This pet Halloween costume AI page turns one ordinary photo of your dog or cat into a costume portrait, without wrestling anyone into a real outfit. Pick one of six looks taken from this year's pet costume trends, load it into Hy Image 3.5 Edit, upload your photo and check the result before you buy, sew or post anything.
Three looks were tested on a dog and three on a cat, but every prompt works for either. Choose "Use this look" to load the prompt into Hy Image 3.5 Edit below, then upload your own pet photo. Each prompt says what must stay the same before it says what to add.
Foil-Wrapped Baked Potato
A plush russet potato in crinkled silver foil, split open on top with a dollop of sour cream and chives. Soft, round and very silly on a small dog.
Pet photo editing prompt
Edit the uploaded pet photo. Keep the exact same animal: same species and breed, same coat length and colour, every marking and patch in the same place, same eye colour, ear shape, muzzle and body size, and the same pose and camera angle. Dress the pet in a soft plush baked-potato costume that wraps the body like a loose sleeping bag: a brown russet potato skin made of fleece, half covered in loosely crinkled silver foil, split open along the back with a fluffy white sour-cream cushion, a pat of yellow butter and a few green chive strips on top. The pet's head, face, eyes, ears and front paws stay fully out of the costume and uncovered; nothing touches the eyes or covers the nose. Place the pet on the same floor with a few real potatoes and a small foil tray nearby, warm kitchen daylight, shallow depth of field, realistic pet photography with crisp fur detail. Avoid: changing the breed or markings, human hands, extra legs or tails, a second animal, costume over the eyes, text, logos or watermarks.
Glow-in-the-Dark Dino Coat
A hoodless green dinosaur coat with soft spikes that glow lime in a dim room. The trend list's glowing dinosaur, built so the face stays clear.
Pet photo editing prompt
Edit the uploaded pet photo. Preserve the animal exactly: breed, fur colour and texture, all markings, eye colour, ear shape and position, face and body proportions, and the original pose. Put the pet in a lightweight hoodless dinosaur coat in deep forest green that starts at the shoulders: a row of soft rounded felt spikes along the spine from the shoulders to a short stuffed tail, small dark scale patches on the sides, and the spikes and scale edges glowing a gentle phosphorescent lime green. There is no hood and nothing on the head: the head, eyes, ears, nose and whiskers stay completely bare and visible. Turn the room into an evening scene: lights low, a deep blue window behind, the costume glow casting a faint green rim light on the fur, a few glow stars on the wall. Realistic low-light pet portrait, sharp eyes, natural fur detail, slight film grain. Avoid: a different animal, glowing eyes, fangs, a real dinosaur head, a hood or hat, extra limbs, text, logos or watermarks.
Hundred-Leg Centipede Suit
A segmented rust-and-black body suit with a fringe of little stuffed legs down both sides. Funny on long dogs, surprisingly cute on short ones.
Pet photo editing prompt
Edit the uploaded pet photo and keep the same pet with no changes to its identity: same species and breed, same fur colour and markings, same eye colour, same head shape and ears, same size and pose. Dress it in a playful centipede costume: a soft padded back cover made of glossy rust-orange and black segments running from the shoulders to the tail, with two rows of many small stuffed felt legs in tan hanging down both sides, plus a pair of bendy antennae on a loose headband placed behind the ears. The costume is clearly fabric and plush, not a real insect, and the pet's face, eyes, ears, real legs and paws stay uncovered and natural. Background: an autumn porch with dry leaves and a carved pumpkin, golden late-afternoon light, eye-level pet photography with crisp fur and fabric texture. Avoid: real bugs, creepy or scary insect detail, changing the animal's legs or adding real extra limbs, covering the eyes, text, logos or watermarks.
Salted Butter Stick
A pale yellow butter-block tunic with a half-peeled paper wrapper. Simple, very photogenic on an orange or cream cat.
