The context of the creation of self-learning contents knows a revolution by the use of digital tools. The process to realize video was based previously on the use of analog tools. It moves nowadays towards digital ones: studios are equipped to work with digital data and digital materials under the pressure of international standards and components. On the same time, capture and storage pictures by videotapes evolve towards hybrid supports (Digital/Video), which will become entirely digital soon (computers hard-discs and networks). The current creation process which uses a kind of simulated interactivity, waits new materials and synopsis that will allow a true and complete interactivity. This interactivity needs to be used through a full digital multimedia way. The current evolution of the market digital systems seems to empower the role of the creator and the designer. They might start their creation with previous and genuine videos, by exploiting the ones which suit the best what they need, picking them up on video databases. But the problem is to be able to locate the video wanted. Many systems of automatic recognition (voice, pictures, videos...) are already functional to help them in this heavy task. But the question of using them in a relevant context, as the educational one per example, is still not resolved