SOUL 2 Vs. Nano Banana Pro 💥 Higgsfield ha lanzado su nuevo generador de imágenes SOUL 2 ⚡ Puedes subirle hasta 80 imágenes de referencia de tu personaje para mantener mejor la constancia 👀 Y para compararlo bien, lo he puesto a prueba junto a Nano Banana Pro que hasta el momento es mi generador de imágenes favorito 💕 La verdad es que hay algunos resultados de SOUL 2 que me han sorprendido bastante... No está nada mal, pero sigo prefiriendo Nano Banana para la mayoría de las ocasiones 😅 Os dejo algunas imágenes que he generado y espero leer vuestras opiniones en comentarios 💌 Y si quieres los prompts de todas las imágenes comenta "ARIA" y te los mando por mensaje!

Aria Cruz | Influencer AI

@soy_aria_cruz · Digital creator

INSTAGRAM · 2026-02-26Source

1.3Klikes

920comments

Remix This

Recreate with AI Image Generator

Make your own AI viral video

Prompt

[Subject] A split-screen low-light bedroom portrait comparison showing the same young woman lying in bed while illuminated mostly by a smartphone screen. She appears to be in her 20s with fair skin, dark brown to black hair pulled loosely back into a high ponytail or gathered style, thin round silver eyeglasses, and silver hoop earrings. Her face is relaxed and intimate, with a tired late-night expression. In the right panel, part of a phone is visible in the lower foreground, casting cool blue light upward onto her face. Reflections of a bright app or screen interface are visible in both eyeglass lenses. The right panel should appear slightly cleaner, more dimensional, and more realistic than the left while preserving the same identity.

[Environment] Dark bedroom setting with a white pillow behind the subject's head, deep blue-black shadows, and no visible room clutter beyond bed linens. The image is presented as a two-panel mobile comparison card with rounded vertical panels and a narrow dark divider. At the bottom left, include the label “Higgsfield SOUL 2” in white. At the bottom right, include “NANO-BANANA PRO” in white with a small colorful sparkle-style icon above or near the text. Keep the mood intimate, nocturnal, and minimal.

[Composition/Camera] Vertical split-screen close-up portraits, framed from forehead to upper chest, with the subject's head resting on a pillow. Both panels are shot at close selfie distance and from a slightly high or side angle as if the camera is above the face. The left panel is tighter on the face; the right panel is slightly wider and shows more pillow plus the blurred phone edge in the foreground. Maintain matched identity and low-light conditions across both panels for direct comparison.

[Lighting] Primary illumination comes from the smartphone screen, producing a cool blue-white glow on the face and visible rectangular reflections in the glasses. Ambient room light is very dim. Keep the skin softly lit with realistic low-light falloff, shadow depth, and subtle screen-light gradients. Avoid flash or warm lamp lighting.

[Style/Rendering] Hyper-real low-light selfie comparison graphic, intimate bedroom realism, natural skin texture, believable digital screen reflections, subtle noise-free darkness, and clean social comparison design. The image should feel like a realistic test of how well different models render phone-lit faces in a dark room.

[Detail constraints] Keep exactly one female identity duplicated across two panels. Preserve the round glasses, hoop earrings, white pillow, cool phone light, app reflection in the lenses, and bottom text labels. Do not add extra people, bedroom decor, heavy makeup, fantasy glows, or bright room lights. The main point is controlled low-light realism comparison between two image models.

Negative prompt: extra people, mismatched identity between panels, warm bedside lamp, bright room, no glasses reflections, blurry low-res selfie, glamour retouch, fantasy glow, incorrect text, random bedroom clutter, cartoon shadows, distorted face, duplicated phone, watermark, heavy grain, overexposed skin.

Suggested parameters: aspect ratio 4:5, close selfie lens feel, low-light phone illumination, steps 28-40, CFG/style strength 6.5-8, sampler DPM++ 2M Karras or equivalent, seed suggestion 613984251.

