# `LlmCall.Media.Capabilities`
[🔗](https://github.com/gilbertwong96/llm_call/blob/main/lib/llm_call/media/capabilities.ex#L1)

The media surface a codec declares.

List-valued fields enumerate accepted values: `sizes: []` means no `size` is
accepted, and `max_n: :unsupported` means no `n` is accepted. A request knob
left `nil` means "provider default" and is always accepted.

`max_references: nil` means no reference is accepted; an integer is the number
the endpoint takes (Seedance's first/last frame pair is `2`, MiniMax's first
frame is `1`). Declarations are static per codec, because the core has no model
catalog; a model-specific limit the codec cannot know surfaces as a provider
error.

# `t`

```elixir
@type t() :: %LlmCall.Media.Capabilities{
  aspect_ratios: [String.t()],
  async: boolean(),
  durations_s: [pos_integer()],
  max_n: pos_integer() | :unsupported,
  max_references: non_neg_integer() | nil,
  modalities: [LlmCall.Media.Request.modality()],
  negative_prompt: boolean(),
  seed: boolean(),
  sizes: [String.t()]
}
```

---

*Consult [api-reference.md](api-reference.md) for complete listing*
