Create with Alibaba Happy Horse model now! Try here 👉

Pricing

Create with Alibaba Happy Horse model now! Try here 👉

COMMUNITY PAGE

Run Moonvalley Marey on Floyo

Home / Model / Moonvalley Marey on Floyo

AI VIDEO GENERATION

Run Moonvalley Marey on Floyo

The world's first commercially safe AI video model. Trained on licensed data only. Pose transfer, motion transfer, image-to-video, and text-to-video with cinematic camera control. 1080p at 24fps. Built for professional filmmakers.

Run Moonvalley's Marey through ComfyUI in your browser. No API key, no installs, no local GPU.

Resolution

1080p @ 24fps

Training Data

100% licensed (no scraping)

Duration

Up to 10 seconds

Key Feature

Pose + motion transfer

Try Marey Now → Browse All Models

No installation. Runs in browser. Updated May 2026.

What to get?

Moonvalley Marey (Realism v1.5) is the world's first commercially safe AI video model, trained exclusively on licensed, high-resolution footage. No scraped content, no user submissions, no legal gray zones. Four generation modes: text-to-video, image-to-video, pose transfer (human performance replication), and motion transfer (apply movement from one video to new content). 1080p at 24fps, up to 10 seconds. 3D-aware camera control, trajectory manipulation, studio-grade lighting, and multi-reference compositing. Built by former DeepMind researchers in collaboration with Hollywood filmmakers. Available as ComfyUI API nodes on Floyo.

MAREY WORKFLOWS ON FLOYO

Moonvalley Marey Text to Video

Moonvalley Marey Image to Video

Moonvalley Marey Pose Transfer Video

Moonvalley Marey Motion Transfer Video

What is Moonvalley Marey?

Marey is Moonvalley's foundational AI video model, launched July 2025 with general availability. It is the first AI video generator trained entirely on licensed, high-resolution footage. No scraped internet content. No user-generated submissions. About 80% of the training data came from filmmakers and agencies who intentionally licensed B-roll and footage through partnerships with platforms including Vimeo.

The model was built by former researchers from DeepMind, Google, Meta, and Microsoft, working alongside Oscar-nominated filmmakers and Hollywood VFX veterans. Ed Ulbrich (whose credits include Titanic, The Curious Case of Benjamin Button) joined Moonvalley's executive team specifically because of Marey's legally clean foundation. The model has been used in pilot programs by major studios and ad agencies, and was deployed for a Carl Sagan documentary through Moonvalley's studio Asteria.

Marey is not a standard text-to-video model. It uses a hybrid approach where directors can manipulate specific parameters for style, camera movement, blocking, and cinematic elements. The model has 3D awareness, so camera motion feels like real cinematography: pan, zoom, push-in, crane, and near-360-degree motion all maintain spatial coherence. Pose transfer replicates human performances from reference video. Motion transfer applies movement patterns to new content.

The current version is Marey Realism v1.5. Output is 1080p at 24fps, up to 10 seconds per generation (extendable). Aspect ratios span widescreen to vertical. The recommended prompt structure follows a cinematic framework: [Camera Movement] + [Scale/Perspective] + [Core Visual] + [Environmental Details] + [Lighting/Technical Specs].

On Floyo, Marey runs through ComfyUI API nodes. Four workflows cover text-to-video, image-to-video, pose transfer, and motion transfer. No Moonvalley account or API key management required.

What are Moonvalley Marey's technical specifications?

Marey Realism v1.5 generates 1080p video at 24fps for up to 10 seconds per clip. 3D-aware camera control with trajectory manipulation. Trained exclusively on licensed, high-definition content. Four modes: T2V, I2V, Pose Transfer (human performance), and Motion Transfer (movement patterns). Supports reference images as start frames, negative prompts, and seed control for reproducibility.

Spec	Details
Developer	Moonvalley AI Inc. (Toronto, Los Angeles, London)
Model Version	Marey Realism v1.5
Training Data	100% licensed, high-resolution footage (~80% from filmmakers/agencies via Vimeo partnerships)
Resolution	Up to 1080p (with upscaling options)
Frame Rate	24fps
Duration	Up to 10 seconds per generation (extendable)
Aspect Ratios	Widescreen to vertical
Modes	Text-to-Video, Image-to-Video, Pose Transfer, Motion Transfer
Camera Control	3D-aware: pan, zoom, push-in, crane, near-360-degree, trajectory manipulation
Pose Transfer	Preserves human movements, poses, and facial expressions from reference video
Motion Transfer	Applies movement and composition from reference video to new content
Additional Controls	Reference image for start frame, negative prompts, seed value, trajectory drawing
Commercial Safety	Commercially safe (licensed training data, legal assurance for commercial use)
Team	Former DeepMind, Google, Meta, Microsoft, TikTok, Disney, DreamWorks
ComfyUI Access	API-based nodes on Floyo (4 workflows)
Release Date	July 8, 2025 (GA)

What can you create with Moonvalley Marey?

