Generate Asset
Start a new image, video, or audio generation using your chosen model and inputs.
Authorizations
Body
- GenerateVideoRequest
- GenerateTextToSpeechRequest
- GenerateTextToSoundRequest
- GenerateImageRequest
- GenerateImageUpscaleRequest
- GenerateVideoUpscaleRequest
- GenerateIsolatedAudioRequest
- GenerateSpeechToSpeechRequest
- GenerateVoiceCloneRequest
- GenerateVideoToVideoRequest
- GenerateMotionControlRequest
Inputs for generating the video.
Optional agent thread ID to associate this generation with.
Optional pre-reserved generation ID. If provided, this ID will be used instead of generating a new one. For batch operations (batch_size > 1), use generation_ids instead.
Optional list of pre-reserved generation IDs for batch operations. Length must match batch_size. Mutually exclusive with generation_id.
Optional pre-reserved asset ID. Used as the produced media+asset resource_id so the client knows the asset's identity at request time. For batch operations (batch_size > 1), use reserved_asset_ids instead. Named distinctly from the response's asset_id (the produced asset) so the two don't collide across the request/response inheritance chain — mirrors generation_id (request) vs id (response).
Optional list of pre-reserved asset IDs for batch operations. Length must match batch_size, parallel to generation_ids. Mutually exclusive with reserved_asset_id.
"video"Optional human-friendly name for the generated asset.
255Deprecated. Use model_slug to select the model.
Slug of the model to use for the generation. Alternative to ai_model_id.
The id of the Image asset to use as the start keyframe.
The URL of the image to use as the start keyframe.
1 - 2083The id of the Image asset to use as the end keyframe.
The URL of the image to use as the end keyframe.
1 - 2083The id of the Audio asset to use, or a list of ids for multi-speaker generation.
Optional TTS parameters for server-side audio generation. If provided (and audio_id is not), audio will be generated from these params before video generation.
Audio start offset in milliseconds. Negative values prepend silence (e.g., -1000 adds 1s silence before audio). Positive values crop from the beginning of the source audio (e.g., 2000 skips the first 2s). Use with generated_video_inputs.duration_ms to control total output length.
The id(s) of the audio(s) to reference in the generation.
The id(s) of the image(s) to reference in the generation.
The id(s) of the video(s) to reference in the generation.
The id of the Video asset to use as input. For V2V (motion control) models this is the driving video; for Google Veo models this triggers video extension.
Number of video variations to generate (1-8). When > 1, batch_results will contain all generation results.
1 <= x <= 8Response
Successful Response
- GenerateVideoResponse
- GenerateTextToSpeechResponse
- GenerateTextToSoundResponse
- GenerateImageResponse
- GenerateImageUpscaleResponse
- GenerateVideoUpscaleResponse
- GenerateIsolatedAudioResponse
- GenerateSpeechToSpeechResponse
- GenerateVoiceCloneResponse
- GenerateVideoToVideoResponse
- GenerateMotionControlResponse
Inputs for generating the video.
The id of the generation created.
The id of the video asset resulting from the generation.
Date the generation was submitted.
Status of the generation
complete, error, processing, queued, finalizing Current progress to completion. Between 0-1
Optional agent thread ID to associate this generation with.
Optional pre-reserved generation ID. If provided, this ID will be used instead of generating a new one. For batch operations (batch_size > 1), use generation_ids instead.
Optional list of pre-reserved generation IDs for batch operations. Length must match batch_size. Mutually exclusive with generation_id.
Optional pre-reserved asset ID. Used as the produced media+asset resource_id so the client knows the asset's identity at request time. For batch operations (batch_size > 1), use reserved_asset_ids instead. Named distinctly from the response's asset_id (the produced asset) so the two don't collide across the request/response inheritance chain — mirrors generation_id (request) vs id (response).
Optional list of pre-reserved asset IDs for batch operations. Length must match batch_size, parallel to generation_ids. Mutually exclusive with reserved_asset_id.
"video"Optional human-friendly name for the generated asset.
255Deprecated. Use model_slug to select the model.
Slug of the model to use for the generation. Alternative to ai_model_id.
The id of the Image asset to use as the start keyframe.
The URL of the image to use as the start keyframe.
1 - 2083The id of the Image asset to use as the end keyframe.
The URL of the image to use as the end keyframe.
1 - 2083The id of the Audio asset to use, or a list of ids for multi-speaker generation.
Optional TTS parameters for server-side audio generation. If provided (and audio_id is not), audio will be generated from these params before video generation.
Audio start offset in milliseconds. Negative values prepend silence (e.g., -1000 adds 1s silence before audio). Positive values crop from the beginning of the source audio (e.g., 2000 skips the first 2s). Use with generated_video_inputs.duration_ms to control total output length.
The id(s) of the audio(s) to reference in the generation.
The id(s) of the image(s) to reference in the generation.
The id(s) of the video(s) to reference in the generation.
The id of the Video asset to use as input. For V2V (motion control) models this is the driving video; for Google Veo models this triggers video extension.
Number of video variations to generate (1-8). When > 1, batch_results will contain all generation results.
1 <= x <= 8Estimated time until completion in seconds. May be None if no historical data available.
Unique identifier linking all generations in a batch.
All generation results in the batch. Always populated (even for batch_size=1). The main response fields (id, asset_id, etc.) reflect the first successful generation.