GPT Image 2
OpenAI's image model for reference-heavy edits, product scenes, typography tests, and high-resolution image work in Banana Pie.
Informations clés
| Développeur | OpenAI |
|---|---|
| ID officiel du modèle | gpt-image-2 |
| Date de sortie | 2026-04-21 |
| Disponibilité sur Banana Pie | Disponible |
| Crédits à partir de | 20 crédits / image |
| Images de référence | Jusqu’à 8 |
GPT Image 2 is a strong fit when an image has to respect references, materials, text, or layout instead of simply looking plausible. It is not the cheapest starting point, so use it when the prompt has enough constraints to benefit from a more deliberate image pass.
Points forts
- Reference-heavy edits
Banana Pie lists support for up to 8 reference images, and the evidence includes six-reference object binding, identity and garment transfer, and product preservation across scene change.
- Product and lighting control
The test suite includes product material and lighting, coherent scene relighting, and product preservation, which are practical checks for commerce and campaign images.
- Text and layout scenarios
Evidence includes multilingual typography, structured UI graphics, and in-image text replacement, making it a sensible option when text and layout are part of the image brief.
- High-resolution output
With 4K listed as the maximum resolution, GPT Image 2 can be used for high-resolution image outputs when the composition is worth refining.
Limites
- Higher starting cost
At 20 credits minimum, GPT Image 2 is better suited to serious drafts and finals than throwaway prompt exploration.
- No comparative winner claim
Evidence runs were selected by the auto:only-run rule, so the notes show tested coverage rather than side-by-side ranking against other models.
- Image-only workflow
The evidence covers image scenarios only, and the runtime facts list this as an image model.
- Reference count is bounded
Reference capacity goes up to 8, so projects needing broader batch identity or catalog conditioning may need a different workflow.
Choisissez GPT Image 2 si
- You need an image model that can work from multiple references.
- You are making product, campaign, or catalog visuals where material, lighting, and preservation matter.
- Your prompt includes typography, UI-like structure, or in-image text edits.
- You want up to 4K output for a still image.
Envisagez une alternative si
- You are only exploring rough concepts and want the lowest-credit way to test many prompts.
- You need video generation rather than a still image.
- Your image depends on more than 8 reference inputs.
- You need a claim that it outranks other models; the provided evidence does not establish that.
Résultats en conditions réelles
Les mêmes prompts fixes sont utilisés pour chaque modèle : une exécution chacun, publiée telle quelle.
Prompt
A premium product photograph of exactly one transparent rectangular perfume bottle, half filled with amber liquid, standing upright on wet black stone. Light comes from the upper left, creating coherent refraction, a contact shadow, and one subtle reflection. No text, logo, plants, or extra objects.
Paramètres
- Resolution: 1K
- Aspect Ratio: 1:1
Prompt
Design a clean cream-colored vertical event poster. Show exactly four centered text lines and no other text: "MOONLIGHT MARKET", "月光市集", "18 OCT", and "RIVER HALL". Render "月光市集" in red and all other lines in black.
Paramètres
- Resolution: 1K
- Aspect Ratio: 2:3
Prompt
A photorealistic close-up of an adult potter shaping a clay bowl on a spinning wheel. Both hands are fully visible, each with five natural fingers touching the clay. No other people or hands. Soft window light.
Paramètres
- Resolution: 1K
- Aspect Ratio: 3:4
Prompt
On a matte gray table, exactly three red wooden cubes form a row on the left, exactly two blue glass spheres sit on the right, and one yellow ceramic mug stands centered behind them. The mug handle points right. No other objects or text.
Paramètres
- Resolution: 1K
- Aspect Ratio: 4:3
Prompt
Create a vertical social ad for a fictional running shoe named AERO. Keep the top 15 percent empty. Directly below it, place the headline "RUN LIGHT" in the upper-left. Show exactly one silver shoe in the lower-right, an orange trail curving from the bottom-left, and a round badge reading "42 KM". No other shoes or text.
Paramètres
- Resolution: 1K
- Aspect Ratio: 9:16
Prompt
Create a clean horizontal pricing comparison graphic titled "CHOOSE YOUR PLAN". Use exactly three equal columns labeled "STARTER", "PRO", and "TEAM". Under each column, show exactly three aligned rows labeled "PROJECTS", "STORAGE", and "SUPPORT", followed by one blue button labeled "SELECT". White background, dark navy text, no additional columns or text.
