VideoCapabilitySection

Declared video generation contract from the model card.

Mirrors the card's video section so a client can build a valid POST /v1/videos request without a copy of the card: which modes it serves, the clip length range, the canvas rules, whether it renders a synchronized audio track, and the named adapters with their trained step counts.

modesstring[]required

Generation modes the card serves.

Possible values: [t2va, fl2va, ref2va]

min_secondsMin Seconds (integer)required

Shortest clip length accepted.

max_secondsMax Seconds (integer)required

Longest clip length accepted.

fpsFps (integer)required

Output frame rate.

frame_grid_multipleFrame Grid Multiple (integer)required

Frame counts snap to this multiple plus the offset.

frame_grid_offsetFrame Grid Offset (integer)required

Frame grid offset.

canvas_multipleCanvas Multiple (integer)required

Canvas dimensions must be multiples of this.

default_short_edge object

Trained short edge when size is omitted.

anyOf

• integer
• null

integer

max_pixels object

Pixel budget per frame.

anyOf

• integer
• null

integer

aspect_ratiosstring[]

Advisory aspect ratios the card lists.

audio_outputAudio Output (boolean)required

Whether renders carry a synchronized audio track.

audio_sample_rate object

Audio sample rate of rendered clips.

anyOf

• integer
• null

integer

audio_channels object

Audio channels of rendered clips.

anyOf

• integer
• null

integer

default_stepsDefault Steps (integer)required

Sampling steps when no adapter is selected.

reference_limits object

Reference attachment limits for reference-to-video cards.

anyOf

• VideoReferenceLimitsSection
• null

max_imagesMax Images (integer)

Reference images accepted per request.

Default value: 0

max_videosMax Videos (integer)

Reference clips accepted per request.

Default value: 0

max_audio_clipsMax Audio Clips (integer)

Reference audio clips accepted per request.

Default value: 0

max_files object

Total attachments accepted, when bounded.

anyOf

• integer
• null

integer

clip_min_seconds object

Shortest reference clip accepted.

anyOf

• integer
• null

integer

clip_max_seconds object

Longest reference clip accepted.

anyOf

• integer
• null

integer

total_clip_seconds object

Combined reference clip length accepted.

anyOf

• integer
• null

integer

adapters object[]

Named adapters (LoRAs) selectable through the job lora field.

Array [

nameName (string)required

Adapter name accepted by the video job lora field.

modesstring[]

Generation modes the adapter is trained for; empty means every mode.

Possible values: [t2va, fl2va, ref2va]

steps object

Sampling steps the adapter was trained for.

anyOf

• integer
• null

integer

strength object

Default adapter strength.

anyOf

• number
• null

number

video_shift object

Video sigma shift a render with this adapter uses when the request sets none; null keeps the card's.

anyOf

• number
• null

number

audio_shift object

Audio sigma shift a render with this adapter uses when the request sets none; null keeps the card's.

anyOf

• number
• null

number

]

styles object[]

Style embeddings selectable through the job styles field.

Array [

nameName (string)required

Style name accepted by the video job styles field.

modesstring[]

Generation modes the style is declared for; empty means every mode.

Possible values: [t2va, fl2va, ref2va]

]

samplersstring[]

Every sampler the job sampler field accepts: the engine's list, not a recommendation.

default_samplerDefault Sampler (string)

Sampler a render uses when the request names none.

Default value: res_multistep

schedulersstring[]

Every sigma schedule the job scheduler field accepts.

default_schedulerDefault Scheduler (string)

Sigma schedule a render uses when the request names none.

Default value: simple

video_shift object

The card's trained video sigma shift, used when neither the request nor its adapter sets one; null keeps the model's built-in value.

anyOf

• number
• null

number

audio_shift object

The card's trained audio sigma shift, used when neither the request nor its adapter sets one; null keeps the model's built-in value.

anyOf

• number
• null

number

shift_boundsnumber[]

Inclusive range the job video_shift and audio_shift fields accept.

reference_fidelitiesstring[]

Reference sizings the job reference_fidelity field accepts; empty when the card serves no ref2va.

default_reference_fidelity object

Reference sizing a ref2va render uses when the request names none.

anyOf

• string
• null

string

codecsstring[]

Every codec the job codec field accepts.

default_codecDefault Codec (string)

Codec a render is written with by default.

Default value: h264

guides object[]

Guides the card derives from a control clip, selectable through the job control_kind field; empty when it has no ControlNet.

Array [

kindKind (string)required

Value of the video job control_kind field: pose, depth, or edges.

Possible values: [pose, depth, edges]

modesstring[]required

Generation modes the card derives this guide for.

Possible values: [t2va, fl2va, ref2va]

weights object[]

Preprocessor weights the guide loads; empty when it needs none.

Array [

nameName (string)required

The card companion's name.

repository object

Hosting repository when it is not the card's own.

anyOf

• string
• null

string

license object

License identifier the hosting repository declares.

anyOf

• string
• null

string

]

]

default_guide object

The guide a control clip gives when control_kind is omitted.

anyOf

• string
• null

string

Possible values: [pose, depth, edges]

control_strength_boundsnumber[]

Inclusive range the job control_strength field accepts.

default_control_strength object

ControlNet strength a render uses when control_strength is omitted: the card's ControlNet's own, else 1; null when the card has no ControlNet.

anyOf

• number
• null

number

default_control_windownumber[]

Schedule fractions control_start and control_end take when omitted: the ControlNet steers across the whole schedule.

VideoCapabilitySection

{
  "modes": [
    "t2va"
  ],
  "min_seconds": 0,
  "max_seconds": 0,
  "fps": 0,
  "frame_grid_multiple": 0,
  "frame_grid_offset": 0,
  "canvas_multiple": 0,
  "default_short_edge": 0,
  "max_pixels": 0,
  "aspect_ratios": [
    "string"
  ],
  "audio_output": true,
  "audio_sample_rate": 0,
  "audio_channels": 0,
  "default_steps": 0,
  "reference_limits": {
    "max_images": 0,
    "max_videos": 0,
    "max_audio_clips": 0,
    "max_files": 0,
    "clip_min_seconds": 0,
    "clip_max_seconds": 0,
    "total_clip_seconds": 0
  },
  "adapters": [
    {
      "name": "string",
      "modes": [
        "t2va"
      ],
      "steps": 0,
      "strength": 0,
      "video_shift": 0,
      "audio_shift": 0
    }
  ],
  "styles": [
    {
      "name": "string",
      "modes": [
        "t2va"
      ]
    }
  ],
  "samplers": [
    "string"
  ],
  "default_sampler": "res_multistep",
  "schedulers": [
    "string"
  ],
  "default_scheduler": "simple",
  "video_shift": 0,
  "audio_shift": 0,
  "shift_bounds": [
    0
  ],
  "reference_fidelities": [
    "string"
  ],
  "default_reference_fidelity": "string",
  "codecs": [
    "string"
  ],
  "default_codec": "h264",
  "guides": [
    {
      "kind": "pose",
      "modes": [
        "t2va"
      ],
      "weights": [
        {
          "name": "string",
          "repository": "string",
          "license": "string"
        }
      ]
    }
  ],
  "default_guide": "pose",
  "control_strength_bounds": [
    0
  ],
  "default_control_strength": 0,
  "default_control_window": [
    0
  ]
}