Quick Answer
- Use POST /v1/images/generations with the Nano Banana 2 model id documented by ToAPIs.
- Supports prompt generation and reference_images for ad, e-commerce, social, and batch creative assets.
- The workflow is async: submit the task, receive a task id, then poll the image task status endpoint.
- Official comparison pricing is roughly $0.067 / $0.101 / $0.151 per image for 1K / 2K / 4K; ToAPIs currently shows $0.030 / $0.040 / $0.060 per image on the landing comparison, with final billing based on account pricing.
- At the ToAPIs 1K listed rate of $0.030 per image, a $10 top-up is roughly 333 additional 1K images; at the 4K listed rate of $0.060 per image, it is roughly 166 4K images.
Key Parameters
- model | string | required | nano_banana_2 | - | Use the Nano Banana 2 model id shown in ToAPIs docs or pricing.
- prompt | string | required | - | - | Primary image generation prompt.
- size | string | optional | 1:1 | 1:1 | 16:9 | 9:16 | 4:3 | 3:4 | Output aspect ratio; available values follow docs and model capability.
- resolution | string | optional | 1K | 1K | 2K | 4K | Output resolution tier, affecting per-image price and generation cost.
- reference_images | string[] | optional | - | URL array | Reference image URL array for guided generation and style/subject control.
- response_format | string | optional | url | url | b64_json | Response format; URL is recommended for most workflows.
Common Errors
- 400 invalid_request_error: trigger=Missing model/prompt or invalid resolution, aspect ratio, or reference image fields.; fix=Validate model, prompt, size, resolution, and reference_images against ToAPIs docs.; retry=Retry only after correcting the payload.
- 401 authentication_error: trigger=Missing Authorization header or invalid API key.; fix=Verify bearer token format and key scope.; retry=Retry after fixing authentication.
- 429 rate_limit_exceeded: trigger=Request rate, concurrency, or current credits hit limits.; fix=Apply exponential backoff, reduce concurrency, and check account credits.; retry=Use 1s/2s/4s backoff with jitter; queue requests if it persists.
