The Stability AI model card uses Stable Audio Open 1.0 for text to generate short audio, describing English text, up to about 47 seconds, and stereo audio output conditions. The sound quality has not been measured in this article. For commercial use, please check the applicable agreement separately.
Prompt to describe the sound first rather than the picture
Simulating the environmental sound of the product video can clarify the sound source, distance, duration and background, avoiding the need to write only a visual story. After generation, the editor will listen to whether it contains unexpected vocals, mutations or unsuitable rhythms before deciding whether to use it.
Check the beginning and end when splicing
There may be obvious seams in short clip loops, and the connection, volume, and other sound masking should be checked on the actual video timeline. Just because it's suitable for audition alone, doesn't mean it doesn't need to be processed when put into the finished film.
Archive the model with the source material
Save prompts, generation date, model version and manual processing records, and indicate the properties of the generated materials. The terms of use may vary depending on the version and purpose, and the community license cannot be generalized to free commercial use in all situations. This article is a workflow design and does not promise availability or production costs.
Extended reading: AI model.