Class: Google::Apis::AiplatformV1::CloudAiLargeModelsVisionGenerateVideoExperiments

Inherits:
Object
  • Object
show all
Includes:
Core::Hashable, Core::JsonObjectSupport
Defined in:
lib/google/apis/aiplatform_v1/classes.rb,
lib/google/apis/aiplatform_v1/representations.rb,
lib/google/apis/aiplatform_v1/representations.rb

Overview

Experimental parameters for video generation.

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(**args) ⇒ CloudAiLargeModelsVisionGenerateVideoExperiments

Returns a new instance of CloudAiLargeModelsVisionGenerateVideoExperiments.



125
126
127
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 125

def initialize(**args)
   update!(**args)
end

Instance Attribute Details

#anchor_last_frame ⇒ Boolean Also known as: anchor_last_frame?

Optional. If true, anchors the last frame in video generation by generating a custom border mask. Corresponds to the JSON property anchorLastFrame

Returns:

  • (Boolean) 


33
34
35
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 33

def anchor_last_frame
  @anchor_last_frame
end

#cfg_scale ⇒ Float

CFG scale for video-transform, perf-generation, a2v, video-textures models. Corresponds to the JSON property cfgScale

Returns:

  • (Float) 


39
40
41
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 39

def cfg_scale
  @cfg_scale
end

#codec ⇒ String

Optional. Video codec to use for output. Corresponds to the JSON property codec

Returns:

  • (String) 


44
45
46
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 44

def codec
  @codec
end

#conditioning_frames ⇒ Array<Google::Apis::AiplatformV1::CloudAiLargeModelsVisionGenerateVideoExperimentsConditioningFrame>

Conditioning frames for veo experimental models ONLY, not to be confused with keyframes (ID:31) in GenerateVideoRequest. Corresponds to the JSON property conditioningFrames

Returns:



50
51
52
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 50

def conditioning_frames
  @conditioning_frames
end

#custom_parameters ⇒ Hash<String,Object>

Generic key-value pairs for experimental parameters. This allows adding new parameters without changing the dataplane binary. Corresponds to the JSON property customParameters

Returns:

  • (Hash<String,Object>) 


56
57
58
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 56

def custom_parameters
  @custom_parameters
end

#human_pose ⇒ Google::Apis::AiplatformV1::CloudAiLargeModelsVisionHumanPose

Human pose parameters for Pose Control Corresponds to the JSON property humanPose

Returns:



61
62
63
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 61

def human_pose
  @human_pose
end

#model_name ⇒ String

Model names, as defined in: xyz Corresponds to the JSON property modelName

Returns:

  • (String) 


66
67
68
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 66

def model_name
  @model_name
end

#num_diffusion_steps ⇒ Fixnum

Number of diffusion steps Corresponds to the JSON property numDiffusionSteps

Returns:

  • (Fixnum) 


71
72
73
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 71

def num_diffusion_steps
  @num_diffusion_steps
end

#original_request_json ⇒ String

The original REST API request JSON sent by the user, in the instances/ parameters format. Preserved for saving alongside output artifacts so users can reproduce their requests. This field is populated by the API handler and is not user-settable. Corresponds to the JSON property originalRequestJson

Returns:

  • (String) 


79
80
81
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 79

def original_request_json
  @original_request_json
end

#prompt_inputs ⇒ Google::Apis::AiplatformV1::CloudAiLargeModelsVisionPromptInputs

"Direct" prompting for Experimental Video Generation. These will be sent directly to the LDM without being rewritten. Corresponds to the JSON property promptInputs

Returns:



85
86
87
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 85

def prompt_inputs
  @prompt_inputs
end

#request_origin_tag ⇒ String

Optional tag for tracking the source of this request. Allowed values: "colab", "comfyui", "curl", "flowresearch", "vertexstudio". Unrecognized tags are recorded as "unknown" in metrics. Tags do not affect video generation behavior. Up to 16 characters, ASCII alphanumeric, hyphens, and underscores only. Corresponds to the JSON property requestOriginTag

Returns:

  • (String) 


93
94
95
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 93

def request_origin_tag
  @request_origin_tag
end

#scheduling_config ⇒ Google::Apis::AiplatformV1::CloudAiLargeModelsVisionGenerateVideoExperimentsVeSchedulingConfig

Scheduling configuration for Veo Experimental. Corresponds to the JSON property schedulingConfig

Returns:



98
99
100
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 98

def scheduling_config
  @scheduling_config
end

#seamless ⇒ Google::Apis::AiplatformV1::CloudAiLargeModelsVisionSeamless

Parameters for seamless (edge-continuous) video generation, enabling temporal looping and/or spatial tessellation. Corresponds to the JSON property seamless

Returns:



104
105
106
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 104

def seamless
  @seamless
end

#truncate_input_video ⇒ Boolean Also known as: truncate_input_video?

If true (default), truncate input videos that exceed the model's maximum frame count by applying a frame_selection_config to video_file inputs. Set to false to preserve the existing fail-fast behavior. Corresponds to the JSON property truncateInputVideo

Returns:

  • (Boolean) 


111
112
113
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 111

def truncate_input_video
  @truncate_input_video
end

#video_transform_mask_gcs_uri ⇒ String

GCS URI of the grayscale video mask for Differential Diffusion. Maps to sdedit_video_tmax_scale_map Corresponds to the JSON property videoTransformMaskGcsUri

Returns:

  • (String) 


118
119
120
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 118

def video_transform_mask_gcs_uri
  @video_transform_mask_gcs_uri
end

#video_transform_strength ⇒ Float

SDEdit: Scalar noise level (0.0 to 1.0) Maps to sdedit_tmax Corresponds to the JSON property videoTransformStrength

Returns:

  • (Float) 


123
124
125
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 123

def video_transform_strength
  @video_transform_strength
end

Instance Method Details

#update!(**args) ⇒ Object

Update properties of this object



130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
 
# File 'lib/google/apis/aiplatform_v1/classes.rb', line 130

def update!(**args)
  @anchor_last_frame = args[:anchor_last_frame] if args.key?(:anchor_last_frame)
  @cfg_scale = args[:cfg_scale] if args.key?(:cfg_scale)
  @codec = args[:codec] if args.key?(:codec)
  @conditioning_frames = args[:conditioning_frames] if args.key?(:conditioning_frames)
  @custom_parameters = args[:custom_parameters] if args.key?(:custom_parameters)
  @human_pose = args[:human_pose] if args.key?(:human_pose)
  @model_name = args[:model_name] if args.key?(:model_name)
  @num_diffusion_steps = args[:num_diffusion_steps] if args.key?(:num_diffusion_steps)
  @original_request_json = args[:original_request_json] if args.key?(:original_request_json)
  @prompt_inputs = args[:prompt_inputs] if args.key?(:prompt_inputs)
  @request_origin_tag = args[:request_origin_tag] if args.key?(:request_origin_tag)
  @scheduling_config = args[:scheduling_config] if args.key?(:scheduling_config)
  @seamless = args[:seamless] if args.key?(:seamless)
  @truncate_input_video = args[:truncate_input_video] if args.key?(:truncate_input_video)
  @video_transform_mask_gcs_uri = args[:video_transform_mask_gcs_uri] if args.key?(:video_transform_mask_gcs_uri)
  @video_transform_strength = args[:video_transform_strength] if args.key?(:video_transform_strength)
end