Promptsref supports Nano Banana Lite image generation at $0.03 per image

• Promptsref's launch post says Promptsref now supports Nano Banana Lite, with image generation starting at $0.03 per image.
• Google's launch post and OfficialLoganK's launch post frame Nano Banana 2 Lite as the fast, cheap Gemini image model: under 4 seconds at about $0.034 per 1K-resolution image.
• underwoodxie96's follow-up comparison still preferred GPT Image 2.0 at that price point for more consistent image quality and finish.
• the Leonardo availability note, the Pollo AI launch post, and Magnific's unlimited post show the model spreading through creator tools almost immediately.

Promptsref linked a full selfie-style prompt built like a JSON production bible, while Google's pricing page gives the exact token math: $30 per million image output tokens, or $0.0336 for a 1K output. The odd split is speed versus finish: one Promptsref comparison favored GPT Image 2.0 for polish, while a Hacker News commenter who claimed early access said Lite behaves closer to Nano Banana 2 than Nano Banana 1 on text rendering.

Promptsref's first example was not a tiny prompt. It was a production spec for a late-night phone selfie with camera angle, finger obstruction, wardrobe, skin texture, background clutter, constraints, and negative prompts.

The reusable structure is clear:

• Camera: high-angle handheld selfie, close-range wide-angle, vertical crop.
• Imperfection: blurred finger obstruction, motion blur, low-light phone noise, messy hair.
• Lighting: desk lamp, laptop screen, city window light, warm-cool shadow mix.
• Continuity: preserve face shape, hair strands, clothing, accessories, and room details.
• Guardrails: adult subject, no explicit nudity, no plastic skin, no watermark, no anime face.

That is where the $0.03 price gets interesting for prompt builders: it makes long, constraint-heavy prompt testing cheap enough to treat as sketching.

Google calls the underlying model Nano Banana 2 Lite, or gemini-3.1-flash-lite-image. Google's developer launch post says it is available in Google AI Studio, the Gemini API, and Gemini Enterprise Agent Platform, with rollout across AI Mode in Search, Gemini, NotebookLM, Photos, Stitch, Flow, and Google Ads.

Google's model documentation lists the practical limits: 65,536 input tokens, 4,096 output tokens, up to 14 input images per prompt, 1K output resolution, and supported aspect ratios from 1:8 to 21:9. The same docs list image generation, image editing, multi-turn image editing, image generation from video input, system instructions, thinking, C2PA content credentials, batch inference, and provisioned throughput.

Google's pricing page puts standard paid output at $30 per 1M image output tokens, equivalent to $0.0336 per 1K image. Batch pricing halves that to $0.0168 per 1K image.

underwoodxie96 posted the most blunt value check: Nano Banana Lite is live, but GPT Image 2.0 still had the more consistent finish in their test.

Creator comparisons are already treating Lite as the fast draft tier rather than the final-quality ceiling. MayorKingAI's Leonardo thread tested Nano Banana 2 Lite against Nano Banana 2 and Nano Banana Pro, then expanded the comparison across a small prompt pack:

• The first Leonardo prompt used a blockbuster explosion portrait.
• The third Leonardo prompt used a rain-soaked overhead portrait with neon reflections.
• The fifth Leonardo prompt used a raw studio boxer portrait.
• The seventh Leonardo prompt used a 1960s convertible fashion editorial.

MayorKingAI's speed-and-quality reply said Lite was surprising on speed and quality. AIwithSynthia's Pollo AI post called it the fastest and most cost-efficient option in the Nano Banana family for high-volume comparison visuals, product concepts, and social content.

The first wave is less about a single app launch and more about distribution. Nano Banana 2 Lite is already being wrapped into image, video, and design workflows.

• Leonardo: the Leonardo availability note says the model can create readable text, make precise edits, and use both image and text references.
• Pollo AI: AIwithSynthia's Pollo AI post says Nano Banana 2 Lite is available there with a 50% off promotion.
• Magnific: Magnific's unlimited post says Nano Banana 2 Lite is unlimited on Magnific and frames the wait time at 5 seconds.
• Weave: Figma Weave's post says Nano Banana 2 Lite and Gemini Omni Flash landed in Weave with 50% off the first month for eligible plans.

Magnific also published a three-step story workflow in its tutorial post: generate the first image with Nano Banana 2 Lite, go to Edit and Variations, then add the storyboard as a reference on Seedance 2.0 4K.

Promptsref is positioning the model inside a larger prompt-discovery system, not as a standalone generator. underwoodxie96's Promptsref note says the site has collected over 10,000 prompts and uses AI algorithms to connect similar styles, characters, scenes, and use cases.

The site's SREF feed gives that library a style index:

• promptsref's retro neon cartoon post highlights --sref 2113922359 --v 8.1 --sv 6 for vintage animation shapes, neon palettes, grotesque fairy-tale characters, music posters, indie game art, and brand visuals.
• promptsref's neon doodle post highlights --sref 3868059321 --v 8.1 --sv6 for editorial photos covered in candy-colored marker lines, graffiti symbols, and pop-art overlays.

The useful combination is a searchable prompt atlas plus a cheap renderer. For creators who already collect SREFs, camera recipes, and negative prompts, Nano Banana Lite gives Promptsref a low-cost way to turn browsing into rapid output testing.