Back
Google DeepMind SynthID: AI Content Watermarking
webCredibility Rating
4/5
High(4)High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: Google DeepMind
SynthID is a practical industry deployment of AI content watermarking, relevant to discussions of AI transparency, synthetic media governance, and technical approaches to reducing AI-enabled misinformation.
Metadata
Importance: 55/100tool pagehomepage
Summary
SynthID is Google DeepMind's technology for embedding imperceptible watermarks into AI-generated content to enable identification of synthetic media. It operates across multiple modalities including images, audio, video, and text without degrading output quality. The system aims to help combat misinformation and improve transparency around AI-generated content.
Key Points
- •Embeds invisible watermarks in AI-generated images, audio, video, and text without perceptible quality degradation
- •Enables identification of synthetic media to help combat misinformation and deepfakes
- •Deployed in Google products including Imagen and other generative AI platforms
- •Represents a technical approach to AI content provenance and transparency at scale
- •Watermarks are designed to be robust against common transformations like compression or cropping
Review
SynthID represents an innovative approach to content authentication in the era of generative AI, providing a method to trace and verify synthetic media. By embedding invisible watermarks that survive common transformations like cropping, compression, and filtering, Google has developed a technical solution to the growing challenge of distinguishing AI-generated from human-created content. The methodology relies on subtle modifications to generation probabilities in different media types - adjusting pixel values in images, embedding inaudible audio signals, and manipulating token probability scores in text. This approach is particularly significant for AI safety, as it offers a potential mechanism to increase transparency and accountability in AI-generated content. While promising, the technology's effectiveness will depend on widespread adoption and the ability to withstand increasingly sophisticated attempts to circumvent or remove watermarks.
Cited by 2 pages
| Page | Type | Quality |
|---|---|---|
| AI Safety Solution Cruxes | Crux | 65.0 |
| AI Disinformation | Risk | 54.0 |
Cached Content Preview
HTTP 200Fetched Apr 10, 20263 KB
SynthID — Google DeepMind Skip to main content SynthID A tool to watermark and identify content generated through AI
Try in Gemini Overview
How it works
Hands-on
What is SynthID?
Generative AI can help us all to be more creative, productive, and innovative. But it can be hard to tell the difference between content that’s been AI-generated, and content created without AI.
SynthID is our new watermarking tool, designed specifically for AI-generated content. It empowers users to identify AI-generated (or altered) content, helping to foster transparency and trust in generative AI.
How SynthID works
SynthID embeds digital watermarks directly into AI-generated images, audio, text or video. The watermarks are embedded across Google’s generative AI consumer products, and are imperceptible to humans – but can be detected by SynthID's technology.
Slide 1 of 3 AI-generated image and video
SynthID adds an invisible digital watermark to an AI-generated image (or video segment). The watermark doesn’t change the image or video quality. It’s added the moment content is created, and designed to stand up to modifications like cropping, adding filters, changing frame rates, or lossy compression.
AI-generated audio
SynthID embeds a watermark into any audio generated or published through our AI music generation model Lyria or the podcast generation feature of Notebook LM. It’s inaudible to the human ear, and can’t be altered by common modifications like adding noise, MP3 compression, or changing the speed of the track.
AI generated-text
We’ve expanded SynthID to watermarking and identifying text generated by the Gemini app and web experience. Large language models generate text one word (token) at a time. Each word is assigned a probability score, based on how likely it is to be generated next. So for a sentence like “My favourite tropical fruits are mango and…”, the word “bananas” would have a higher probability score than the word “airplanes”. SynthID adjusts these probability scores to generate a watermark. It's not noticeable to the human eye, and doesn’t affect the quality of the output.
Detecting SynthID watermarks in Gemini
Want to check if an image, video or audio clip was generated, or edited, by Google AI? Ask Gemini.
Simply upload the image, video or audio clip to your chat, and ask if it’s been created or altered by Google AI. Gemini will check for a SynthID watermark, and let you know if it finds one.
Try in Gemini How to verify in Gemini Your browser does not support the video tag. Your browser does not support the video tag. Your browser does not support the video tag. Your browser does not support the video tag. SynthID Detector
Earlier this year, we launched the SynthID Detector, a verification portal, to verify if a piece of content was watermarked with SynthID. Just upload an image, video, audio file, or text snippet.
We are currently collaborating with journalists and media professionals to test the portal a
... (truncated, 3 KB total)Resource ID:
fc492fd338071abd | Stable ID: sid_z2P9fh4IAW