AI Jellycat Style Video Generator: Turn a Photo, Pet or Object into a Plush Toy
The VdoBloom AI Jellycat Style Video Generator rebuilds whatever is in your photo as a soft toy and then animates it — stitched fabric texture, round button eyes, rounded plush proportions and a pastel palette, moving with the slightly weighted wobble that real stuffed animals have. Upload a person, a pet or an object, pick an action, and the AI hands back a short cuddly clip a few minutes later.
The conversion works on outline rather than identity. Nothing in this pipeline looks for a face: the model reads the subject's silhouette, inflates it toward plush proportions, rounds off every hard edge and then re-textures the surface as fabric. That is why a dog, a teapot and a scuffed old car all convert convincingly, and it is also why framing matters more than facial detail. Give it a full silhouette and it has something to round; give it a cropped face and it has an outline that was never meant to become a toy shape.
Three templates set what the finished plush does. Classic Jellycat performs the transformation itself and settles into gentle, cosy motion. Jellycat Dance gives the plush a bouncy little jig, which stresses the fabric physics hardest and shows most clearly whether the conversion held. Jellycat Peek-a-boo animates a hide-and-reveal, so part of the plush leaves frame and comes back — the one that rewards a clean, uncluttered background. All three load their text into an editable prompt box, so you can specify fabric colour, plush size, or where the toy is sitting.
On credits: this tab has no model of its own, so it opens on Runway Gen-3 at 10 credits for a 720p five-second render, and Seedance 1.5 Pro does the same for 8. New accounts get 10 free credits, which covers exactly one Runway Gen-3 clip or one Seedance clip with two to spare. That first render is real and full length, but free-tier playback carries a VdoBloom watermark and the download is the watermarked copy — the clean file comes with a paid plan. Upload your own pets and objects freely; photos of people should be of adults who have agreed to it.
How it works
- 1
Upload any subject
A person, a pet, or an object — all three are supported, because the transformation works from the outline rather than from a detected face. Photos where the subject is fully in frame with a little space around it convert best, since the plush version needs room for its rounder proportions.
- 2
Choose the plush action
Classic Jellycat for the transformation and a cosy settle, Jellycat Dance for a bouncy jig, or Jellycat Peek-a-boo for a playful hide-and-reveal. Each loads a complete prompt you can read before you spend anything.
- 3
Name the fabric details
The prompt box is editable, so you can specify the fabric colour, ask for a particular pastel palette, add a soft blanket or a windowsill behind the plush, or request a smaller and rounder body. Wording like 'matte fabric, visible stitched seams' pushes the surface further from realistic fur.
- 4
Pick a model
Runway Gen-3 is the tab default at 10 credits for 720p and five seconds. Seedance 1.5 Pro is the value pick at 8, ByteDance V1 Pro Fast I2V is the lowest-priced option in the picker at 7, and PixVerse leans more animated, which suits the toy look. Forty-four image-to-video variants are available in total.
- 5
Set duration and shape
Five seconds carries the transformation comfortably; Runway Gen-3's other option is 10 seconds at 20 credits, which suits the Dance template more than the others. Choose 9:16 for Reels and TikTok, 1:1 for a feed square, or 16:9 if the clip is going into a longer edit or a slideshow.
- 6
Generate and check the silhouette
The render runs server-side in a few minutes. When it lands, judge it on the outline first — if the plush shape reads as a toy from across the room, the fabric detail will hold up close as well.
Why use this tool
Silhouette-driven, so any subject works
The transformation reads the subject's outline instead of hunting for a person, which is why pets, plants, cars and household objects convert as convincingly as portraits do. No face is required anywhere in the process.
Three behaviours that test the plush differently
A cosy settle keeps the toy still and shows texture. A bouncy dance stresses the fabric physics and exposes any conversion that did not fully take. A peek-a-boo sends part of the plush out of frame and brings it back, which needs a clean background to work.
Fabric cues, not generic cartoon-ification
The prompts ask specifically for stitched seams, button eyes and a soft matte nap on the surface. That is the difference between something that reads as a stuffed toy and something that just reads as a smoothed-over 3D render.
Colour, size and setting respond to plain wording
Ask for a specific pastel, a rounder body, a smaller toy, a blanket underneath or a windowsill behind, and the model takes it. Nothing you type is quietly discarded in favour of the preset text.
