Google has launched a new site called SynthID that lets anyone verify whether an image, video, or audio clip was generated using AI. This tool, which first appeared at Google I/O last year, is now available to the public.
SynthID supports various media formats, making it adaptable for different content types. For images, it accepts formats like JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, and GIF. If you’re working with videos, you can upload MP4, MOV, and WEBM formats. For audio, the site supports WAV, MP3, OGG, FLAC, AAC, and M4A.
At the heart of SynthID is a watermarking technology that Google introduced in 2023. This tech is embedded in the company’s AI models, including Nano Banana, Veo, and Lyria, as well as other generative tools like Gemini, Flow, ProducerAI, and Vids. All these tools use SynthID to watermark any media they create, ensuring that such content can be traced back to its origin.
Several major companies also see the value in SynthID. OpenAI, NVIDIA, and Kakao have integrated support for this verification tool. OpenAI maintains its own site for checking content, providing users with an additional resource. Meanwhile, Apple is said to be adding support for SynthID soon.
Google has incorporated SynthID verification into its Gemini app and Google Chrome, emphasizing its commitment to transparency in AI-generated content. The company said people currently make 1 million requests to verify content every day.
While Google’s SynthID marks a significant step forward in identifying AI-generated content, it has its limitations. This tool specifically targets content created by Google’s own AI models and may not reliably identify outputs from other systems, such as those generated by ChatGPT. Additionally, competing companies like Microsoft and Meta have their own watermarking and verification standards. However, these tools are not infallible, often failing to identify content created by their makers’ models.
As SynthID gains traction, its reliance on Google’s watermarking technology raises questions about the effectiveness of verifying AI-generated content from other systems, highlighting the challenge of establishing universal standards in a fragmented landscape.
