
Kunlun Wanwei's SkyProduction workbench...
Send the logline or a picture prompt first, then decide whether to lock the camera and the light, instead of regenerating the whole clip. A music video needs the song address, and can include character images and a background, with switches for lyric subtitles and lip sync. Reference images, video, and audio can be uploaded as assets and reused after they become active. A first frame can stand alone, a last frame must accompany a first frame, and a draft cannot be mixed with full reference inputs in one request. Script translation uses the text endpoint and returns the translation directly. Video and song jobs need their task id until they finish, fail, or time out. It fits short-drama teams who revise shots, music videos timed to a song, and people batching commerce clips through the API.
Kunlun Wanwei's SkyProduction workbench turns a one-line story into a short drama, and it can also generate footage and cut a finished video for short-form, social, and commerce use. You can adjust camera, lens, focal length, and aperture, and change the color, edge, and direction of the lights. The same product also covers text-to-video, a first frame or reference image, music videos with a song and lyric subtitles, songs from a prompt or lyrics, scoring a picture or video, and script translation. Video and song jobs must be checked again after they are created.
Difficulty: Intermediate
Logline to short drama
Start a short drama from one logline, and also generate footage and cut a finished video for short-form or commerce.
Camera and lighting
Adjust camera, lens, focal length, and aperture, and change light color, edge, and direction so shots in one scene match.
Song-based music video
Submit a song and a picture prompt, add character and background images, choose lyric subtitles and lip sync, then upscale if needed.
Songs and picture scores
Generate one song from a prompt or lyrics, or upload a picture or video and receive only a matching music track.
Script translation
Put the script in the chat messages for a synchronous translation that keeps scene and dialogue formatting, separate from video jobs.
A short-drama unit revises shots
After the writer hands over a logline, the director changes camera and light, then generates the next shot in the same scene.
A musician pictures a song
Someone with a finished song submits a clip, character images, and a background to get a music video with lyric subtitles.
Commerce teams batch clips
An operator creates video jobs from a reference image and a prompt, keeps the task id, and downloads the clip when it finishes.
You need a publicly reachable song address and a picture prompt. Add character images and a background when the person or the set should stay fixed. Lyric subtitles and lip sync are separate switches and both start off. You can use only part of the song. Higher resolution is an extra upscale setting on the same job, and upscale does not add frames or change playback speed.
After the video request is accepted, the job is usually queued and the response only has an id and a status. Poll with the same key until it succeeds, fails, or times out. If the create call itself times out, retry the original body with the original idempotency key. A changed body conflicts. Only the key that created the job can read it.
No. Script translation puts the source text in the chat messages and returns the translation in that call, which suits scene and dialogue formatting. The separate script field is not what gets translated, so the text must be in the messages. Video, image, and text share one idempotency namespace, but video is still polled. Translation accepts text only.
Real reviews and feedback from users