character reference to Gemini Omni as well. Result: Gemini Omni surprised me with style quality and got closer than I expected in prompt adherence. But Seedance still feels ahead for storyboard execution, motion energy, camera language and environmental interaction. Gemini looks good. Seedance feels directed. You can see them in the video and the storyboard sheet is shown at the end. Which one do you think is more useful for this kind of work?

Seedance 2.0 vs Gemini Omni, tested under the same conditions 👀 I was working on a new storyboard. After generating it with Seedance 2.0, I gave the exact same prompt + storyboard reference + character reference to Gemini Omni as well. Result: Gemini Omni surprised me with style quality and got closer than I expected in prompt adherence. But Seedance still feels ahead for storyboard execution, motion energy, camera language and environmental interaction. Gemini looks good. Seedance feels directed. You can see them in the video and the storyboard sheet is shown at the end. Which one do you think is more useful for this kind of work?

Kōda

36,082 次观看 • 3 个月前

I just built a Claude skill that makes animated ads in any style (Stop Motion, Vox, Origami...) Paste any design you like. It replicates the whole style. Here's what it does: - You paste a reference image, or just a brief. - It breaks the look down into an editable spec. - It generates the on-brand stills with Nanobanana 2. - You approve the ones you like. - It animates them into stop-motion with Seedance 2.0. The frame opens empty. The cut-out pieces slide in and snap into place. Everything runs in Claude through the connector, so you never juggle models. Your only job is taste. Paste a reference, pick the shots. I'm giving the whole skill & workflow away. Comment "Skill" and I'll send it to you.

Raph Guilhem

61,821 次观看 • 1 个月前

🎬 Want to create cinematic AI videos in seconds? Here’s exactly how I do it using Seedance 2.0 on Yapper: Most people get this wrong… your results depend on how well you *direct* the AI. Here’s the workflow 👇 ✅ Step 1: Open Yapper Go to Create → Video ✅ Step 2: Choose your model Select Seedance 2.0 This model delivers smooth motion, dynamic camera angles, realistic lighting, and strong scene consistency. ✅ Step 3: Write a strong prompt Your prompt is your director. Be specific: Camera movement (pan, zoom, tracking shot) Lighting (golden hour, neon, soft shadows) Environment (city, desert, luxury interior) ✅ Step 4: Add a reference image (optional) Use this to guide style, composition, or subject for more control and consistency. ✅ Step 5: Hit Generate In seconds, your idea turns into a cinematic sequence. 🚀 Pro tip: If it’s not perfect, tweak your prompt — small changes = big difference. From idea → cinematic video in seconds.

🎬 Want to create cinematic AI videos in seconds? Here’s exactly how I do it using Seedance 2.0 on Yapper: Most people get this wrong… your results depend on how well you direct the AI. Here’s the workflow 👇 ✅ Step 1: Open Yapper Go to Create → Video ✅ Step 2: Choose your model Select Seedance 2.0 This model delivers smooth motion, dynamic camera angles, realistic lighting, and strong scene consistency. ✅ Step 3: Write a strong prompt Your prompt is your director. Be specific: Camera movement (pan, zoom, tracking shot) Lighting (golden hour, neon, soft shadows) Environment (city, desert, luxury interior) ✅ Step 4: Add a reference image (optional) Use this to guide style, composition, or subject for more control and consistency. ✅ Step 5: Hit Generate In seconds, your idea turns into a cinematic sequence. 🚀 Pro tip: If it’s not perfect, tweak your prompt — small changes = big difference. From idea → cinematic video in seconds.

Kuria | AI

43,685 次观看 • 4 个月前

pro tip the easiest way to get exact character movements in nano banana pro + Veo workflow isn't through prompting instead of wrestling with text descriptions, try this: - generate your base character in Nano Banana Pro - upload your character + a reference screenshot of someone doing the exact pose/action you need - generate with a simple prompt: "make the second person match the pose/action of the first person" - animate in Veo using your scene-by-scene prompt structure (will attach it in guide under this post) this method is significantly faster, more accurate, and way less frustrating than trying to describe complex movements through text prompts alone

pro tip the easiest way to get exact character movements in nano banana pro + Veo workflow isn't through prompting instead of wrestling with text descriptions, try this: - generate your base character in Nano Banana Pro - upload your character + a reference screenshot of someone doing the exact pose/action you need - generate with a simple prompt: "make the second person match the pose/action of the first person" - animate in Veo using your scene-by-scene prompt structure (will attach it in guide under this post) this method is significantly faster, more accurate, and way less frustrating than trying to describe complex movements through text prompts alone

MAX

42,417 次观看 • 7 个月前

We just tested a lip sync workflow in Seedance 2.0… and it might actually compete with HeyGen and other avatar models 👀 You upload: – a reference image – a blacked out video with audio OR audio file – a prompt (camera motion + VO) The prompted VO is really important here to make sure the model doesn’t hallucinate or riff off of the original speaker verses. Feels like a legit breakthrough. Lip sync has been a pain point for a while, and this gets us a lot closer.

Curious Refuge

19,504 次观看 • 4 个月前

Seedance 2.0 is insane. Blender previs is having a moment in AI filmmaking for a reason. I wanted to see how far I could push the workflow beyond a simple camera move in an ambush scene. The workflow: 1. Make a start image in Midjourney 2. Block it out in Blender, animate the camera 3. Feed both to Seedance The more I test this pipeline, the more convinced I am that hybrid pipelines are the future. The conversation is moving fast but previs is one of the clearest unlocks so far. Seedance still isn't perfect at matching the reference but it's close. You can experiment with staging, timing, and camera movement before you ever touch the video model. Solve the scene before you generate the shot.

Seedance 2.0 is insane. Blender previs is having a moment in AI filmmaking for a reason. I wanted to see how far I could push the workflow beyond a simple camera move in an ambush scene. The workflow: 1. Make a start image in Midjourney 2. Block it out in Blender, animate the camera 3. Feed both to Seedance The more I test this pipeline, the more convinced I am that hybrid pipelines are the future. The conversation is moving fast but previs is one of the clearest unlocks so far. Seedance still isn't perfect at matching the reference but it's close. You can experiment with staging, timing, and camera movement before you ever touch the video model. Solve the scene before you generate the shot.

Reid Hannaford

226,194 次观看 • 2 个月前

We're giving away the prompt we used to make this AI UGC video. Getting Sora 2 to output hyper-realistic, consistent footage isn't just about the tool. The prompt has to be structured in a very specific way... character description, cinematography, camera motion, lighting, dialogue, audio, authenticity keywords. Get it wrong and the footage looks off. Get it right and everything downstream becomes easier. We’re sharing the exact Claude template we use as step one of our AI UGC workflow. It takes a basic brief (who the subject is, where they are, what they're saying, what device it should look like, their accent) and structures all of that into the exact format Sora 2 needs automatically. It sits at the start of a full workflow that runs through Sora 2 for the hook, ElevenLabs for voice cloning, ChatGPT for subject consistency, Nano Banana for B-roll generation and Kling for animation. The whole thing is broken down in the latest D2C Diaries episode. But this prompt is the foundation. Without it, the rest of the workflow is harder than it needs to be. If you're building in this space or testing AI UGC for your brand, this is worth having. Retweet this post and comment PROMPT and I’ll send it over.

Joe Marston | Ecom Growth

17,556 次观看 • 6 个月前