Marey covers text-to-video generation, image animation, pose transfer (replicate human performances), motion transfer (apply movement patterns to new scenes), cinematic camera control, trajectory-based object movement, previsualization, storyboarding, B-roll generation, and commercial ad production. All output is commercially safe for professional use with no copyright risk from training data.

Capability	What It Does	Use Case
Pose Transfer	Preserve an actor's movements, poses, and facial expressions from a reference video and apply them to a new generated video. Optionally set a reference image as the first frame.	Performance capture, acting previsualization, talent replacement
Motion Transfer	Apply movement and composition from a reference video to new content. Restyle footage while keeping timing and energy intact. A bison running becomes a car racing through the same environment.	Video restyling, physics-aware transformation, creative adaptation
Image-to-Video	Transform a 2D image into a 3D-aware cinematic video. Marey builds a 3D environment from the image so you can direct camera movement as if you are on set.	Product showcases, architectural walkthroughs, photo animation
Text-to-Video	Generate 1080p cinematic video from text prompts with camera control, lighting direction, and environmental detail. Uses a structured prompting framework for filmmaker-level control.	Short films, B-roll, storyboarding, commercial ads
Camera Control	3D-aware cinematic camera: pan, zoom, push-in, crane, tracking, near-360-degree motion. Specify in your prompt or use trajectory drawing for precise object/camera paths.	Cinematic sequences, dynamic product reveals, VFX shots
Pipeline Integration	Chain with other models in ComfyUI. Use Marey for pose/motion transfer, then upscale with Topaz, add voiceover with Fish Audio S2 or ElevenLabs. All in one workflow.	Multi-model production pipelines

What are Moonvalley Marey's key features?

Marey's feature set is designed for professional filmmakers, not casual content creators. The commercially safe training data, pose transfer, motion transfer, and 3D-aware camera control are capabilities that matter in production environments where legal clearance and directorial precision are both required.

Commercially Safe (100% Licensed Data)

Marey is the first AI video model trained entirely on licensed footage. About 80% came from filmmakers and agencies who intentionally provided B-roll through platforms like Vimeo. No scraped internet content. No user-generated submissions. This is why major studios and ad agencies have adopted it for pilot programs. Commercial outputs carry legal assurance that models trained on scraped data cannot provide.

Pose Transfer

Upload a reference video of a human performance. Marey preserves the actor's movements, poses, and facial expressions and applies them to a new generated video. Optionally set a reference image as the starting frame. The performance from one actor transfers to a completely different character or environment while maintaining the timing and physicality of the original.

Motion Transfer

Upload a reference video to define movement patterns. Marey keeps the timing and energy intact while letting you change everything else. A bison running through grass becomes a Cadillac racing through the same environment with realistic physics. The composition and motion rhythm carry over, but the content transforms completely.

3D-Aware Camera Control

Marey builds a 3D understanding of each scene. Camera motion feels like real cinematography because the model knows where objects are in 3D space. Pan, zoom, push-in, crane, tracking, and near-360-degree motion all maintain spatial coherence. You can specify camera movement in your prompt or use trajectory drawing for precise control.

Studio-Grade Lighting

Lighting responds with cinematic logic. Shadows fall naturally. Subjects brighten or darken with intent. Lighting continuity holds across frames so scenes feel like they were lit on a real set. Trained on high-resolution footage from professional cinematographers, the model understands how light interacts with surfaces, skin, and environments.

Filmmaker-Designed Workflow

Built through direct collaboration between AI researchers (from DeepMind, Google, Meta) and professional filmmakers. The structured prompting framework ([Camera Movement] + [Scale/Perspective] + [Core Visual] + [Environmental Details] + [Lighting/Technical Specs]) mirrors how cinematographers think about shots. This is not a chat prompt. It is a shot description.

How does Moonvalley Marey compare to other video models?

Marey's unique advantage is commercial safety: the only production-grade video model with fully licensed training data. HappyHorse 1.0 leads on arena ranking and joint audio. Vidu Q3 leads on duration (16 seconds) and native audio. Kling 3.0 Omni offers 4K at 60fps. Wan 2.7 leads on open-source flexibility. Marey is the pick when legal clearance matters as much as visual quality.

Model	Licensed Data	Pose Transfer	Duration	Resolution
Moonvalley Marey	100% licensed	Yes (human perf.)	10 seconds	1080p
HappyHorse 1.0	Not disclosed	SV2V (subject swap)	15 seconds	1080p
Vidu Q3	Not disclosed	Reference-to-video	16 seconds	1080p
Wan 2.7	Not disclosed	No	5-10 seconds	Up to 4K

Source: Moonvalley official website, BusinessWire press release (July 8, 2025), CineD review, Deadline Hollywood reporting, fal.ai model documentation, Curious Refuge review, and VP Land reporting as of May 2026.