Pet photo editing prompt
Edit the uploaded pet photo. Keep the same cat or dog exactly as it is: same species and breed, identical fur colour, stripes, patches and white areas, same eye colour, ear shape, face and body proportions, and the same seated pose and framing. Add a soft costume shaped like a stick of salted butter: a loose rectangular tunic in pale buttery yellow with slightly rounded corners, a white and pale blue waxed-paper wrapper peeled halfway back like a cape, and a smooth glossy surface with a single shallow knife curl on top. The costume sits on the body only; the head, eyes, ears, whiskers and paws stay completely uncovered. Setting: a clean kitchen counter style backdrop with a wooden butter board and a small dish, soft morning light, minimal and bright, realistic pet photography with sharp fur detail. Avoid: printed brand names or any lettering on the wrapper, melting or greasy textures on the fur, changing the animal, a costume on the head, extra paws, logos or watermarks.
Midnight Rider Cape
A nod to the Headless Horseman trend that keeps the head very much attached: a tall-collared black cape, a tiny felt lantern and a misty lane.
Pet photo editing prompt
Edit the uploaded pet photo. The pet keeps its full head, clearly visible and unchanged: same species and breed, same fur colour and markings, same eye colour, ears, whiskers and face, same pose and camera angle. Dress the pet as a storybook midnight rider: a short black velvet cape with a tall stiff collar that stands behind the neck (not over the face), a dark burgundy lining, a small brass clasp at the chest, and a tiny felt jack-o'-lantern bag hanging at its side with a soft warm glow inside. Do not hide, remove or cover the head in any way; the costume is only a cape. Background: a misty country lane at dusk with bare trees, a wooden fence and a few fallen leaves, cool blue fog with the warm lantern glow on the fur. Cinematic but friendly pet portrait, sharp eyes, realistic fur. Avoid: a missing or hidden head, a horse, weapons, blood, scary faces, a second animal, text, logos or watermarks.
Pumpkin Patch Pet
A round orange pumpkin body with a curly green stem and leaf collar. The classic, easy to read at thumbnail size.
Pet photo editing prompt
Edit the uploaded pet photo. Keep the exact same animal and do not change it: species and breed, coat colour and pattern, white markings, eye colour, ears, face shape, size and pose all stay as in the photo. Put the pet in a soft round pumpkin costume: a plump orange ribbed body piece around the torso, a short curly green stem and two felt leaves forming a small collar at the neck, and a few curling green vine tendrils at the sides. The head, eyes, ears, whiskers and paws stay uncovered and free. Place the pet in a pumpkin patch at golden hour with real pumpkins of different sizes, straw and a hay bale behind, warm low sun and soft backlight on the fur. Bright, cheerful, realistic pet photography with sharp focus on the eyes. Avoid: a carved face on the costume over the pet, changing the animal's colours, extra tails or legs, a person, text, logos or watermarks.
The two pets in these examples do not exist: a fictional terrier mix and a fictional ginger tabby were generated for this page, then dressed with these prompts during testing, using an open image model (FLUX.2 klein) for most looks and ChatGPT image generation for the dino coat. Your result comes from Hy Image 3.5 Edit or whichever model you select, and will vary with your photo.
Put the costume on your own pet
The look you picked is loaded into Hy Image 3.5 Preview · Edit. Upload one photo of your dog or cat, keep the 3:4 ratio for a phone-friendly portrait, and check the credit quote before you generate. New accounts get one free Hy Image 3.5 image.
Add inputsAttach images, videos, or audio that are part of what you want to create.
0/20
JPG, PNG, WebP · 20 MB max
Just add the material you need. The generator will use it automatically.
Real Hy Image 3.5 Preview is selected by default, with 2K or 4K output and a separate reference-editing mode. Your first image is free; after that each model shows its own credit cost.
Six original pet looks: foil baked potato, glow-in-the-dark dinosaur, stick of butter, centipede, a caped midnight rider and a pumpkin.
Every prompt locks the breed, fur colour, markings and eye colour, and keeps the eyes, ears and face uncovered.
The best input is a sharp, eye-level photo with one animal, visible paws and even daylight.
Hy Image 3.5 Edit runs at 2K or 4K, so a result can go to a phone screen or a printed card.
Take a pet photo the AI can read
A costume edit can only keep what it can see. If the photo shows a blurred tail and half an ear, the model has to guess the rest, and guesses are where a spaniel turns into a different spaniel. Spend two minutes on the input and the costume will sit on your pet rather than on a lookalike.
Get down to eye level. Shoot from the height of your pet's eyes, not from standing height. Top-down photos shrink the body and make the costume look like a blanket thrown over a head.
