Nano Banana Prompts | a free prompt library for pro AI images

Prompt design still does most of the heavy lifting in AI image generation. The Nano Banana collection makes that craft visible and repeatable through a large open library of structured prompts, organized examples, and a companion dataset designed for identity consistent work.

The library covers a wide range of visual categories. Playful styles like voxel worlds and stylized posters sit alongside practical production cases: infographics, product mockups, and storyboard frames. The structure behind each prompt is what separates this from a generic collection of one liners.

What the Library Contains

Each prompt in the collection specifies lensing, lighting, palette, material, and layout explicitly. Where most prompt collections give you a starting point, Nano Banana gives you a template with variables you can swap for your specific project.

Control over text areas and negative space is built into many of the templates. For boards, mood pieces, or pitch stills, that kind of purpose built structure cuts trial and error from hours to minutes. Teams working toward a shared visual brief can align on a look without arguing over adjectives.

The examples come with production context. Rather than standalone prompts, many entries show a goal, the prompt structure used to reach it, and the output. That transparency makes the library useful for learning prompt logic rather than only copying results.

The Nano Consistent 150K Dataset

Alongside the prompt library, the maintainers released Nano consistent 150K, a large synthetic dataset for training or evaluating identity consistent image editing. The dataset covers face consistent generation across style changes, outfit swaps, and background variations.

Even for teams that will never train a model, the dataset explains why the prompts work the way they do. They were designed with repeatability in mind, which is visible in how precisely each template defines what can change and what must stay stable.

The 150K scale is meaningful for evaluation purposes. Studios testing identity preservation across a set of brand characters or talent likenesses have enough volume to measure consistency before committing to a production pipeline.

Identity Consistent Editing

The library's editing examples cover virtual makeup, outfit swaps, and packaging changes that keep faces stable while altering surrounding elements. For brand work, posters with recognizable talent, and any sequence where continuity will be obvious to viewers, that stability matters.

Altering a single attribute without changing the rest of the frame requires a different prompt structure than generating from scratch. The Nano Banana templates encode that distinction explicitly: they show which parameters to pin and which to leave open for variation.

For production, this translates to fewer fixes in compositing and less time rebuilding elements in a design tool when a client requests a change. A consistent identity across a campaign can be maintained through prompt discipline rather than repeated manual retouching.

The Likeness Controversy

In 2026, a prompt circulating online demonstrated that Nano Banana 2 could reproduce a specific creator's likeness from a text description alone, with no image reference uploaded. The model had learned the face well enough that image inputs became unnecessary.

That demonstration raised questions about what it means when a generative model reaches that threshold. The broader legal and ethical implications of text only likeness generation are covered in our analysis of the YouTube creator likeness crisis.

The incident is relevant to anyone using the library for production work involving recognizable individuals. Consent documentation and a clear rights check should precede any generation involving real faces, regardless of which model is used.

The controversy also highlighted a gap in how creative tools communicate capability to users. A prompt library that was designed for expressive image composition becomes something different when the model running those prompts can produce accurate likenesses from minimal text input. That shift happens at the model level, not the prompt level, which means library users benefit from tracking model capability changes separately from prompt technique changes.

For production contexts, the practical response is a per project rights review at the start of generation work, not at the end. Reviewing model capabilities and consent requirements before generating protects both the production and anyone whose likeness could appear in the output. The Nano Banana library's prompt templates are neutral on this question, but the models you run them on are not.

How to Use the Library in Production

Start by browsing the showcase and selecting a style already close to your brief. Copy the full prompt rather than summarizing it. Small tests at low resolution establish the seed, guidance, and step count before committing to a high resolution run.

Pay attention to the parts of the template that define composition and text areas. If you need space for headlines or UI elements, the sections governing negative space and text placement are where you make those calls before moving to final resolution.

When the look is right, move to your refiner or high resolution pipeline and keep the seed fixed so variations remain comparable. Treat every prompt like a scene recipe. Save the version you showed to stakeholders and label it with the model, seed, and settings used.

That labeling discipline pays off on longer projects. When a client returns six weeks after a shoot asking for a variation on a look you approved in preproduction, a logged seed and prompt get you back to the same starting point in minutes rather than hours.