How does Moonvalley Marey work?

Marey uses a hybrid generation approach that goes beyond standard text-to-video prompting. The model builds a 3D understanding of each scene, which enables physically accurate camera motion, lighting, and object interaction. Directors can manipulate specific parameters for style, camera movement, blocking, and cinematic elements alongside their text prompts.

For pose transfer, the model extracts skeletal pose data, facial expressions, and body dynamics from a reference video. These are applied to a new generation while the model handles physics, clothing dynamics, and environmental interaction. The original actor's timing and physicality transfer to a completely different character. An optional reference image can anchor the starting frame's appearance.

For motion transfer, the model extracts motion patterns, camera composition, and temporal rhythm from a reference video. New content is generated that follows the same movement logic but with entirely different subjects and environments. The physics are aware of the transformation: a running animal becomes a speeding vehicle with appropriate environmental responses (dust, wind, weight).

On Floyo, Marey runs through ComfyUI API nodes. Each workflow accepts a text prompt, optional reference video (for pose or motion transfer), optional reference image (for start frame), negative prompts, and seed value. Results are MP4 video at 1080p/24fps. You can chain Marey with other models: upscale with Topaz, add voiceover with ElevenLabs or Fish Audio S2, or use Marey's output as input to another generation.

Fair warning: Marey is API-based, not open source. Pose transfer works for human performances only; animal motion will not be preserved. Temporal consistency can break down in complex scenes with fast camera motion or multiple subjects. The model prioritizes cinematic quality over speed. Some reviewers note that prompt adherence varies: atmosphere and lighting are strong, while complex multi-action prompts may lose detail. API pricing applies through your Floyo API Wallet.

Frequently Asked Questions

Common questions about running Moonvalley Marey on Floyo.

Is Moonvalley Marey free to use on Floyo?

You can start with Floyo's free pricing plan. Floyo gives $0.25 in free API credits on signup. To continue using the service beyond the free tier, upgrade your Floyo pricing plan. Marey runs as an API node, so generation costs come from your API Wallet (separate from your plan's GPU time).

How do I run Marey without installing anything?

Open Floyo in your browser, search "Moonvalley" or "Marey" in the template library, and pick a workflow (T2V, I2V, Pose Transfer, or Motion Transfer). Click Run, write your prompt, optionally upload reference media, and generate. Floyo handles the ComfyUI environment and API connection. No local install, no Python setup.

Who made Moonvalley Marey?

Moonvalley AI Inc., a company based in Toronto, Los Angeles, and London. The team includes former researchers from DeepMind, Google, Meta, Microsoft, and TikTok, plus VFX veteran Ed Ulbrich (Titanic, Benjamin Button). Moonvalley is a member of the Creative Rights in AI Coalition. Marey launched July 8, 2025 with general availability.

Is Marey's output commercially safe?

Yes. This is Marey's primary differentiator. The model was trained entirely on licensed, high-resolution footage. No scraped content. No user-generated submissions. Moonvalley provides legal assurance for commercial use. This is why studios, ad agencies, and production companies have adopted it for professional work.

What is the difference between pose transfer and motion transfer?

Pose transfer replicates a human actor's movements, poses, and facial expressions from a reference video and applies them to a new character or scene. Motion transfer applies broader movement patterns and composition from a reference video to entirely new content (not limited to humans). Use pose transfer when you want to preserve a specific performance. Use motion transfer when you want to preserve a movement rhythm or energy.

Can I combine Marey with other AI models in one workflow?

Yes. Floyo runs ComfyUI, which lets you chain multiple models. Use Marey for pose/motion transfer, upscale with Topaz Video AI, add voiceover with Fish Audio S2 or ElevenLabs, generate reference images with Nano Banana. All in one pipeline, all in your browser.

Can Marey transfer animal performances?

No. Pose transfer preserves human performances only. Animal poses and movements will not be captured through the pose transfer workflow. For animal motion, use the motion transfer workflow instead, which applies broader movement patterns to new content regardless of subject type.

How should I write prompts for Marey?

Use the structured cinematic framework: [Camera Movement] + [Scale/Perspective] + [Core Visual] + [Environmental Details] + [Lighting/Technical Specs]. For example: "Slow push-in, medium close-up, a woman reading on a park bench, autumn leaves, warm afternoon light, shallow depth of field." Marey responds well to cinematographic language. Use negative prompts to steer away from unwanted results.

Try Moonvalley Marey on Floyo

The world's first commercially safe AI video model. Pose transfer, motion transfer, cinematic camera control, and 100% licensed training data. Run it in your browser.

Try Marey Now → Browse All Models