May 24, 2026

6 min read

By Team AIGN, step-by-step guide | Updated May 2026

🚀 Fast track: our free generator → enter English Pony prompt → wait 15 seconds → download. No install, no signup.

Quick answer

To undress a photo with AI in 2026, you have two paths: (1) text-to-image . Describe what you want, AI generates a new image, easiest through our free in-browser generator; (2) img2img inpainting . Take an existing photo, mask the clothing area, AI replaces it with body. Path 2 requires local install of ComfyUI or Forge (8+ GB VRAM GPU needed). Path 1 takes 30 seconds. Path 2 takes 30 minutes to set up but works on your own photos.

⚡ The easiest path

Open the generator →

Nothing to install. Nothing to learn. Type, generate, download.

Path 1: Text-to-image (recommended for beginners)

This generates a brand-new image based on your description. You can’t use it on a specific existing photo, but it’s instant and free.

Step 1: Open the generator

Go to aiimagegeneratornsfw.com/#generator. You’ll see a prompt input and a Generate button.

Step 2: Write a Pony-format prompt

Pony XL was trained on English booru-style tags. Basic formula for an undressed character:

score_9, score_8_up, score_7_up, source_anime,
1girl, nude, [appearance],
[scene and pose],
masterpiece, best quality, detailed eyes
Negative: score_4, score_5, source_furry, source_cartoon,
clothing, dress, shirt, watermark, deformed

Real example:

score_9, score_8_up, 1girl, nude, long blonde hair,
blue eyes, slim body, lying on bed, bedroom,
soft window light, detailed eyes, masterpiece
Negative: score_4, score_5, clothing, dress, watermark

Don’t want to write English prompts? Our free prompt generator builds them from a plain description.

Step 3: Hit Generate

15-20 seconds. Right-click on the result → Save Image. Done.

Step 4: Iterate

Each run uses a different random seed. Re-run to get variations. No daily limit.

Path 2: Img2img inpainting on a specific photo

This works on an existing photo: you mask the clothing area, and AI replaces it. Requires local install but gives full control.

What you need

NVIDIA GPU with 8+ GB VRAM (RTX 3060 12GB optimal . Check your card with our VRAM calculator)
30 minutes for first-time install
30 GB free disk space

Installing Stable Diffusion WebUI Forge

Option A: Forge (easier, recommended for first-time)

Download Stable Diffusion WebUI Forge . Windows portable ZIP
Extract to any folder
Download Pony Diffusion XL v6 (~6.5 GB) from Civitai . Needs Civitai email account
Drop the safetensors file into models/Stable-diffusion/
Run webui-user.bat
Open http://127.0.0.1:7860 in browser

Option B: ComfyUI (more powerful, steeper curve) . See our ComfyUI guide.

Compare both: ComfyUI vs Forge vs A1111.

Inpainting workflow in Forge

Open the img2img tab
Click Inpaint sub-tab
Upload your photo
Draw a mask with the black brush over the clothing area you want replaced
Prompt:

   score_9, score_8_up, 1girl, nude body, detailed skin texture

Negative prompt:

   score_4, score_5, clothing, dress, shirt, pants, watermark, deformed, bad anatomy

Sampling steps: 28
Sampler: Euler a
CFG scale: 6-7
Denoising strength: 0.85 (strong override . Necessary)
Inpaint area: Only masked
Mask blur: 8 pixels
Click Generate

5-30 seconds later, you have the result. Lower denoising (0.7) for subtler change, higher (0.9) for stronger replacement.

💡 Reminder

Don’t want to install Stable Diffusion? Use our generator for text-to-image →. For img2img on a specific photo, local install is required.

Path 3: Civitai On-Site Generator (middle ground)

Civitai’s browser generator supports img2img inpainting without local install:

Sign up at civitai.com . Email + 18+ toggle
Verify email
Enable mature content in Account → Browsing settings
Click Create → Image
Select Pony Diffusion XL v6 from the model dropdown
Upload your photo to Image2Image
Draw mask, enter prompt, generate

Free daily Buzz credits cover ~100 generations.

