属性
系列
集合
参数
| 参数名 | 描述 | 类型 | 必填 | 枚举值 |
|---|---|---|---|---|
| prompt | Describe the scene, motion, or edit to produce. | string | 是 | - |
| video | Publicly reachable HTTP(S) URL of the source video. Its aspect ratio is retained. Source length must be 3 to 15 seconds. | string | 是 | - |
| referenceImages | Optional visual references for the edit or continuation. | string[] | 否 | - |
| resolution | Output resolution. | string | 否 | 360p, 720p, 1080p, 4k |
| seed | Optional generation seed. Omit it to leave seed selection to the model. | integer | 否 | - |
价格
输入图片($/img): 0.0020
输入视频($/sec): 0.0090
单位: $/sec
| 维度 | 价格 |
|---|---|
| resolution: 360p | 0.0230 |
| resolution: 720p | 0.0700 |
| resolution: 1080p | 0.1040 |
| resolution: 4k | 0.2090 |
相关模型
- google/gemini-omni-1.2-flash-upscale: Video resolution enhancement for preparing existing clips for larger displays and higher-resolution delivery. It preserves the source timing and aspect ratio while increasing the output resolution.
- google/gemini-omni-1.2-flash-video-extend: Video continuation for extending an action, continuing a camera move, or completing a short scene. It uses the source ending and optional image references to generate a new segment and returns it joined to the retained source footage.
- google/gemini-omni-1.2-flash-r2v: Reference-guided video generation for carrying visual appearance and motion ideas into new scenes. It combines image and video references with a text prompt for character concepts, product scenes, and mixed-reference storyboards.
- google/gemini-omni-1.2-flash-i2v: Image-to-video generation for animating product imagery, characters, and visual concepts. A starting image anchors the opening frame, and an optional ending image guides the transition while the model generates synchronized audio.
- google/gemini-omni-1.2-flash-t2v: Google’s video generation model for turning text into short scenes with synchronized sound. It suits product concepts, narrative clips, and storyboard exploration across landscape and portrait formats.
- google/gemini-omni-1.1-flash-v2v: Google’s video editor for changing the scene, mood, style, or lighting of a short clip, with synchronized sound. It keeps the source length and frame shape, and can render the result at 360p, 720p, 1080p, or 4K.
- google/gemini-omni-1.1-flash-i2v: Google’s image-to-video model for bringing a still to life with synchronized sound. It uses one starting frame in a 3 to 10 second clip at 360p, 720p, 1080p, or 4K.
- google/gemini-omni-1.1-flash-t2v: Google’s text-to-video model for short social and concept clips with synchronized speech, music, and sound effects. It makes a 3 to 10 second clip at 360p, 720p, 1080p, or 4K, in widescreen or vertical frames.
- alibaba/happyhorse-1.1-t2v: Alibaba’s text-to-video model for cinematic clips and spoken scenes, with native sound and lip sync. It makes a 3 to 15 second clip at 720P or 1080P, including ultrawide and tall frames, with clearer motion than HappyHorse 1.0.
- alibaba/wan2.7-t2v: Alibaba’s text-to-video model for multi-shot scenes with synchronized sound. It makes a 2 to 15 second clip at 720P or 1080P, and can follow a supplied audio track.
- bytedance/seedance-2.0-t2v: ByteDance’s text-to-video model for cinematic clips with native sound, including a 4K master. It makes a 4 to 15 second clip at 480p, 720p, 1080p, or 4K.
- xai/grok-imagine-video: xAI’s text-to-video model for short social clips and animated scenes described in a prompt, with generated audio. It makes 1 to 15 second clips at 480p or 720p.
- xai/grok-imagine-video-1.5-i2v: xAI’s image-to-video model for animating one still with clearer motion, physics, and native audio, including lip-synced speech. It makes 1 to 15 second clips at native 1080p, 720p, or 480p.
- bytedance/seedance-2.0-i2v: ByteDance’s image-to-video model for animating a start frame, bridging two frames, or following up to nine reference images, with native sound. It makes a 4 to 15 second clip at 480p, 720p, 1080p, or 4K.
- alibaba/happyhorse-1.1-i2v: Alibaba’s image-to-video model for animating a still into a spoken or cinematic scene, with native sound. It makes a 3 to 15 second clip at 720P or 1080P.
- alibaba/happyhorse-1.1-r2v: Alibaba’s reference-to-video model for a scene that keeps several characters consistent, with native sound. It uses 1 to 9 reference images in a 3 to 15 second clip at 720P or 1080P.
- alibaba/wan2.7-i2v: Alibaba’s image-to-video model for animating a start frame, bridging two frames, or continuing a clip, with synchronized sound. It makes a 2 to 15 second clip at 720P or 1080P.
- alibaba/wan2.7-r2v: Alibaba’s reference-to-video model for a clip guided by images or short videos, with synchronized sound. It makes a 720P or 1080P result in widescreen, square, or vertical frames.
