Almost every bad result from a Cyberpunk Edgerunners AI filter comes from one of three things: the photo you uploaded, a style or format that does not suit that photo, or the normal variation of an AI model between renders. On Cyber Anime Filter the fix is therefore always one of three moves: use a different photo, pick a different style or format, or run the set again. There is nothing to type and no prompt to edit.
This guide goes through the seven problems people run into, says which move fixes each one, and is plain about what cannot be fixed.
Key takeaways
- If the same fault appears in all 4 images, the cause is almost always the photo. Change the photo before you spend another credit.
- If one image of the 4 has the fault, it is variation. Run the set again.
- If the fault belongs to the outfit or the scene, such as a metal hand or a wall of signs, pick a style that does not have that element.
- A set of 4 images costs 1 credit and every rerun costs 1 more. Renders that fail or are blocked return their credit automatically. Renders that finished are not refundable.
What a set is and what a rerun costs
One render on Cyber Anime Filter is a set of 4 images of the same character, drawn with GPT Image, an image model made by OpenAI: a chest-up portrait, an action shot, a wide street shot and a key-art poster.
The four are drawn separately from the same photo, the same outfit and the same scene. A fault that shows up in all four came from something they share, which is your photo or the style. A fault in one came from that one draw.
A set costs 1 credit whichever of the 8 styles and 3 formats you choose, and running it again costs 1 more. If a render fails or is blocked, the credit goes back to your account on its own. If it finished and you do not like it, the credit is spent. Finished sets stay downloadable for 7 days.
Quick reference: seven problems and the fix for each
| Problem | Most likely cause | Fix |
|---|---|---|
| It does not look like you | Photo hides hair or face shape, or is filtered | Different photo |
| Semi-realistic, not cel-shaded | Variation, or a heavily processed photo | Run again, then a plainer photo |
| Extra fingers, wrong metal hand | Variation on a hard detail | Run again, or a style without a mechanical hand |
| Stray text or a fake logo | Variation, more likely in sign-heavy scenes | Run again, or a style with a plainer scene |
| Eyes hidden | Sunglasses or hair over the eyes in the photo | Different photo |
| Wrong person, blended faces | More than one person in the photo | Solo photo |
| Mask on the face, wrong hair colour | Variation, or coloured light in the photo | Run again, or a photo in neutral light |
Problem 1: it does not look like you
This is the complaint that matters most, and it helps to know why it happens. Anime simplifies a face to a few lines. Likeness rests on a short list: your hair, your face shape, your eye colour, your eyebrows, your skin tone, your expression, and anything you wear on your face such as glasses or a beard.
If the photo does not show those clearly, the drawing cannot keep them. The weakest results come from tight face crops that cut off the hair, selfies with a heavy beauty filter, photos taken from far away, and hats or hoods over the hairline.
The fix: use a different photo. Pick one that is sharp, front-facing and evenly lit, with your hair and shoulders in frame. Running the same weak photo again rarely helps, because the missing information is still missing.
If your photo is already good, judge likeness on the portrait, where your face is largest. The street image shows you small and at full length, so it will never resemble you as closely. That is the shot, not a fault. A style change can also help: Neon Street Kid is the most forgiving of the eight, because its outfit is simple and the face fills more of the frame.
Problem 2: it looks semi-realistic instead of cel-shaded
The look you are paying for is a frame of hand-drawn animation: black ink outlines, flat colour, one hard-edged shadow per area, shadows in purple or teal. Every render is told to draw that, and told what it must not be: a painting, a 3D render or a photo with a filter. Most renders follow it. Occasionally one image comes out softer than the rest, with blended skin or a painted background.
The fix: run the set again. When one image of four drifts, it is variation, and a new draw usually lands back on the flat look.
If two sets in a row come out soft across all 4 images, look at the photo. A portrait that has been through strong smoothing, a studio retouch or an artificial background blur gives the model a painted-looking starting point. Swap it for a plain, unedited photo.
