AI Photo Effects: Glitch, Double Exposure, Duotone, Hologram

You can apply all five of these effects — glitch, double exposure, duotone, hologram and splash — in the AI image editor by uploading a photo and describing the effect in plain language. No masks, layers or channel mixing are required for most of them: you write what you want, pick a model, and the edit comes back in a minute or two. The exact credit cost is shown before you run it.

Updated October 2026.

What each effect actually is

These five are popular because each one changes the mood of a photo without changing the subject. Knowing what each one does technically makes the prompts much easier to write.

Glitch. Digital damage — horizontal displacement, pixel sorting, chromatic colour separation, scan lines and datamosh artefacts. It reads as cyberpunk, retro-tech or corrupted footage.
Double exposure. Two images blended so both remain visible, traditionally a portrait with a landscape inside the silhouette. It reads as editorial, atmospheric and slightly dreamlike.
Duotone. The image reduced to two colours — a shadow colour and a highlight colour — with all the midtones mapped between them. It reads as poster, album cover or brand campaign.
Hologram. Iridescent, oil-slick, foil-like colour shifts across the surface of the subject. It reads as Y2K, fashion-forward and chrome-adjacent.
Splash. Liquid thrown at, around or out of the subject — paint, milk, water, ink. It reads as energetic advertising.

How to apply any of them on VdoBloom

The AI image editor is prompt-driven: you describe the change and the model makes it. The flow is the same whichever effect you want.

Upload your image. For double exposure you can upload two — many models accept several reference images at once, and the Nano Banana 2 and 2.1 models take up to fourteen, which is more than these effects need.
Describe the edit in plain language. Say what should change and what should stay. Naming the parts you want kept matters more than naming the parts you want changed.
Choose a model. The default is Auto, described as balanced quality, which prices at 7 credits and tries WAN 2.7 first before falling back to Seedream and Qwen. For a specific effect, a specific model is often better.
Set the output resolution and shape if the model offers them, then apply. Results come back in a few minutes.
Iterate. Effects like these are rarely right on the first attempt. Change one word at a time — the strength of the effect, or the area it covers — rather than rewriting the whole prompt.

Two specialised modes also sit alongside the prompt box. The Flux Erase mode lets you paint the areas you want removed and costs 10 credits; Flux Outpainting expands the canvas in a chosen direction — top, right, bottom or left — for 12 credits. The Qwen 2.1 edit model adds a mask mode and a transparent background toggle for 3 credits at 1K and 5 at 2K, which is the cheapest way to cut a subject out before applying an effect.

Glitch

The trick with glitch is restraint. A full-frame glitch destroys the photo; a glitch that affects one band of the image leaves the subject readable, which is what makes it usable.

“Add a subtle digital glitch to this portrait: thin horizontal RGB displacement across the middle third only, faint scan lines, the face itself perfectly clean and sharp.”
“Make this image look like a corrupted video frame: datamosh banding across the lower half, chromatic separation on the edges, heavy grain, subject unchanged.”

A model with fine control over the edit area gives the best results here, because most of the frame is supposed to stay unchanged. Nano Banana 2.1 is the cheapest of the narrow-editing options at 3, 4 and 6 credits depending on resolution.

Double exposure

Traditional double exposure needs the silhouette of one image to hold the detail of another. You can supply that second image as a reference, or describe it and let the model generate it.

“Double exposure of this portrait with a pine forest at dusk: the trees and mist visible inside the silhouette of the head and shoulders, the background outside the figure almost white, high contrast, fine grain.”
“Double exposure of the product with ocean waves, water texture contained within the pack silhouette, clean white background around it.”

Pick a model that accepts a reference image when you want a specific second photo rather than one the model invents. Nano Banana Edit takes up to ten images per edit and is a flat 5 credits.

Duotone

Duotone is the most predictable of the five because it is a colour operation, not a structural one. The main choice is the colour pair.

“Convert to a duotone in deep navy and warm cream, high contrast, keep all detail in the shadows, no third colour anywhere in the image.”
“Two-colour screen-print look in burnt orange and off-white, poster style, flat graphic feel, slight paper texture.”
“Magenta and cyan duotone, club-poster energy, posterised midtones, black point preserved for depth.”

Because duotone is a global change, almost any editing model handles it; the 3-credit tier on the GPT Image 2 and 2.5 variants is the cheapest route.