What matters for good results

Source photo quality (Path 2 only)

Higher resolution = better inpainting (1024×1024+)
Good lighting
Clear separation between clothing and skin/background
Minimal pose complexity (standing, sitting straightforward)

Complex poses with layered clothing produce artifacts. Sometimes you need multiple passes with different masks.

Anatomy quality (both paths)

Pony XL base gives 6/10 hand and eye anatomy. Enable ADetailer in Forge:

face_yolov8n.pt . Auto-fix faces
hand_yolov8n.pt . Auto-fix hands

Quality jumps to 9/10. ADetailer comes built-in with Forge.

Negative prompts (both paths)

Critical. Without strong negatives, you’ll often get partial clothing artifacts:

score_4, score_5, source_furry, source_cartoon,
clothing, dress, shirt, pants, bra, watermark,
deformed, bad anatomy, missing limbs, extra fingers

Use our free negative prompt builder . Click toggles for issues, get optimized negative.

Three paths compared

Path	Speed	Difficulty	Specific photo?	Privacy
Our generator	15-20 sec	⭐ (just open)	❌ Text-to-image only	Good
Local install Forge	5-30 sec	⭐⭐⭐ (30 min setup)	✅ Full img2img	Excellent
Civitai	30-60 sec	⭐⭐ (signup needed)	✅ Img2img supported	Medium

Legal and ethical boundaries

You can undress AI:

✅ Fictional characters (text-to-image generates imaginary people)
✅ Your own photo for private use

You cannot undress AI:

❌ Real people without consent . Illegal in US (federal AI Disclosure Act 2025), UK (Online Safety Act), EU (AI Act), and most jurisdictions
❌ Anyone under 18 . Universally illegal, severe criminal penalties
❌ Use for harassment, blackmail, sextortion . Separate criminal charges

Our generator is for fictional characters via text-to-image. Real-person deepfakes violate our terms.

🎯 Ready?

Open the generator →

No install. No signup. 15 seconds to first image.

Free tools to pair with it:

🎲 Prompt generator . Auto-build Pony prompts
🧮 VRAM calculator . Check if your GPU runs local install
❌ Negative prompt builder
🎯 Model recommendation quiz

Frequently asked questions

How do I undress a photo with AI?

Two ways: text-to-image via our free generator (instant, no install), or img2img inpainting via local Forge/ComfyUI (works on specific photos, requires GPU). Detailed steps above.

Can AI undress any photo?

Img2img inpainting works on photos with reasonable resolution, clear lighting, and standard poses. Very small images, heavy shadows, or complex multi-layer clothing produce artifacts.

Do I need a powerful GPU?

For our cloud generator . No, any device with a browser works. For local install . Yes, RTX 3060 12GB minimum.

How do I write a prompt?

Use English booru-style tags with Pony score prefix. Example: score_9, score_8_up, 1girl, nude, blonde hair, bedroom, masterpiece. Or use our free prompt generator to auto-build.

Why are the results sometimes still clothed?

Increase CFG to 7-8, add stronger negatives (clothing, dress, shirt, pants, bra, undergarments), and raise denoising to 0.9 in img2img. For text-to-image, make sure nude is in the positive prompt.

Can I do this on my phone?

Our generator works on mobile browsers (iOS Safari, Android Chrome). Local install requires a desktop with GPU.

Is it free?

Yes. Our generator, local install of Pony XL, and Civitai’s free daily Buzz credits are all free. Only paid Telegram DeepNude bots cost money (and they’re worse).

Is this legal?

Generating fictional characters . Legal in most jurisdictions. Creating deepfakes of real people without consent . Illegal. See “Legal” section above.

Related guides

Verdict

The fastest way to undress a photo with AI in 2026 is our free in-browser generator . Text-to-image, 15 seconds, no install. For img2img on a specific photo, install Forge with Pony XL locally. The middle ground is Civitai’s browser generator with email signup.

Skip the Telegram bots. They’re worse quality, paid, and risk your privacy.

Try it now →

How to Undress a Photo with AI 2026: Step-by-Step Guide