One more check before you call it a fault. The four Season 2 look styles are meant to be softer than the Classic neon ones. They follow Cyberpunk: Edgerunners 2, which premieres on Netflix on October 20, 2026 (official site) and which Netflix's Tudum describes as having a reimagined animation style. If you want the hardest ink, pick a Classic neon style: Sandevistan Rush, Moonlit Netrunner, Chrome Solo or Neon Street Kid.
Problem 3: extra fingers or a metal hand drawn wrong
Hands are one of the details AI image models still get wrong, and a mechanical hand is harder than a human one. Every render asks for five fingers per hand. That is followed most of the time, and not always. You have three fixes, in order of cost.
Use the images where the hands are fine. A hand error is usually in one image of the four. The portrait is chest up, so the hands are often out of frame, and the street image shows them tiny. If the action image has a six-fingered fist and the other three are clean, you may already have what you came for.
Run the set again. A hand error is variation. It does not come from your photo.
Pick a different style. Some styles put a hand at the centre of the picture:
| Style | Hands and arms |
|---|---|
| Chrome Solo | Both forearms replaced by chrome cyberarms |
| Burned-Out Legend | One red mechanical hand |
| Chrome Gang | One red mechanical hand, large in the foreground of the action image |
| Nomad Netrunner | Fingerless gloves, one hand raised in the action image |
| Neon Street Kid, Moonlit Netrunner, Sandevistan Rush, Camcorder REC | No mechanical hands |
Problem 4: stray text, letters or a fake logo
Cyber Anime Filter asks the model for no words, letters, numbers, logos, signatures, watermark or border anywhere in the image. Signs and graffiti are meant to be abstract shapes, and the site adds no watermark of its own. AI models sometimes write anyway: a few nonsense characters on a sign, a squiggle like a signature, a made-up brand mark on a jacket.
The fix: run the set again. Stray marks are variation.
If it keeps happening, the scene may be giving the model too many surfaces to write on. Moonlit Netrunner, Nomad Netrunner and Burned-Out Legend have the plainest scenes, with few signs or walls. Camcorder REC draws viewfinder brackets and a recording dot on purpose, and the model sometimes adds stray characters beside them.
If a mark sits at the very edge of an otherwise good image, cropping it off in your phone's photo app is quicker than a rerun.
Problem 5: something covers the eyes
The anime look depends on the eyes. They are the largest feature of the drawn face and carry most of the expression. If the eyes in your photo cannot be seen, the model has to invent them, and invented eyes are where likeness is lost.
Sunglasses hide the eyes completely. The drawing either keeps the dark lenses or guesses at what is behind them. Ordinary glasses are fine: they are kept, and in Neon Street Kid they pick up a neon reflection.
Hair across the eyes is kept as it is, so the eyes stay hidden in the drawing too.
The fix: use a different photo, with both eyes visible and looking at the camera. No rerun and no style change gives the model eyes it never saw.
Problem 6: a group photo or two people in frame
Every image is drawn with one person in it. If you upload a photo of two or more people, the model has to decide whose face to draw. It can pick the wrong person, or borrow hair from one and a jaw from another.
The fix: use a solo photo. If your best photo is a group shot, crop it to yourself before you upload, as long as your face stays large and sharp.
If your friends want to join, each person makes a set from their own photo. Pick the same style and format, and the images sit together as a crew when you post them. Upload a photo of someone else only with their permission, and everyone in a photo must be 18 or older.
Problem 7: a mask on the face or the wrong hair colour
These look unrelated, but both are the drawing overriding something from your photo.
The mask
Chrome Gang includes a tactical half mask. It is drawn pulled down around the neck on purpose, in all 4 images, because a mask over the mouth hides half of what makes the drawing look like you. The model follows this in most renders. Now and then it pulls the mask up.
The fix: run the set again, or pick another Season 2 look style if you would prefer no mask at all. If the mask was in your photo to begin with, the fix is a different photo. The model cannot draw a mouth and jaw it was not shown.
Hair and skin colour
Hair colour is kept from the photo, including dyed hair, which can become the accent colour of the image. When it comes out wrong, the usual cause is the light in the photo. A selfie under a red LED strip, club lighting or a strong sunset gives the model no way to tell brown hair in pink light from pink hair. Warm indoor bulbs can shift skin tone in the same way, most visibly in a daylight style such as Camcorder REC.
