An idea
Use Text to Image to create a new visual.

Your ideas. A clearer picture.
Create images from text, transform existing photos, or combine visual references in PWZ Studio. Follow a workflow for your starting material, then choose the model and output settings that fit your project.
Start Creating AI ImagesStart with what you have. Build toward what you need.
Use Text to Image to create a new visual.
Use Image to Image and describe a focused edit.
Assign each image a role in the new composition.
Use a supported transparent-background workflow.
Check the output settings and credits before generating, then refine the result with one clear change at a time.
Describe the picture before you generate it.
Use text to image for campaign concepts, illustrations, thumbnails, posters, and visual ideas that do not require an existing photo.

Open PWZ Studio, select Image, choose a model, and switch to Text to Image. Start with a single composition so the result is easy to evaluate.
Include the subject, setting, composition, lighting, and style. For graphics with text, put the exact wording in quotation marks and specify where it belongs.

PROMPT /Create a vertical poster for a weekend flower market. Place a sculptural arrangement of orange tulips in the lower half against a pale lavender background. Add the headline “WEEKEND IN BLOOM” at the top in bold, elegant lettering, with “SATURDAY 10 AM” beneath it. Soft directional light, generous margins, no additional text.
Quick tip: Describe placement with phrases such as “lower half,” “centered,” or “leave clear space on the right.” These give more direction than “make it beautiful.”
Choose an aspect ratio for the intended placement, then select the available quality or resolution. Review the credit amount and generate. Check the overall layout before spending time on small details.
Change the detail. Keep the direction.
Use AI image editing to explore a new background, outfit, material, color palette, or artistic style from an existing image.

Switch to Image to Image and upload the photo you want to edit. Use a clear source that shows the details you want to retain.
State the requested change first, followed by the subject details and composition that should remain recognizable.
PROMPT /Replace the background behind the portable speaker with a warm modern living room. Keep the speaker’s shape, dark green finish, front grille, buttons, and camera angle. Place it naturally on a walnut side table with a soft contact shadow. No extra products or text.
Generate the result and compare it with the source. To continue, use the version you prefer as the reference for another request.
PROMPT /Reduce the background brightness and move the plant farther from the speaker. Keep the speaker and its position unchanged.

Quick tip: For product photos, inspect labels, proportions, controls, and edges. Instructions to preserve details guide the model but do not lock every pixel.
Try Image-to-Video on PWZGive every image a specific job.
Use multiple reference images when one picture defines the person, another shows a product, and a third establishes the environment or visual direction.

Choose clear, relevant images and upload them in an order you can describe. More references are useful only when they add information.
Reference image limits vary by model.
| Model in PWZ Studio | Reference images | Maximum per image |
|---|---|---|
| GPT Image 2.5 | 10 | 10 MB |
| GPT Image 2 | 10 | 10 MB |
| Nano Banana 2 | 14 | 10 MB |
| Nano Banana 2.1 | 14 | 10 MB |
| Nano Banana Pro | 14 | 10 MB |
Supported files: JPG, PNG, and WebP work across these uploaders. Nano Banana Pro also accepts BMP and GIF.
These are upload limits, not guarantees about how many distinct subjects a result will preserve.
Identify images by upload order and visible content. Clarify whether a reference guides identity, product design, setting, or style.
PROMPT /Use Image 1 for the woman’s face and hairstyle, Image 2 for the burgundy handbag, and Image 3 for the sunlit gallery setting. Create one editorial photograph of the woman holding the bag. Preserve its clasp and strap design, use consistent lighting, and keep the scale natural. Do not create a collage.
Review scale, lighting, hand placement, and subject identity. If details conflict, remove a competing reference or explain its role more narrowly.
Quick tip: More references are useful only when they add information.
Create with Reference ImagesMake assets ready for your next layout.
Use transparent images for stickers, isolated illustrations, product cutouts, and elements you want to place over another background.

Select Text to Image for a new asset, or Image to Image to work from an existing picture.
Choose Transparent under Background and select PNG or WebP. JPEG does not preserve transparency.

PROMPT /Create a polished 3D illustration of a friendly orange robot holding a small flower. Show the complete character, centered, with clean edges and a transparent background. No floor, scenery, text, or cast shadow.
Download the result and test it over both light and dark backgrounds. Inspect fine edges and confirm that any visible checkerboard is a preview pattern rather than part of the image.
Match the output to where it will appear.
| Setting | How to choose |
|---|---|
| Aspect ratio | Use square for compact graphics, portrait for posters, and landscape for banners or video frames. |
| Resolution | Nano Banana 2, 2.1, and Pro offer 1K, 2K, and 4K options in Studio. Choose the size your delivery needs. |
| Quality | GPT image workflows provide quality settings. Refine the composition first, then compare higher settings for important details. |
| Image count | GPT image workflows let you request 1–10 images. Begin with fewer outputs while testing the brief. |
| File format | Use a transparency-compatible format for cutouts; choose other available formats for finished images with backgrounds. |
Higher resolution describes output size; it does not guarantee accurate lettering, anatomy, or product details. Review the credit estimate again after changing settings.
Optional search context: Nano Banana 2 and 2.1 expose Google Search and Image Search controls in Studio. Use them when external context helps the subject, and check factual labels before publication. They do not replace your own product or character references.
Choose for the task, then judge the result.
| Creative goal | A starting option |
|---|---|
| Explore several directions quickly | GPT Image 2.5 Flare or Nano Banana 2 |
| Refine small details or a demanding edit | GPT Image 2.5 Sunburst |
| Build a complex layout with several requirements | Nano Banana Pro |
| Try a reference-based workflow with optional search context | Nano Banana 2.1 |
| Create prompt-led graphics or edit an existing image | GPT Image 2 |
Treat these as starting points rather than a quality ranking. Compare a small number of outputs against the same brief before committing to a longer series.
Learn more about AI Image Generators on PWZA clear starting frame makes motion easier to direct.
For image-to-video projects, create a frame with a readable subject, uncluttered surroundings, and room for the intended movement.

Generate an image in the aspect ratio you want for the video.
Download your preferred version, then switch to Video in PWZ Studio.
Open an image-to-video workflow and upload it as the starting frame.
Describe the movement and camera direction rather than repeating the image description.

PWZ STUDIO / READY

No. Text to Image starts from a written prompt. Choose Image to Image when you want an existing picture to guide an edit or composition.
Describe the object, its position, and the exact change in Image to Image. Add a short list of details to preserve, then compare the result with the source before continuing.
GPT image workflows include an image control for 1–10 outputs. A prompt asking for “four panels” may instead produce one composite image, so use the output count when you need separate files.
Do not assume so. Recheck references, aspect ratio, output options, and credits after a model change; the available controls differ.
Review the result on the Studio board and use Download image. When a task returns multiple images, download the outputs you want individually. The Board provides access to your creation history.