Show fur detail. Use daylight from a window or a porch, not a flash. Even light keeps the texture of the coat, which helps the model keep markings in place.
One animal, whole body. Seated or standing with the front paws in view gives the costume something to wrap around. Crop out other pets and people.
Mind dark coats. Black dogs and cats lose shape against dark sofas. Put them on a lighter floor or rug so the outline stays clear.
Use a calm moment. Eyes open, mouth closed or relaxed, ears in their usual position. A mid-yawn photo will be dressed mid-yawn.
Why every look keeps the eyes uncovered
There are two reasons, one practical and one about the picture. Real pet costumes that cover the eyes or ears stress animals: they cannot see where they are going, they shake their heads, they paw at the fabric. If you plan to copy a look in real life, a version that only covers the back and chest is the one your dog or cat will tolerate.
The second reason is identity. Eyes, ears and the shape of the muzzle are what make your pet recognisable in a photo. When a hood, mask or hat covers them, the model fills the gap with a generic face, and the result stops looking like your animal. That is why every prompt on this page says where the costume ends, and why the midnight rider look turns the headless rider idea into a cape with the head left firmly in view.
How the looks map to this year's pet trends
Google's 2026 Halloween trend list for pets is full of food and objects rather than monsters: baked potatoes, sticks of butter, a centipede, a glow-in-the-dark dinosaur, the Headless Horseman. The looks above are original takes on those ideas, made with plain shapes and materials so they read at a glance and do not borrow anyone's character design.
If you want a different idea from the list, keep the same structure: one sentence that locks the animal, one paragraph describing the costume as fabric with its colours and shape, a line saying the head and paws stay free, the setting and light, then what to avoid.
Template for your own pet costume idea
Edit the uploaded pet photo. Keep the same animal: species and breed, fur colour and markings, eye colour, ears, face and pose. Dress it in a soft fabric costume of [your idea], described by colour, shape and material. The head, eyes, ears and paws stay uncovered. Setting: [place], [light]. Realistic pet photography with sharp fur detail. Avoid: changing the animal, extra limbs, costume over the eyes, text, logos or watermarks.
Sizes for sharing and printing
The looks are set to 3:4, which fills a phone screen in a feed and crops cleanly to a 4:5 Instagram post (1080 × 1350) without cutting off ears. For Stories, Reels covers or a lock screen, switch the ratio to 9:16 before you generate rather than cropping later; a costume that was composed for portrait loses its tail when it is squeezed into a tall frame afterwards. Use 1:1 for profile pictures and group chats.
Hy Image 3.5 Edit can output 2K or 4K. 2K is plenty for social media. Choose 4K if you want to print a 5 × 7 card or a small poster for the door, since it leaves room to crop. Download the file from the workspace or your account history; the site does not post anything for you.
Frequently asked questions
Does the pet Halloween costume AI work for cats as well as dogs?
Yes. All six prompts are written for "the pet" and lock the species, so the same look works on a cat, a dog or a rabbit. Cats with very fluffy coats sometimes get a slightly thinner costume, so ask for "loose" or "padded" if you want more volume.
Will my pet still look like my pet?
Usually, if the photo is clear. Each prompt asks to keep the breed, coat colour, markings, eye colour and pose, and Hy Image 3.5 Edit follows those instructions well on sharp, eye-level photos. Check the markings around the face in the result; if one moved, generate again or add a line describing it, for example "white blaze down the nose".
Can I put two pets in one costume photo?
You can upload a photo with two animals and say "keep both pets, dress both in the same costume", and Hy Image 3.5 Edit accepts several reference images. Results are less predictable than with one pet, so start with single portraits and try the pair once you know which look you like.
How much does a costume photo cost?
New accounts receive one free Hy Image 3.5 Preview image. After that, each image uses credits, and the workspace shows the exact quote for the model and resolution you select before you generate.
Is a costume like this safe for my real pet?
The AI version is risk-free. If you make one for real, keep it loose, light and off the head, eyes and ears, and let your pet wear it only for a few minutes of photos while you watch. Skip costumes for animals who freeze, hide or try to wriggle out.
Updated 2026-10-20. Independent Hy Image 3.5 workspace, not affiliated with Tencent. Model facts are Tencent's published figures. Every generation here uses the model you select in the workspace.