The fix: use a photo taken in neutral light. Daylight near a window or outdoor shade works well.
How to choose a photo that avoids most of this
Five of the seven problems start with the photo. This checklist takes a minute and saves credits:
- One person. Just you, with nobody behind you.
- Facing the camera. Both eyes visible and looking at the lens.
- Hair and shoulders in frame. No tight face crops, hats or hoods.
- Sharp and evenly lit. Daylight or neutral indoor light, with no coloured LEDs.
- Unfiltered. No beauty filter, no face-reshaping, no artificial blur.
- Nothing over the face. No sunglasses, masks or hands.
- The right file. JPG, PNG or WebP, up to 10 MB.
There is more on photo choice in our guide to making the Cyberpunk Edgerunners AI filter.
When the picture feels cramped, change the format
Some results are not wrong, just squeezed. Format is chosen before the render, and each style has a shape that suits it. You can compare all eight on the styles page.
- Vertical 9:16 gives height to the after-image trail in Sandevistan Rush and the tilted close-up in Chrome Gang.
- Square 1:1 suits avatars, especially Moonlit Netrunner.
- Wide 16:9 gives room to the ground punch in Chrome Solo, the empty street in Burned-Out Legend and the tape-footage feel of Camcorder REC.
What cannot be fixed, stated plainly
- Likeness is not guaranteed. The drawing keeps your hair, face shape, eye colour and expression as closely as the model manages, and it varies between renders.
- Stray marks happen. AI sometimes adds a letter, a squiggle or a sixth finger. A rerun usually clears it and costs 1 credit.
- Rendered sets are not refunded. Credits come back automatically only when a render fails or is blocked.
Knowing these before you start is the cheapest fix of all. Credits are sold in one-off packs of $9.99 for 10, $24.99 for 30 and $69.99 for 100, with no free tier and no subscription, and they never expire, so an unused credit is still there the next time you want to try a style. See the pricing page for the details, or upload a photo to start.
Cyber Anime Filter is an independent fan-made tool and is not affiliated with, endorsed by or sponsored by CD PROJEKT RED, Studio TRIGGER or Netflix.
Frequently asked questions
Do I lose a credit if my Edgerunners filter result comes out bad?
On Cyber Anime Filter a credit is returned automatically only when a render fails or is blocked. A set that finished drawing has used its 1 credit, even if you dislike it, and running it again costs 1 more. If a finished render clearly went wrong, you can email support and the team will look at it. Unused packs can be refunded within 7 days.
What kind of photo should I upload for an Edgerunners anime filter?
Cyber Anime Filter works best with one sharp, front-facing photo of a single person in even, neutral light. Keep your hair and shoulders in frame, look at the camera, and leave off sunglasses, masks, hats and beauty filters. The file can be JPG, PNG or WebP up to 10 MB. Your clothes do not matter, because the style replaces the outfit.
Can I put me and my friends in one Edgerunners AI picture?
No, Cyber Anime Filter draws one person per image, so a group photo makes the model guess whose face to use. Each friend should make a set from a solo photo, with the same style and format so the images match when posted together. Upload a photo of someone else only with their permission, and everyone must be 18 or older.
How many times should I rerun a set before I change the photo?
Once is enough on Cyber Anime Filter. If a fault shows up in one image of 4, a rerun usually clears it, because it came from normal variation between renders. If the same fault returns in the second set, or appears in all 4 images, the cause is the photo or the style, and another rerun with nothing changed spends 1 credit on the same result.
Why is one image in my set worse than the other three?
Each of the 4 images in a Cyber Anime Filter set is drawn separately, so one can go wrong while the others come out clean. The action image is the hardest shot, with foreshortened limbs and hands close to the camera, which makes it the likeliest place for a finger or metal hand error. Use the three good images, or run the set again for 1 credit.
Sources
- Cyberpunk: Edgerunners 2, official site (CD PROJEKT RED)
- Netflix Tudum: Cyberpunk: Edgerunners 2 release date and news