Google introduced SynthID, an innovative measure unveiled on Tuesday to combat misinformation by implementing an imperceptible, permanent watermark on images to indicate their computer-generated origin.
SynthID, integrated into images produced by Imagen, Google’s latest text-to-image generator, embeds a distinctive mark that persists through alterations such as filters or color adjustments. This technology, labeled SynthID, employs three levels of certainty—detected, not detected, and possibly detected—to identify images likely created by Imagen, enhancing accuracy against common manipulations, according to Google.
Initially available in beta to select Vertex AI customers, Google plans to expand SynthID across its platforms and potentially to third parties. The development of this tool by Google’s DeepMind unit in collaboration with Google Cloud reflects growing industry efforts to address the proliferation of realistic deepfakes and edited media, crucial in the current landscape where distinguishing real from AI-generated content is increasingly challenging.
COMMENTS