Turn a selfie, portrait, pet or landscape into anime-style art. AnimeGAN, a neural network trained on frames from Japanese animated films, repaints your photo with flat cel colors, clean outlines and glowing skies, right inside your browser. Pick one of three styles, set the effect strength, compare before and after, and download the picture as PNG or JPG.
Last updated:
◂ ▸
PhotoAnime
Working…
Anime style
photofull anime
nonebold
flatsmooth
Use your own ONNX model
Load an AnimeGAN-style model (.onnx) from your computer. It must take an RGB image and return an RGB image of the same size.
Advertisement
Advertisement
How to use the Photo to Anime
Add a photo by dropping it on the box, choosing it, or pasting it with Ctrl+V. A sample countryside photo is shown first.
Pick a style: Hayao for soft painted scenes, Shinkai for bright skies and glowing light, Paprika for punchy, saturated colors.
Click Convert to anime. The first time a style is used its AI model (about 9 MB) downloads and is cached; later conversions start immediately.
Fine-tune with Effect strength and drag across the picture to compare it with the original.
Download the anime picture as PNG or JPG.
How AnimeGAN turns a photo into anime
AnimeGAN (Chen, Liu and Chen, 2019; AnimeGANv2 in 2020) is a lightweight generative adversarial network. Its generator is a small convolutional network that takes a photo and outputs a picture of the same size. During training a second network, the discriminator, tried to tell those outputs apart from real frames of animated films, while extra loss terms kept the content of the photo, the colors and the texture of the target style. Three ideas make the results look hand-drawn:
Grayscale style loss: textures and strokes are learned from gray versions of the film frames, so the photo keeps its own colors instead of copying the film palette.
Edge-smoothed training images: the discriminator also sees film frames with blurred edges labelled as fake, which teaches the generator to draw crisp, clean outlines.
Color reconstruction loss in YUV space keeps skin, sky and foliage close to the original hues.
In this tool the generator runs at the AI working size you choose (rounded to a multiple of 32 px), with pixel values scaled from 0-255 to −1…1. The output is scaled back to your picture's size, and Effect strength blends it with the original photo.
High contrast, vivid reds and oranges, punchy shading
Food, festivals, colorful street scenes, bold portraits
Tips for better anime pictures
Start with a sharp, well-lit photo. Noise and heavy JPEG artifacts turn into blotches; clean them first with the Image Denoiser.
Crop to the subject with the Image Cropper. A face that fills a large part of the frame keeps more detail at the same working size.
Use Balanced (512 px) to preview and switch to Detailed (768 px) for the final picture. Larger sizes are slower and need more memory, especially on phones.
Set Effect strength to 70-85% to keep a little photographic texture, which often flatters faces.
It uses AnimeGANv2, a generative adversarial network (GAN). A generator network learned to repaint real photos so that a second network could not tell them apart from frames of animated films. The result keeps the layout of your photo but replaces textures with flat cel-shaded colors, simplified details and clean edges. The model runs in your browser with ONNX Runtime Web.
What is the difference between the three styles?
Each style is a separate model trained on a different set of animated films. Hayao gives soft, slightly muted painterly colors; Shinkai gives saturated blue skies, strong light and glowing highlights; Paprika gives high contrast and very saturated, punchy colors. Try all three: the best one depends on the photo.
Does it work on faces and selfies?
Yes, portraits and selfies work, and the AI smooths skin and simplifies hair into anime-like shapes. AnimeGAN is strongest on scenery, so faces keep their real proportions rather than getting big anime eyes. Use a well-lit, sharp photo and a Detailed working size for the best face results.
Is my photo uploaded anywhere?
No. The model file is downloaded to your browser and the photo is processed on your own device. Nothing is sent to a server, and the tool works offline in Quick anime filter mode.
Why is the first conversion slower?
The first time you use a style, your browser downloads its model (about 9 MB) and the ONNX runtime (a few MB). Both are cached, so later conversions start right away. The AI itself usually takes 1-10 seconds depending on the working size and your device.
What is the Quick anime filter?
It is a classic image filter, not AI: a Kuwahara filter flattens textures into painted patches while keeping edges, brightness is shaded in a few cel tones, Sobel edge detection adds line art, and a color grade plus bloom imitates the chosen style. It is instant and works offline, but it cannot redraw shapes the way the neural network does.
Can I use the anime pictures commercially?
The pictures you create are made from your own photos on your device. The AnimeGANv2 models are released by their authors for non-commercial use, so check the model license before using AI results commercially; Quick anime filter results have no such restriction.
Are all the tools on FreeToolCabin tested for quality and accuracy?
Yes! Every tool on our site undergoes automated Q/A testing in real browser environments to ensure high precision, correct output, and zero errors. If you happen to spot a bug or have a suggestion, please contact us at [email protected] and our team will review it right away.