From photo to reviewed asset
Before a video is submitted, the service reviews the portrait as a private image asset. This step is separate from video generation and can reject an unsuitable image. A sharp photo of one consenting adult gives the workflow a clearer input.
The local photo check validates the image format, dimensions and brightness. It is not an identity check or a guarantee that the provider will approve the portrait.
How the video is directed
The request includes a reviewed image, a fixed party reference and a scene direction. The reference guides the camera movement and performance; your prompt can add wardrobe, expression and mood details. A new fictional scene is generated instead of editing the original recording.
Available outputs are 480p, 720p and 1080p at 10 or 15 seconds, in 16:9 or 9:16. All options use the same Seedance 2.5 model and request MP4 output with generated ambience.
Progress, retries and limitations
A saved task moves through portrait review, submission and rendering. Refreshing the progress checks that saved task; it does not submit a second video. Confirmed failures restore the credits charged for the task.
AI video is probabilistic. Facial details, cup placement and hand motion can vary, and a longer clip may introduce more variation. Start with a focused prompt and judge the actual result rather than expecting an identical copy of an example.