Hologram

Holographic treatments are about specular colour: the surface shifts hue as it curves, like foil or an oil slick. The effect works best on objects with a defined form to catch the light.

“Apply a holographic foil treatment to this product: iridescent pink, mint and violet shifting across the curved surface, sharp specular highlights, clean studio background, no other change.”
“Y2K holographic portrait: skin untouched, but the jacket and hair rendered in shifting chrome-iridescent colour with strong reflections and a soft glow.”
“Subtle holographic sheen over the whole image, low strength, colours shifting gently from teal to lilac, subject and detail fully preserved.”

Hologram is where over-editing is most common. Ask for a low strength first, because once the effect covers the whole frame the subject stops being visible.

Splash

Splash shots are the most commercial of the five and the most likely to need a second attempt, because the liquid has to interact plausibly with the object.

“Milk splashing around this product, frozen mid-air, crown-shaped splash rising from a white surface, crisp droplets, high-key studio lighting, product perfectly sharp and unchanged.”
“Water droplets suspended around this bottle, fine mist at the base, backlit rim light, dark blue gradient background, product label fully legible.”

Say explicitly that the product must stay unchanged, because splash prompts otherwise tempt the model into redrawing the object to match the liquid. If your product is a food or drink item, the same treatment is available as motion in the platform’s food and beverage video tools rather than as a still.

What it costs

Editing prices are per image and depend almost entirely on the model. At the low end, Muse Image is 1 credit, Boogu Image Edit and the 1K tier of Qwen 2.1 and GPT Image 2.5 are 3 credits, and Seedream 5 Flash, FLUX.2 dev and Grok Imagine are 4 to 5. The Auto default is 7. The speciality modes are the expensive ones: Flux Erase at 10 credits and Flux Outpainting at 12, because both do a more specific job than a prompt.

Because an effect rarely lands first time, budget for three or four runs per final image on a cheap model and move only the winning prompt onto an expensive one. Every cost is displayed before you generate, and the plans page explains the allowances.

Common mistakes

Applying the effect everywhere. Every one of these five looks better when part of the frame is left alone.
Not saying what must stay the same. Name the subject, the face and the label as things to preserve.
Rewriting the whole prompt between attempts. Change one variable at a time or you will not learn which word caused the difference.
Using an expensive model to search. Find the prompt on a 1 to 5 credit model, then re-run the winner on the model you want for the final.
Expecting a mask-based effect from a prompt. For solid cut-outs, the background remover at 1 credit for the standard pass or 2 for HD is cheaper and cleaner than describing it.
Judging the result at low resolution. Fine detail in glitch and hologram vanishes when the image is small, so check at full size.

If you want a decade-specific look to go with these effects, our guides to the 1980s and 1990s retro photo look and to fixing a blurry AI image cover the rest of the toolkit, and the AI image editor tool page has the full model list.

Frequently asked questions

Can I add a double exposure effect without Photoshop?

Yes. Upload the main image and, if you have one, the second image as a reference, then describe how you want them blended — for example, the forest visible inside the silhouette of the head with a near-white background. The editor applies it in one pass and you refine from there.

What is a duotone and how is it different from black and white?

A black-and-white image uses one ink on a range of greys. A duotone uses two colours, one for the shadows and one for the highlights, with the midtones mapped between them. It is what gives posters and album covers their flat, printed look, and it is one of the easiest effects to get right because it is a purely colour-based change.

Why does my hologram effect cover the whole photo?

Because the prompt asked for it globally. Ask for the treatment on a named part of the image — the jacket, the packaging, the surface — and specify that the rest is unchanged. Starting at a low strength and increasing it is faster than trying to reduce an effect that is already at full coverage.

How much do AI photo effects cost on VdoBloom?

Each edit is one image on the model you choose, and prices range from 1 credit on Muse Image to 12 on Flux Outpainting. Common choices sit between 3 and 7 credits: Qwen 2.1 and GPT Image 2.5 at 3 credits for 1K output, Nano Banana Edit at 5, and the Auto default at 7. The cost is always shown before the edit runs.

Can I apply these effects to a product photo?

Yes, and glitch, hologram and splash are the three that suit products best. State clearly that the product and any printed label must remain unchanged, then check the label at full resolution — heavy effects sometimes re-letter small text, and AI models are not reliable at reproducing lettering.

Open the AI image editor, upload your photo, describe the effect you want and generate — the credit cost appears before the edit runs.