Synthetic media are audiovisual contents—images, audio, video, or text—artificially generated or modified using artificial intelligence algorithms, especially generative models such as GANs, diffusion models, or transformer-type architectures. Unlike content captured by a camera or microphone, here the material originates from a computational process that learns patterns from large volumes of data.
Their relevance is twofold. On one hand, they open up remarkable creative and productivity possibilities: automatic dubbing, file restoration, visual prototyping, or the generation of accessible synthetic voices. On the other hand, they pose risks of misinformation and impersonation, especially in the form of deepfakes.
Some common uses include:
- Video: digital avatars and generated faces.
- Audio: voice cloning and synthesis.
- Image: non-existent illustrations and photographs.
A practical nuance: not all synthetic content is deceptive. The key lies in transparency, which is why watermarking and labeling initiatives that identify the artificial origin of the material are growing.