Delta prompt strategy:
1. If the two identities drift: “same exact young woman in both panels, matched facial structure, glasses, earrings, and hairline”.
2. If the phone-light effect weakens: “cool smartphone screen light illuminating the face from below or front with visible rectangular screen reflections in the glasses”.
3. If the bedroom becomes too visible: “dark minimal bedroom, only white pillow and bed shadows visible”.
4. If the right panel is not stronger: “right panel slightly more realistic, dimensional, and clean than the left while keeping the same concept”.
5. If the text labels fail: “bottom left text Higgsfield SOUL 2, bottom right text NANO-BANANA PRO”.
6. If the mood becomes too polished: “late-night intimate phone selfie in bed, sleepy realistic mood, no studio polish”.
7. If the glasses disappear: “thin round silver eyeglasses with crisp digital reflections”.
8. If the phone in the right panel is missing: “blurred smartphone edge visible in the lower foreground of the right panel”.
9. If the lighting becomes flat: “deep shadowed room with soft cool screen glow as the main light source”.
10. If it turns into a beauty shot: “authentic nocturnal selfie realism, subtle fatigue, minimal styling, not glamour editorial”.

How soy_aria_cruz Made This Low Light Bed Selfie AI Comparison

This image is a much harder test than it looks. Low-light selfies in bed are familiar to everyone, which means they are easy to judge. People instantly notice when the shadows feel fake, when the screen reflection makes no sense, or when skin turns plasticky under blue phone light. That makes this kind of comparison more revealing than a bright studio portrait.

The setup is also disciplined in the right way. Same woman, same glasses, same pillow, same late-night mood. The main variable is how each model handles darkness, phone illumination, and reflections. That gives the viewer something useful to evaluate instead of just a vague “which one do you like more?” question.

Why this comparison format is effective

The strongest part of the image is that it targets a realism zone where many generators struggle quietly. Low light is unforgiving. Small mistakes in eye reflections, skin falloff, lens glare, or pillow texture become obvious because the scene is so simple. There is nowhere for the model to hide.

The second strength is emotional familiarity. A phone-lit face on a pillow feels intimate and believable as a real-life moment. That familiarity gives the comparison social traction. Viewers are not evaluating an abstract test scene. They are evaluating a kind of image they have actually taken themselves, which makes the difference between outputs feel more tangible.

Signal	Evidence (from this image)	Mechanism	Replication Action
Hard-to-fake lighting logic	Cool phone glow and rectangular reflections appear in the glasses	Incorrect light behavior becomes obvious immediately in simple dark scenes	Use screen-lit or device-lit portraits as realism benchmarks
Controlled comparison setup	Same woman, same pillow, same bedroom mood across both panels	Viewers can attribute differences to model quality rather than concept drift	Lock identity, framing, and environment before comparing models
Universal familiarity	The scene resembles a common late-night phone selfie	People can intuitively judge whether it feels real	Choose ordinary scenarios that viewers know from experience
Minimal background noise	Dark room and plain pillow keep the face as the main test surface	Simplicity exposes subtle realism errors more clearly	Strip the scene down to one face, one light source, and one surface

Where this visual strategy fits best

Model comparison content: ideal for testing phone-light realism, eye reflections, and low-light skin rendering.
Prompt engineering posts: useful for showing how fair A/B tests should reduce variables.
AI influencer consistency checks: strong because the character has to survive intimate low-light conditions, not just bright editorial scenes.
Bedroom or lifestyle realism experiments: helpful when testing how well a model handles ordinary private moments.

Where it is less effective

High-energy social thumbnails: the mood is quiet and intimate rather than loud.
Brand-forward campaigns: the image has almost no product or environmental context.
Fantasy-art audiences: the value here is technical realism, not imaginative styling.

