Cyberpunk Avatars: How to Create Your Own With AI
The visual grammar that makes an image read as cyberpunk, how to describe it so a model follows, and why the lighting in your source photo decides the result.
Cyberpunk is the easiest aesthetic to ask for and one of the harder ones to get right, because most people describe the setting when the style is actually about light.
The visual grammar
Strip away the subject matter and the style is four things:
Opposed coloured rim light. Two hues from opposite sides — cyan and magenta is the canonical pair, but orange and teal or green and violet work the same way. The face is lit from the edges rather than the front.
Dark base, small bright sources. Most of the frame is deep shadow. The light comes from signs, screens, reflections — small, intense, coloured.
Reflective surfaces. Wet streets, glass, metal, damp skin. Reflection is what multiplies a few light sources into a whole scene.
Dense background detail. Signage, cables, layered structures, atmospheric haze. Suggested rather than resolved.
That is the entire grammar. Notice that none of it is "a person with cybernetic implants" — the implants are a motif, not the style. An ordinary portrait with that lighting reads as cyberpunk. A character covered in chrome, flatly lit, does not.
Why your source photo's light decides everything
This is the practical consequence, and it is the thing most people get wrong.
The style works by replacing the light you had with coloured light coming from roughly the same directions. If your photo has a strong light source on one side, the model has an obvious place to put a magenta rim and an obvious opposite side for the cyan. The result has real three-dimensional form.
If your photo is flat — even frontal light, no directional shadow — there is nothing for the coloured light to attach to. You get a tint rather than a transformation. It looks like a photo with a filter on it, because functionally that is what happened.
So: pick a photo with visible directional light. A window on one side, a lamp off to one side, anything that puts one half of your face in more light than the other. Evening and indoor photos are usually better source material than bright daylight, whose shadow directions and warm colour actively contradict the style.
Writing the instruction
Describe the grammar, not a franchise:
Cyberpunk portrait, cyan and magenta rim lighting from opposite sides, dark background with distant neon signage, wet reflective surfaces, atmospheric haze, keeping the facial features defined.
Each clause does work:
- The colour pair is the single strongest signal. Name it explicitly.
- "Rim lighting from opposite sides" is what produces form rather than a tint.
- "Dark background" stops the model brightening everything and flattening the contrast the style depends on.
- "Wet reflective surfaces" buys the reflections without you having to describe each one.
- "Keeping the facial features defined" is the clause people omit, and it is what stops the style eating the likeness.
Keeping it recognisably you
Coloured light removes mid-tones, and mid-tones are where facial structure lives. That is the tension: the more committed the lighting, the less of your face survives.
What helps:
- Crop close. Head and shoulders. More pixels on the face means more structure to keep.
- Start from even, directional light rather than a photo already in heavy shadow — the model needs to see your face before it can relight it.
- Ask for preservation, as above.
- Generate three times. Strong styles vary a lot between runs, and one attempt is not a verdict.
The general principle — that in a stylised avatar the likeness is carried by hair, face shape and accessories rather than by features — applies here too, and is covered in turning a selfie into anime art.
Variations worth knowing
- Solarpunk — same structural idea, inverted palette: daylight, greenery, warm gold and green. Bright base rather than dark.
- Vaporwave — pink and cyan, but flat and hazy rather than high-contrast, with gradients instead of rim light.
- Blade Runner-ish noir — heavier haze, more orange, fewer colours, deeper blacks.
- Synthwave — grid horizons, sunset gradient, less about the portrait and more about the scene.
Each is a different lighting recipe rather than a different subject, which is why describing the light gets you there faster than naming the genre.
Where to use it
A stylised avatar belongs where stylisation is expected — gaming, community platforms, a handle-based account. Not where someone is deciding whether you are who you say you are; that trade-off is in AI avatar vs professional photo.
For one character across several platforms, generate once from one anchor photo and crop per platform: creating consistent avatars.
Restyling runs in Kitana's browser studio; the dedicated avatar tool is among the thirteen in the apps.
Frequently asked questions
- What actually makes an image look cyberpunk?
- Four things, and none of them are the subject: coloured rim light from two opposing hues, usually cyan and magenta; a dark base with small bright sources; wet or reflective surfaces; and dense background detail suggesting a city. Get the light right and almost anything reads as cyberpunk.
- Do I need neon in the source photo?
- No — the model adds it. What you need is directional light, because the style works by replacing the light you had with coloured light from the same directions. A flat, evenly lit photo gives it nothing to build on and usually comes back looking like a photo with a colour overlay.
- Why does my face go unrecognisable?
- Strong coloured lighting removes the mid-tones that carry facial structure, and heavy stylisation compounds it. Keep the face large in the frame and ask the instruction to preserve facial features, which gives the model something to weigh against the style.
- Should I mention specific films or games?
- Describing the visual qualities works better and is less restricted. Naming a title asks the model to recall a specific look it may have seen inconsistently; naming the qualities — rim light, colour pair, reflections, haze — steers reliably.
- Can I use a daytime photo?
- Yes, but expect to fight it. The style is nocturnal and a daylight source has shadow directions and colour temperature that contradict it. An evening or indoor photo with one strong light source is much easier material.
- How do I keep a series consistent?
- Use one anchor photo for every generation and vary only the style wording. Generation varies run to run, so consistency has to come from the input rather than from luck.
Ready to put this into practice?
Create with Kitana using the tool that fits this guide.
Create a cyberpunk avatarReady to try it yourself?
Download Kitana and create your first AI photo in under a minute.