Two engines inside the free allowance
Ten signup credits stretch to one Runway Gen-3 render at 10 credits, or one Seedance 1.5 Pro render at 8 with two left. Both produce a full-length clip; both play and download with a VdoBloom watermark until you move to a paid plan.
Errored jobs cost nothing
Failed generations are refunded automatically, so trying the same photo across templates is limited by patience rather than by credits when the provider misbehaves.
Frequently asked questions
Can I turn my pet into a plush toy video?
Yes, and pets are the subject this effect handles most comfortably. Upload a clear photo of your cat, dog, rabbit or bird and the AI rebuilds it as a soft toy version with fabric texture and button eyes, keeping the markings and proportions recognisable. A side-on or three-quarter shot in decent light works better than a close-up of the face alone, because the plush shape is built from the whole body outline — a head-only crop gives the model nothing to round.
Does it work on objects, or only living things?
Objects work, and often better than people. A teapot, a pair of trainers, a bicycle, a first car — anything with a distinct outline converts, because the pipeline is reshaping a silhouette rather than identifying a creature. Objects with strong, simple profiles do best. Very thin structures such as railings, bike spokes or leafless branches are the weak case, since there is not enough volume there to inflate into something plush.
Why does full-body framing beat a close-up here?
Because the conversion is a rounding pass over the outline. It takes the silhouette, exaggerates the volume, softens every corner and re-skins the surface as fabric. When the whole subject is in frame with a little space around it, there is a complete shape to work with and the result reads as a toy you could pick up. When you upload a tight face crop, the model has only a partial outline, so it tends to produce a smoothed head rather than a stuffed animal, and the fabric texture ends up floating on features instead of wrapping a form.
What do the three templates actually change?
Classic Jellycat runs the transformation and then lets the finished plush settle with gentle movement — the calmest option and the one that shows the fabric texture most clearly. Jellycat Dance gives the toy a bouncy jig, which is the best stress test: if the plush wobbles with weight and the seams stay put, the conversion took properly. Jellycat Peek-a-boo animates hiding and revealing, so part of the toy leaves the frame and returns, which needs a simple background or the model has to reconstruct clutter on the way back in.
The plush came out too realistic — how do I push it further?
Add explicit material wording to the prompt: 'exaggerated round plush proportions, visible stitched seams, matte fabric surface, button eyes, no realistic fur, no photorealistic skin'. Naming what you do not want matters as much as naming what you do, because a realistic-leaning engine will otherwise keep the original texture and only soften the shape. If that does not do it, switch models — PixVerse pushes into stylised territory more readily than Runway does, which suits a toy.
What does the free allowance cover on this tab?
One clip. There is no tab-specific default model here, so it opens on Runway Gen-3 at 10 credits for 720p and five seconds, which is the whole 10-credit signup allowance. Seedance 1.5 Pro at 8 credits leaves you two, not enough for a second run. The free render is complete and full quality, but it plays with a VdoBloom watermark and downloads as the watermarked copy. The watermark-free file is a paid-plan feature.
Which model should I use, and how long does a render take?
Runway Gen-3, the default, gives the smoothest camera movement, which suits the calmer Classic template. Seedance 1.5 Pro at 8 credits holds soft texture well and is the sensible choice if you want to try more than one template. PixVerse pushes furthest into animated motion, which flatters the Dance template. Renders normally complete within a few minutes and run in the background, so you can queue one and carry on elsewhere in VdoBloom.
Can I have two subjects — a person and their pet — in one clip?
Yes, if they are already together in the uploaded photo. The pass applies to everything in the frame, so a person holding their dog becomes two plush figures in the same scene, and the scale relationship between them is usually preserved. What it cannot do is merge two separate photos into one video; this tab works from a single image, so shoot or crop them together first.
Is this tool affiliated with the Jellycat brand?
No. VdoBloom is an independent platform with no affiliation, endorsement or licence from Jellycat. The tool creates video in a soft plush-toy aesthetic inspired by that general style, and it is intended for fun, personal and fan-style content. Bear that in mind if you are considering any commercial use.
Are my uploads private, and what am I allowed to upload?
Your own pets and objects are the easiest case — upload freely. Photos of people must be of adults, either yourself or someone who has agreed to appear. Uploads are used only to run the generation you asked for and stay tied to your account; VdoBloom does not publish them or your finished videos. Individual videos can be deleted from your creations library and the whole account from settings. Every prompt is moderated before it runs and may be rejected.