Gemini gets new SynthID-based tool to detect AI-generated and edited videos

synth-id
1
0

Google has introduced a new feature in its Gemini chatbot that helps users identify whether a video has been created or altered using artificial intelligence. Announced on Thursday, the tool allows users to upload video files to Gemini and check if Google’s own AI models were involved in generating or editing the content. The feature is being rolled out globally and is available to all Gemini users.

This update comes just weeks after Google added a similar AI-detection capability for images, strengthening its efforts to improve transparency around AI-generated content. With rapid advances in AI-powered video and image creation, concerns over deepfakes and manipulated media have grown worldwide.

Google said the new capability is part of its broader push to expand content verification tools. At the core of this effort is SynthID, a digital watermarking technology introduced by the company in 2023. SynthID embeds an invisible watermark into content produced using Google’s AI tools. Over time, the technology has been upgraded to work across text, images, audio and videos.

Beyond watermarking, SynthID can also detect whether content has been modified using Google’s AI, even if it has been cropped, copied, or edited outside the company’s platforms. Initially, access to this technology was limited to researchers and select partners. However, Google has gradually expanded availability, starting with AI image verification in Gemini last November and now extending it to videos.

To use the new feature, users can upload a video to the Gemini app or website and ask a simple question such as, “Was this created or edited using AI?” Gemini then separately analyses the video and audio tracks to check for the presence of the SynthID watermark. For example, it may indicate that AI was detected in a specific portion of the audio while confirming that the visuals remain unaltered.

There are some usage limits at present. Videos uploaded for verification must be under 100MB in size and no longer than 90 seconds in duration. Larger or longer files are not currently supported.

The AI video detection tool is available across all regions and languages where Gemini is supported, including India.