WAN 2.7 Image

The WAN 2.7 Image Edit Pro model supports a maximum resolution of 2K and delivers high-fidelity results.

Nano Banana 2.1

Nano Banana 2.1 (gemini-nano-banana-2.1) is Google's next-generation Nano Banana 2 image model, built on Gemini 3.6 Flash. It improves realism, visual design, text rendering and multi-turn character consistency, generating and editing images at 1K, 2K and 4K with up to 14 reference images.

Seedream 5.0 Flash

Seedream 5.0 Flash Edit is the reference-guided editing mode of Seedream 5.0 Flash. It accepts 1-10 source images for multi-image composition, interactive edits driven by marks or / coordinates, and transparent-background editing of a single PNG input.

Seedream 5.0 Pro

Seedream 5.0 Pro Layer decomposes one design image into a base image and up to 16 transparent PNG layers, ideal for re-editing posters and covers. Target text or subjects with coordinates in the prompt; each output carries its z_index, bounding box, name and description. Billed once per request.

GPT Image 2.5

GPT Image 2.5 Flare Lite is APIPod's cost-oriented routing variant for GPT Image 2.5 Flare rather than a separate OpenAI model. It uses the shared generation and editing contract with lower-cost configured channels, suited to high-volume batch generation on a budget.

GPT Image 2

GPT Image 2 1K low-cost image generation model is a special-priced gpt-image-2 model for 1K resolution. It features an ultra-low price with slightly compromised stability. It is suitable for quick preview and testing and not recommended for production environments.

Nano Banana

Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) is the ultra-fast, low-cost tier of Google's Nano Banana 2 image generation family. It generates and edits images at 1K resolution with the lowest latency in the family, making it ideal for high-concurrency real-time scenarios such as stickers, recoloring and background changes.

Seedream V4.5

The Seedream 4.5 image editing model has achieved an all-round improvement based on the overall scaling of the model: it can accurately identify and stably lock the main subject in multi-image combinations, and maximize the retention of the original image's features and detailed textures.