5 秒內完成生成
體驗超低延遲，以每張 1K 圖像約 4 秒的速度進行圖像生成。速度比旗艦模型快 2.7 倍，非常適合即時面向使用者的應用程式。
Nano Banana 2 Lite (Gemini 3.1 Flash-Lite Image)
Nano Banana Lite API 驅動 Gemini 3.1 Flash-Lite 模型，提供不到 5 秒的影像生成速度。這款輕量級視覺工具專為高速工作流程最佳化，提供 1K 解析度，以及可大規模使用的原生圖像轉圖像編輯功能。
API 呼叫範例
curl --request POST "https://gptproto.com/api/v3/google/gemini-3.1-flash-lite-image/text-to-image" \
--header "Authorization: Bearer $GPTPROTO_API_KEY" \
--header "Content-Type: application/json" \
--data '{
"prompt": "A tiny origami fox sailing a teacup across a moonlit puddle",
"aspect_ratio": "16:9",
"size": "1k",
"output_formart": "png",
"enable_base64_output": false,
"enable_sync_mode": false
}'Nano Banana 2 Lite (Gemini 3.1 Flash-Lite Image) 定價
先從單次樣本成本開始，再選擇測試預算。GPTProto 費率比標價低 40%。
你的用量
$0.0201 / 張
相關模型
所有模型
| 模型 | 解析度 | 輸入 → 輸出 | ||
|---|---|---|---|---|
| $0.02 / 每次 | 1K | 輸入: 文字輸入: 圖像 輸出: 圖像 | |
| $7.60 / $28.50 每 1M | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $7.60 / $28.50 每 1M | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| — | $0.04 / 每次 | 1K · 2K · more | 輸入: 文字輸入: 圖像 輸出: 圖像 | |
| $0.04 / 每次 | 1K · 2K | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.04 / 每次 | 1K · 2K · 4K | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $6.40 / $24.00 每 1M | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.04 / 每次 | 1K · 2K · 4K | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.03 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.03 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.01 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.02 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| — | $0.14 / 每次 | — | 輸入: 文字 輸出: 圖像 | |
| $0.08 / 每次 | 1K · 2K · 4K | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| — | $0.01 / 每次 | — | 輸入: 圖像 輸出: 圖像 | |
| — | $0.02 / 每次 | — | 輸入: 圖像 輸出: 圖像 | |
| $0.03 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.02 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.03 / 每次 | — | 輸入: 文字 輸出: 圖像 | ||
| $0.06 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.03 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 | ||
| $0.05 / 每次 | — | 輸入: 圖像 輸出: 圖像 | ||
| $0.05 / 每次 | — | 輸入: 圖像 輸出: 圖像 | ||
| $0.05 / 每次 | — | 輸入: 文字 輸出: 圖像 | ||
| $0.06 / 每次 | — | 輸入: 文字輸入: 圖像 輸出: 圖像 |
Core Nano Banana Lite API Features
探索 Nano Banana Lite API 為何是大規模視覺任務的首選。
原生視覺編輯
使用自然語言修改圖像。立即新增、移除或變更視覺內容中的元素。非常適合以輕量成本進行創意版本 A/B 測試，以及反覆調整設計。
高吞吐量視覺 OCR
針對大量文件批次掃描進行最佳化。以高準確度從螢幕截圖中擷取表格資料並識別 UI 元件，同時將處理時間降至最低，滿足企業工作流程需求。
高成本效益的擴展
輸入 token 每 1M 僅需 $2.00，每張圖像 $0.034，是目前最具經濟效益的視覺 API。在不犧牲多模態效能的情況下，最大化高流量專案的投資報酬率。
What Is the Nano Banana 2 Lite API?
Nano Banana 2 Lite is Google’s efficiency-focused image model, officially named Gemini 3.1 Flash-Lite Image. It became generally available in June 2026 under the stable ID gemini-3.1-flash-lite-image. Text and image inputs support both generation and natural-language editing.
Google reports approximately four seconds for a 1K text-to-image result. This makes the model a practical default for concepts, previews, thumbnails, social variants, and product-scene drafts before selected assets move to a higher-resolution model.
Is Nano Banana 2 Lite the Same as Gemini 3.1 Flash-Lite Image?
Yes. Nano Banana 2 Lite is the product name; Gemini 3.1 Flash-Lite Image is the official model name. Do not confuse it with the text-only Gemini 3.1 Flash-Lite model or the full Nano Banana 2 model (gemini-3.1-flash-image).
Nano Banana 2 Lite Specifications
| Specification | Value |
|---|---|
| Provider | |
| Official model name | Gemini 3.1 Flash-Lite Image |
| Stable model ID | gemini-3.1-flash-lite-image |
| Status | Generally available; latest model update June 2026 |
| Input | Text and images |
| Output | Images and text |
| Main tasks | Text-to-image generation and image editing |
| Output resolution | 1K only |
| Documented aspect ratios | 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 |
| Google Search grounding | Not supported |
| Reference images | Up to 14 object references in Google’s model; not optimized for reference-heavy or long multi-turn editing workflows |
| Provenance | Generated images include an invisible SynthID watermark |
Best Nano Banana 2 Lite Use Cases
Interactive features: Return a usable 1K draft inside creative apps, learning tools, games, or generative interfaces.
Creative testing: Produce ad concepts, thumbnails, product backgrounds, or social variants at a low cost per rejected draft.
Targeted edits: Change a background, color, object, or style with an input image and a concise instruction.
Pre-production: Explore storyboards or campaigns, then render approved directions with a higher-resolution model.