Paramètres
- Resolution: 1K
- Aspect Ratio: 16:9
Images de référence
Prompt
Use Reference 1 as the base tabletop scene. Replace the red mug with the exact dark navy-blue mug from Reference 2 in the same position and orientation. Replace the folded white towel with the exact green-and-white striped napkin from Reference 3 in the same folded area. Replace the green pear with exactly one orange from Reference 4 in the same position. Replace the blue notebook with the exact mustard-yellow hardcover notebook from Reference 5 in the same position. Add exactly the two lemons from Reference 6 to the right of the mug. Preserve the wooden table, camera, framing, wood grain, lighting, shadows, and all other spatial relationships from Reference 1. Do not copy the white product backgrounds from References 2–6, duplicate any asset, or add other objects.
Paramètres
- Resolution: 1K
- Aspect Ratio: 4:3
Images de référence
Prompt
Replace only the sign text with exactly "NIGHT OWL". Preserve the original font style, spacing, perspective, sign material, lighting, and everything else.
Paramètres
- Resolution: 1K
- Aspect Ratio: 3:2
Prompt
Use the portrait in Reference 1 for the person and the isolated jacket in Reference 2 for the garment. Dress the person from Reference 1 in the exact jacket shown in Reference 2. Preserve the person's identity, face, expression, skin, hair, hands, pose, body proportions, background, framing, and lighting from Reference 1. Preserve the jacket's material, color, collar, buttons, pockets, and sleeve patch from Reference 2. Do not copy the ghost mannequin or white product background.
Paramètres
- Resolution: 1K
- Aspect Ratio: 3:4
Images de référence
Prompt
Change the lighting to warm golden-hour sunlight entering from the left window. Do not move, add, remove, or redesign any object. Update highlights and shadows coherently.
Paramètres
- Resolution: 1K
- Aspect Ratio: 16:9
Images de référence
Prompt
Expand the canvas to a 16:9 landscape by naturally continuing the beach, ocean, and sky on both sides. Keep the complete original image centered without cropping, stretching, letterboxing, or modifying it.
Paramètres
- Resolution: 1K
- Aspect Ratio: 16:9
Images de référence
Prompt
Place the sneaker on a wet outdoor basketball court at dusk. Preserve the exact sneaker shape, black geometric side mark, materials, stitching, sole geometry, and camera angle. Add physically coherent contact, reflections, and dusk lighting.
Paramètres
- Resolution: 1K
- Aspect Ratio: 1:1
Crédits et tarifs
GPT Image 2 starts at 20 credits in Banana Pie. In practice, treat it as a model for intentional image attempts where references, text, layout, or preservation matter enough to justify the spend.
Questions fréquentes
Is GPT Image 2 available in Banana Pie?
Yes. GPT Image 2 is listed as available in Banana Pie for image generation.
Can I use reference images with GPT Image 2?
Use it when you need to combine or edit several references in one image. Banana Pie lists support for up to 8 reference images, and the test set included six-reference object binding, identity and garment transfer, and product preservation across scene changes.
What resolution can GPT Image 2 generate?
Banana Pie lists a maximum resolution of 4K for GPT Image 2.
What kinds of image tasks was it tested on?
The published test selections cover product lighting, multilingual typography, structured UI graphics, text replacement, relighting, outpainting, and reference-based edits. They show the kinds of prompts used to evaluate the model, not a guarantee for every image.
When should I not spend credits on it?
Choose another model if you need very low-cost drafts, fast iteration above all else, or video generation. GPT Image 2 is positioned here as a careful image model for reference-heavy and detail-sensitive work.
Méthode de test et transparence
Conclusions come from a fixed prompt suite and the published selected runs for GPT Image 2, covering product material and lighting, multilingual typography, anatomy and contact, counting and attribute binding, advertising composition, structured UI graphics, reference-based edits, text replacement, relighting, outpainting, and product preservation.
Banana Pie propose un accès payant à ce modèle, aux côtés d’autres modèles dans un même studio. Nos verdicts reposent sur des tests réalisés via le même processus que celui proposé à nos utilisateurs.