What is SynthID?
SynthID is an AI detector developed by Google. It allows users to identify whether content—image, video, audio, or text—was generated by AI. SynthID embeds invisible watermarks into AI-generated content during creation. These watermarks help verify content authenticity without altering quality. The tool helps users check if something was created using Google AI systems, including Gemini and Lyria.
SynthID addresses the need for identifying AI-generated content to support transparency, especially when it’s difficult to visually detect AI involvement. The AI detector helps users recognize modified or original synthetic media.
Features & Benefits
- Invisible watermarks – Embed imperceptible digital watermarks in AI-generated images, video, text, and audio.
- Detect modified content – Identify AI-generated media even after edits like cropping, compression, or filter use.
- Text watermarking – Apply watermarking to AI-generated text using probability-based token adjustments without reducing readability.
- Audio watermarking – Embed inaudible watermarks in audio content generated with tools like Lyria and Notebook LM.
- Real-time detection – Detect AI-generated content by uploading a file or pasting text into the SynthID system.
- Integrated into Google AI tools – Automatically watermark content created via Gemini, Lyria, and other Google generative platforms.
- Supports transparency – Help users and platforms verify synthetic content and improve content trust.
- Cross-media coverage – Use a single AI detector for multiple content types: text, image, video, and audio.
SynthID Platforms
Web app
SynthID Tasks
- Detect AI-generated images with AI detector
- Identify AI-generated text using AI detector
- Verify synthetic audio files using AI detector
- Confirm AI-generated videos with AI detector
- Test Google AI content with watermark detection
- Upload and check files for AI-generated origin
SynthID Integrations
n/a
Real-world applications
Content moderators may use SynthID to check whether images or videos posted online were generated by AI. This supports content labeling in social media, especially where manipulated or misleading visuals are common. By uploading the image or video, the AI detector verifies if content originated from tools like Gemini.
Podcasters and musicians generating content with Google’s Lyria model can benefit from SynthID’s audio detection. If a track is questioned for authenticity, SynthID can identify if it was created through AI. The watermark remains intact even after common edits like compression or pitch shift.
Writers and researchers may use SynthID to validate blocks of text suspected to be AI-written. This can help publishers and educators screen AI-generated essays or articles. By analyzing token probabilities, SynthID confirms whether content was produced through Gemini.
Video editors working with AI footage may use SynthID to label video segments. Even after resolution changes or speed adjustments, the tool’s AI detector can find the watermark. This helps maintain accurate content attribution across platforms.
