Capability Map

This document summarizes the currently supported model capabilities, API types, and applicable use cases, helping you quickly select the right endpoint.

bach-1.0-preview

bach-1.0-preview is the default video generation model. It supports text-to-video, image-to-video, multi-element video generation, and Montage video generation.

Capability	Endpoint	Duration	Resolution	Frame Rate
Text to Video	POST /videos/text2video	Fixed 6 seconds	720p, 1080p	24fps
Image to Video	POST /videos/image2video	1–5 seconds	720p, 1080p	24fps
Elements to Video	POST /videos/elements2video	1–6 seconds	720p, 1080p	24fps, 30fps
Montage Video	POST /videos/montage	10–45 seconds	720p, 1080p	Fixed 24fps

Video Capabilities

Text to Video

Generates a video from a text prompt. Well suited for concept visuals, creative short clips, story fragments, and rapid visual prototyping.

• Create task: POST /videos/text2video
• Query task: GET /videos/text2video/{task_id}
• Primary input: prompt
• Output: video task result and video_url

Image to Video

Generates a dynamic video from a single static image. Motion and style changes can be guided via the text prompt.

• Create task: POST /videos/image2video
• Query task: GET /videos/image2video/{task_id}
• Primary inputs: image_url, prompt
• We recommend using Batch Upload first to obtain image URLs

Elements to Video

Generates a video from multiple reference images and a text prompt. Supports both single-image references and subject references, making it ideal for scenarios that require consistency of characters, objects, or brand elements.

• Create task: POST /videos/elements2video
• Query task: GET /videos/elements2video/{task_id}
• Primary inputs: reference_inputs, prompt
• Supports up to 9 reference inputs
• Supports two reference input types: image and subject
• We recommend using Batch Upload first to obtain image URLs

Montage Video

Generates a longer video from a story description and optional image assets. Supports background music, sound effects, and narration configuration, making it well suited for brand films, narrative videos, and asset-remix scenarios.

• Create task: POST /videos/montage
• Query task: GET /videos/montage/{task_id}
• Primary inputs: prompt, image_urls
• Supports up to 20 reference images
• When no images are provided, the video can be generated from the text story alone

Subject Capabilities

Text to Subject

Generates subject images from a text description. Ideal for creating character, object, or brand-element references for use in Elements-to-Video generation.

• Create task: POST /subject/text2image
• Query task: GET /subject/text2image/{id}
• Primary inputs: name, description, style, subject_type
• Supported subject types: character, object

File Capabilities

Batch Upload

Used to upload image files in advance and obtain image URLs that can be used in the video generation endpoints. Recommended before calling Image to Video, Keyframe to Video, Elements to Video, or Montage Video.

• Submit file upload: POST /batch-upload/submit-with-files
• Query upload status: GET /batch-upload/status/{batch_id}
• Maximum size per image: 10MB
• Maximum images per batch: 300
• Supported image formats: jpg, jpeg, png