PixVerse Motion Control Mimic: Image-to-Video Motion Transfer on Models and API | RunComfy

PixVerse / PixVerse Motion Control Mimic#

This image-to-video model copies movement from a reference clip and applies it to a still subject. Upload one clear character image, provide a motion reference video, pick a quality tier, and generate a short animated clip without writing a prompt.

The workflow suits dance remakes, pose transfer, mascot animation, and social motion tests where the look comes from the still and the choreography comes from the reference. Teams keep identity locked to the image while exploring new movement takes on RunComfy models and API.

Highlights#

• Motion from video: Drive the still with a person-focused reference so limbs, timing, and staging follow the source performance.

• Still-led identity: Keep face, wardrobe, and styling anchored to the uploaded image while the body follows the reference path.

• Simple controls: Only image, video, and quality are required—no prompt field to tune for every take.

• Quality tiers: Choose 360p for fast drafts, 540p for balanced review, or 720p when the cut needs a sharper delivery frame.

• Clip-length billing: Charges follow the reference video length (rounded up, minimum 1 second, maximum about 30 seconds).

Parameters#

Pricing#

Billing follows the reference video duration, rounded up to the next whole second, from 1 to 30 seconds.

How to Use#

• Open the image-to-video form on the RunComfy model UI and confirm the motion-mimic template is selected.

• Upload a subject image with a readable silhouette, clear face or head, and enough body visible for the intended motion.

• Add a motion reference video where a person is the primary mover; keep the clip within about 30 seconds and prefer mp4 or mov.

• Choose quality: start at 360p for exploration, then raise to 540p or 720p once the motion match looks right.

• Generate a draft and compare limb timing, stance, and framing against the reference performance.

• If motion drifts, swap a cleaner reference (single performer, stable camera) or tighten the subject crop on the still.

• When the take is solid, lock the same inputs and call the HTTP API for repeatable production runs.

• Prefer short, single-routine references when you need crisp mimicry rather than crowded multi-person scenes.

Prompt & Reference Tips#

• Lead with a clear subject: Frontal or three-quarter stills of a person or animal transfer more cleanly than cluttered group shots.

• Keep one motion hero: References work best when one dancer or actor owns the frame for the full take.

• Match framing loosely: Similar body scale between still and reference reduces awkward stretch or crop artifacts.

• Stabilize the camera: Locked or gently moving references usually mimic more reliably than chaotic handheld cuts.

• Trim dead air: Cut long idle intros so billed seconds stay on the useful motion beat.

• Respect media limits: Stay near or under ~30s, ~1920 max resolution, and ~100MB for the reference file when possible.

• Iterate quality last: Prove the motion match at 360p, then spend the higher tier once the choreography reads correctly.

How PixVerse Motion Control Mimic compares to other models#

• Motion source: Based on publicly available information, this path copies performance from a reference video onto a still, while many image-to-video tools rely mainly on text prompts for motion.

• Input simplicity: Two media inputs plus a quality enum keep the loop short for dance and pose experiments.

• Use-case focus: Strong fit for remakes, character animation tests, and social clips where identity stays on the image and timing comes from video.

• Resolution set: Three practical tiers (360p–720p) cover draft-to-publish needs without extra camera or audio switches.

More Models to Try#

• PixVerse C1 Image To Video for prompt-directed animation from a single still.

• PixVerse V6 Transition for bridging a defined first and last frame.

• PixVerse V6 Extend for continuing an existing clip with new directed motion.

• Other RunComfy image-to-video models when you need text-led camera moves instead of reference mimicry.