Three transfer recipes

Laptop-glow transfer
Keep: dark room, one digital light source, intimate framing.
Change: phone to laptop or tablet, pillow to desk or couch context.
Slot template (EN): {same subject} lit only by {device screen} in a dark {everyday setting} shown as a model comparison
Night-call transfer
Keep: close face crop, reflections in glasses, minimal background.
Change: screen content, expression, and angle.
Slot template (EN): {young woman} in a low-light close-up with {screen reflections} compared across {two image models}
Wake-up realism transfer
Keep: pillow context, intimate bedroom framing, soft tired expression.
Change: light source to dawn screen mix or alarm-clock glow.
Slot template (EN): {subject} lying in bed under {single cool light source} in a split-screen realism benchmark

Aesthetic read: why the image feels believable

The strongest visual detail is the lens reflection. The rectangular phone interface reflected in the glasses instantly tells the viewer where the light is coming from. That is a tiny detail, but it creates a lot of trust. The entire image feels more grounded because the lighting has a readable source.

The pillow and the darkness are also working hard. White bedding gives the face a soft framing surface, while the dark room removes distractions and increases the importance of the screen glow. That makes the scene feel intimate without becoming overly designed.

The blue cast is restrained, which is another good choice. If the phone light were too intense, the image would look stylized or artificial. Instead, it stays close to the slightly tired, very real mood of late-night scrolling.

Observed	Why it matters
Rectangular screen reflections in the glasses	Creates believable lighting logic and a strong realism cue
White pillow against a dark room	Frames the face while keeping the environment minimal
Cool phone-lit skin tones	Defines the nocturnal mood without needing extra props
Right panel includes part of the phone in the foreground	Makes the source of the light and the selfie context more explicit
Matched subject identity across both halves	Keeps the comparison fair and easy to interpret

Prompt technique breakdown

To recreate a useful comparison like this, the prompt should behave almost like a lab setup. The challenge is not wardrobe or background design. It is how accurately the model can render a face under one small digital light source in darkness.

Prompt chunk	What it controls	Swap ideas (EN, 2–3 options)
Late-night face on a pillow	Defines the intimate bedroom context and framing	in-bed selfie; pillow close-up; nocturnal bedroom portrait
Smartphone screen as main light source	Creates the technical realism challenge	phone glow portrait; device-lit face; cool screen illumination
Round glasses with reflections	Adds a hard transparency and lighting test	digital reflections in lenses; screen glare on glasses; reflective eyewear in darkness
Minimal dark background	Removes distractions and exposes small rendering errors	dark bedroom; shadowy bed scene; low-light minimal interior
Split-screen model labels	Makes the image instantly readable as a comparison post	two-panel benchmark card; A/B comparison graphic; model-versus layout
Right panel slightly more realistic	Produces a meaningful visual outcome without altering the concept	cleaner right-side realism; stronger right-panel dimensionality; more believable right output

Starter prompt block

split-screen low-light bed selfie of the same young woman with round glasses and hoop earrings, lying on a white pillow in a dark bedroom, illuminated by a smartphone screen with visible app reflections in the lenses, left labeled Higgsfield SOUL 2 and right labeled NANO-BANANA PRO, hyper-real comparison graphic

Remix playbook

The cleanest way to iterate on this image is to protect the lighting logic first.

Baseline lock

Lock the same face, glasses, and pillow position first.
Lock the same screen-light angle and reflection pattern second.
Lock the same dark-room environment before changing only the model.

One-change rule

If one version gets a brighter room, a different expression, or cleaner styling, the comparison becomes much less useful. Keep the scenario stable and change only the rendering engine.

Run 1: establish the pillow, face crop, and dark bedroom mood.
Run 2: refine the cool phone light and lens reflections.
Run 3: standardize the split layout and bottom labels.
Run 4: judge only realism quality, not different stylistic choices.

That is what makes the image valuable. It is not dramatic, but it is disciplined. And disciplined comparisons are the ones people trust.