Nano Banana 2 Lite vs Nano Banana 2, Seedream 5.0 Pro, and GPT Image 1.5
| Model | Output options | Workflow strength | Choose it when… |
|---|---|---|---|
| Nano Banana 2 Lite | 1K only | Fast drafts, affordable volume, lightweight edits | You need many usable images quickly and do not require 2K or 4K output |
| Nano Banana 2 | 0.5K, 1K, 2K, and 4K | General-purpose generation, search-grounded imagery, multi-reference work | You need higher resolution, current web context, or stronger character and reference consistency |
| Seedream 5.0 Pro | 1K, 1.5K, and 2K | Dense layouts, multilingual text, precise element placement, multi-reference editing | You are creating posters, product graphics, technical visuals, or layout-sensitive assets |
| GPT Image 1.5 | 1024×1024, 1024×1536, and 1536×1024; low, medium, and high quality | Detailed instruction following and controllable generation or editing | Your workflow needs quality tiers, portrait/landscape size control, or compatibility with OpenAI image tooling |
Nano Banana 2 Lite vs Nano Banana 2
Choose Lite when speed and cost matter more than maximum resolution. Choose Nano Banana 2 for 2K or 4K delivery, Search grounding, and reference-heavy work.
Nano Banana 2 Lite vs Seedream 5.0 Pro
Use Lite for fast 1K ideation and bulk variants. Choose Seedream 5.0 Pro for denser compositions, multilingual copy, exact placement, or output above 1K.
Nano Banana 2 Lite vs GPT Image 1.5
Lite offers one inexpensive 1K tier. GPT Image 1.5 provides three output sizes and three quality levels, making it a better fit when those controls or an existing OpenAI workflow matter more.
What to Know Before Using Nano Banana 2 Lite in Production
The Lite label represents a real trade-off: output is limited to 1K, and Search grounding is unavailable. Although Google documents up to 14 object references, it says this model is not optimized for reference-heavy or long sequential editing workflows.
Review small typography, long text blocks, character identity, and left-versus-right placement. Google lists these as remaining limitations, so keep visible copy short and verify every final image.
Nano Banana 2 Lite API Prompt Patterns
Fast product variant
Create a 1K studio product photo of [product] on [surface], with [lighting], [camera angle], and [background color]. Keep the product shape, label, and logo unchanged. No extra objects.
Targeted image edit
Using the provided image, change only [specific element] to [new description]. Preserve the subject, camera angle, lighting, composition, and all other objects exactly as they are.
Short-text social graphic
Create a [aspect ratio] social graphic for [topic]. Place the exact headline “[short headline]” in large, high-contrast type. Use [palette], one clear focal subject, generous spacing, and no additional text.
For dense posters, long paragraphs, repeated characters, or multi-round precision edits, route the approved concept to a model with more resolution or stronger reference handling.
Why Access Nano Banana 2 Lite on GPTProto?
GPTProto places 200+ text, image, video, and audio models under one key and balance. Compare the same brief across Nano Banana 2, Seedream 5.0 Pro, or GPT Image 1.5 without maintaining separate provider credit pools.
Use the request format in the fixed API Usage section. The current GPTProto rate is $0.02016 per 1K generation; check the live panel before estimating a production budget.
Nano Banana Lite API 常見問題
關於 Nano Banana Lite API、Gemini 3.1 Flash-Lite 功能與整合需求的答案。
Who provides Nano Banana 2 Lite?
Google provides the model. Nano Banana 2 Lite is the product name for Gemini 3.1 Flash-Lite Image; GPTProto provides third-party access and consolidated billing.
How much is the Nano Banana 2 Lite API price on GPTProto?
The current rate is $0.02016 per 1K generation, about 40% below Google’s $0.034 launch benchmark. A $100 balance covers approximately 4,960 generations before any bonus. Confirm the live panel because rates can change.
How do I get Nano Banana 2 Lite API access?
Create a GPTProto key, then use the request shown in the fixed API Usage section. The same account balance works with other supported models.
Can Nano Banana 2 Lite edit an existing image?
Yes. It accepts image input and natural-language edit instructions. State what should change and what must remain unchanged.
What resolution and aspect ratios does Nano Banana 2 Lite support?
It outputs 1K images only. Google documents ten ratios: 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, and 21:9.
Does Nano Banana 2 Lite support Google Search grounding?
No. Choose the full Nano Banana 2 model if your image needs current web or image-search context.
How many reference images can Nano Banana 2 Lite use?
Google documents up to 14 object references, but says Lite is not optimized for reference-heavy or long multi-turn editing. Check the selected GPTProto task’s input limits.
Does Nano Banana 2 Lite add a visible watermark?
Google states that generated images include invisible SynthID. This is different from a visible logo or corner overlay.
相關文章
與本模型相關的指南、對比與更新。
所有文章
6 Cheapest AI Image Generators in 2026: Real Cost per Image
Compare 6 cheapest AI image generators in 2026, from about $0.0035 per image. See batch costs, hidden fees, and the best API for startups.
5 Best Midjourney API Alternatives in 2026: Real Model APIs, Not Discord Wrappers
See how five Midjourney API alternatives compare on image quality, editing, text rendering, pricing, and the production tasks each handles best.
How to Make an AI Story Video for Kids with GPT Image 2 and Seedance 2.5
Learn how to make an AI story video for kids with GPT Image 2 and Seedance 2.5, including the full prompt, captions, sound, editing, and a real test.
Which AI Image Model Is Best for E-Commerce in 2026? 6 Models Tested on Product Photos
See which AI image model is best for e-commerce after 3 real product tests. Compare product fidelity, ad text, first-pass results and cost.