- alibaba/wan2.7-videoedit: Alibaba’s video editor for instruction changes, restyles, and copied camera moves. It edits a clip at 720P or 1080P, for 2 to 10 seconds, and can keep the original sound.
- bytedance/seedance-2.0-v2v: ByteDance’s video editor for restyling or rebuilding a clip from reference video, images, and audio, with native sound. It uses 1 to 3 videos in a 4 to 15 second result at up to 4K.
- xai/grok-imagine-video-edit: xAI’s video editor for prompt-guided changes such as restyles, wardrobe, and props, while keeping the rest of the clip close to the source. The result keeps the source duration and shape, up to 720p.
- xai/grok-imagine-video-extend: xAI’s video extension model for continuing a clip from its last frame with new prompt-guided action. It appends 2 to 10 seconds and keeps the source shape, up to 720p.
- kling/kling-v3-omni-t2v: Kling’s storyboard text-to-video model for multi-beat scenes and consistent characters, with optional native sound and automatic shot changes. It makes a 3 to 15 second clip at 720p, 1080p, or 4K, in widescreen, square, or vertical frames.
- kling/kling-v3-t2v: Kling’s text-to-video model for cinematic scenes and longer shots, with optional native sound and automatic shot changes. It makes a 3 to 15 second clip at 720p, 1080p, or 4K, in widescreen, square, or vertical frames.
- kling/kling-v3-turbo-t2v: Kling’s fast text-to-video model for short-form drafts and high-volume clips. It makes a 3 to 15 second clip at 720p or 1080p, in widescreen, square, or vertical frames.
- kling/kling-v3-i2v: Kling’s image-to-video model for animating a still into a longer scene, with an optional end frame, native sound, and automatic shot changes. It makes a 3 to 15 second clip at 720p, 1080p, or 4K.
- kling/kling-v3-omni-i2v: Kling’s reference-to-video model for a scene that keeps subjects and style consistent, with optional native sound and automatic shot changes. It uses 1 to 7 reference images in a 3 to 15 second clip at 720p, 1080p, or 4K.
- kling/kling-v3-turbo-i2v: Kling’s fast image-to-video model for turning one still into a short clip. It makes a 3 to 15 second video at 720p or 1080p.
- kling/kling-v3-omni-video: Kling’s video-to-video model for restyling or changing a clip from a reference video, with optional stills and sound. It uses one reference video and up to four images in a 3 to 15 second clip at 720p, 1080p, or 4K.
- bytedance/seedance-2.5-t2v: ByteDance’s text-to-video model for longer stories told in one clip, with native sound. It makes a 4 to 30 second clip at 480p, 720p, or 1080p, in widescreen, square, or vertical frames.
- bytedance/seedance-2.5-v2v: ByteDance’s video model for editing, extending, or rebuilding a clip from reference video, with native sound. It uses up to 10 videos, plus optional images and audio, and can make a 4 to 30 second result at up to 1080p.
- alibaba/wan3.0-t2v: Alibaba’s text-to-video model for longer scenes with optional sound. It makes a 2 to 30 second clip at 480P, 720P, or 1080P, in widescreen, square, or vertical frames.
- alibaba/wan3.0-i2v: Alibaba’s image-to-video model for animating a start frame, bridging two frames, or following up to 10 reference images, with optional sound. It makes a 2 to 30 second clip at 480P, 720P, or 1080P.
- alibaba/wan3.0-v2v: Alibaba’s video model for rebuilding a clip from reference video, images, and audio. It uses 1 to 5 videos in a 2 to 30 second result at 480P, 720P, or 1080P.
- vidu/viduq4-r2v: Vidu’s reference-to-video model for a newly composed shot that keeps characters, style, and voices consistent across camera changes. It uses 1 to 15 reference images and up to three audio clips, in a 3 to 16 second clip from 540p through 4K, including 3:4 and 4:3 frames.
- kling/kling-4.0-t2v: Text-to-video generation for advertising concepts, storyboards, and short narrative scenes. It combines prompt-directed composition and widescreen framing to turn scene ideas into finished clips.
- kling/kling-4.0-flash-t2v: Fast text-to-video generation for product previews, advertising drafts, and short social clips. It combines prompt-directed scenes and flexible framing for frequent creative iteration.
- kling/kling-4.0-flash-i2v: Single-image animation for product showcases, character motion, and social video. It turns a starting image into a short clip, using the reference to guide the opening composition.
- kling/kling-4.0-i2v: Image-to-video generation for animating product shots, characters, and illustrated scenes. First, middle, and final keyframes guide the visual progression for planned motion and scene transitions.
- kling/kling-4.0-flash-video-edit: Video editing in the Flash series for scene restyling, object changes, and reference-guided revisions. It combines source footage with image, video, or subject references to shape the resulting scene.
- kling/kling-4.0-video-edit: Reference-guided video editing for removing objects, restyling scenes, and revising existing footage. Image, video, and subject references help guide changes while keeping the surrounding scene and recognizable subjects consistent.

