POST /api/generate/v2, including via an API token. Some models are surfaced only on certain MeiGen surfaces (for example, not inside the mobile apps), but that never restricts API access.
Query Parameters
string
Filter by output type:
image or video. Omit to get both.string
default:"true"
Pass
false to include retired models as well. Retired models cannot be used for generation — this is only useful for resolving historical modelId values.Response
{ success: true, models: [...] }. Each model object contains:
string
The unique model identifier. Use this as
modelId when generating.string
Display name.
string
Model provider family (e.g.
OpenAI, Google, ByteDance, Midjourney). Display only — not a request parameter.number
A nominal fallback value. For video models, do not quote it to end users — derive the real cost from
extra_config.pricingPerSec × duration, or from extra_config.pricingPerCall[tier][duration] for Veo 3.1. Always send duration explicitly for Veo 3.1: if it is omitted, billing falls back to this nominal value. The authoritative charge is creditsUsed in the generate response.string[]
Aspect ratios accepted by
aspectRatio. Seedance additionally accepts "adaptive".number
Maximum number of reference images accepted (
0 means the model does not accept references).string
"image" or "video".object
Model-specific configuration. Keys you may rely on:
The pricing-related keys above are part of the contract. Other keys may appear and are informational.
Example
Example shows two representative models. The live endpoint returns all active models. Responses are cached for 1 hour (plus stale-while-revalidate), so newly added or retired models can take up to an hour to appear.