Image and video generation

This section documents the public PopGo image and video models. All examples use PopGo's unified OpenAI-compatible API. Current availability, pricing, and enabled capabilities are defined by the live models and pricing page.

Before you start

1. Set the client Base URL to https://api.popgo.site/v1.
2. Authenticate with Authorization: Bearer YOUR_API_KEY.
3. Copy the exact public model ID from the models and pricing page and replace YOUR_MODEL.
4. Verify a minimal text-to-image or short-video request before adding references or high-cost options.

Public model catalog

Type	Models
Image	gpt-image-2, gpt-image-2 4K
Image	nano-banana-1, nano-banana-2
Image	grok-imagine-image, grok-imagine-image-quality
Video	grok-imagine-video, grok-imagine-video-1.5
Video	kling-video-v3, kling-video-v3-omni, kling-video-v3-turbo
Video	seedance-2.5-B, seedance2.0-F-1080p, seedance2.5-A, seedance2.0-B, seedance2.0-fast-F-720p, seedance2.0-F-720p, seedance2.0-E-720p, seedance2.0-fast-B, seedance2.0-A, seedance-2.0-C

Models in one row share a protocol family, not necessarily every limit. Use the model-specific page for resolution, duration, references, and billing dimensions.

Endpoint map

Purpose	Method	PopGo path
Text-to-image	POST	/v1/images/generations
Image editing	POST	/v1/images/edits
Create video task	POST	/v1/videos
Poll video task	GET	/v1/videos/{task_id}
Download video	GET	/v1/videos/{task_id}/content

Clients always use PopGo's public video paths. They do not need to know a channel's internal create or polling path.

Reference media

• Image editing normally uses multipart/form-data with an image file field.
• A video reference URL must be a public direct link that the server can fetch without login, cookies, or a sharing-page redirect.
• Base64 input must be a complete Data URL such as data:image/png;base64,...; raw Base64 without a MIME prefix is not a URL.
• Image, video, audio, frame-pair, and count support is model-specific.

Response rule

Image endpoints return an OpenAI Images-shaped data array. Video endpoints return a task ID first and a downloadable result after completion. See Video task workflow for states and progress handling.

Troubleshooting order

Check the model ID, path, authentication header, field names, reference accessibility, duration, resolution, aspect ratio, reference count, and model pricing configuration in that order.