Prompt Structure for Production Workflows

The most effective approach starts with the technical parameters: aspect ratio, lighting direction, color temperature, lens character, and period or setting. A prompt that establishes those conditions before describing the subject gives the model cleaner constraints to work from.

Subject description follows the technical framing. After establishing scene conditions, describe the subject specifically. Material properties, scale relative to the environment, surface condition, and frame position all benefit from explicit specification rather than inference.

For prompts with text elements in frame, the templates include guidance on how to specify typeface register, language, and scale. Including that context produces cleaner text output than leaving it to model inference. The difference is visible in signage, labels, and UI elements where character accuracy matters.

Seed management is the most frequently overlooked production practice for batch image generation. If you are generating a set of images that need visual consistency across a project, generate the first satisfactory frame with a logged seed, then vary only the subject or scene parameters while holding the seed constant across the batch.

A batch generated with a shared seed produces frames with consistent material language and color temperature from a common visual starting point. Without the seed, even identical prompts applied to the same model will produce perceptibly different results in lighting and surface quality. Documenting the seed is the minimum version control required for any batch that will be presented to a client or reviewed by a creative director.

Two artists working on the same project and using different seeds will produce outputs that look like different productions, even with identical prompts. A shared prompt document that includes the approved seed resolves this without additional review cycles.

The Nano Banana library's prompt templates are structured to make this discipline easy. Because each template already specifies the parameters that matter most for visual consistency, adopting the library also adopts the habit of explicit parameter documentation. That habit extends naturally to seed management without requiring a separate process.

Licensing and Commercial Use

The Awesome Nano Banana images repository uses Apache 2.0, which permits reuse and adaptation of the templates. The prompts themselves are freely available for creative and commercial use under that license.

The models and datasets you run them on carry their own licenses and restrictions. Apache 2.0 on the prompt library does not extend to generated outputs if the underlying model has a more restrictive license. Before publishing commercial work, check the model card and asset terms independently.

For larger studios, a short rights checklist added to the generation workflow prevents approval delays at delivery. When the legal side is handled up front, the speed advantage of a structured prompt library carries through to the end of the project rather than being lost to clearance work at delivery.

Shared Prompt Libraries Across Production Teams

A prompt library stops being a personal tool and becomes a production asset when multiple artists are generating images for the same project. Without shared prompts and seeds, different artists will produce frames with different color temperatures, lens characters, and material qualities even when targeting the same reference. Consistent prompting is the generation equivalent of a lens list or a color bible.

The practical implementation is a single shared document per project that stores the base prompt structure, any seeds that produced approved frames, the model version, and a note on what each template is designed for. New artists joining a production can generate on brief from the first session without a lengthy handoff.

Prompt libraries also accumulate across projects. A template that worked for a campaign with warm industrial lighting is reusable on the next project that calls for the same look. Over a year of production, a well maintained library represents a measurable reduction in the time spent rediscovering prompt combinations that someone on the team already figured out.

When a frame is approved for use in production documentation, log the full prompt, the seed, the model version, and any refiner settings that produced it. Future generations referencing the same scene type can start from that record rather than rebuilding the prompt from scratch. These records become a searchable reference that reduces iteration time on subsequent projects and lets you reproduce approved work months after the original generation session.

The habit matters more than the format. A single shared document, a Notion page, a plain text file, any consistent system for storing approved prompts reduces the cost of revisiting past work. Teams that maintain this discipline report faster creative review cycles because client presentation is built on a library of approved outputs rather than regenerated from scratch each time.

Using Nano Banana Prompts With AI FILMS Studio

AI FILMS Studio's image workspace supports the same generation workflow the Nano Banana templates are designed for. Multiple models are available in a single interface, which makes prompt testing across models straightforward without switching between tools.

For a complete walkthrough of every setting in the image workspace, from model selection and reasoning levels to the Node Graph Editor, the Nano Banana Pro 2 tutorial covers the full workflow.

The image to image workflow applies the same prompt structure to transform and restyle existing images.

---

Sources

Project page and showcase: picotrex.github.io/Awesome-Nano-Banana-images GitHub: PicoTrex/Awesome-Nano-Banana-images