Introduction

Synth ID Detector tells you whether a file carries signs of being made or edited by AI. Send it an image, a song, a voice message or a video and it replies with a verdict, the generator if the file names one, and the signs it found. It is free and works inside Telegram.

What it checks

• AI labels in the file. Generators leave open markers: C2PA Content Credentials, IPTC and XMP fields, audio tags and, for Stable Diffusion and ComfyUI, the generation parameters written into PNG files. The bot reads them. See checking an image.
• The SynthID watermark in images. Google puts an invisible watermark into images made with Gemini, Nano Banana and Imagen. A neural network estimates whether it is there even when there are no labels.
• An AI voice in audio. Another neural network listens for synthesized speech in audio, voice messages and the soundtrack of a video. See checking audio and voice.
• Video by its sound. For a video the bot checks the soundtrack and the labels of the file. It does not analyse the picture. See checking a video.

What it cannot do

It is not a lie detector. A file with no labels may still be AI-made: labels disappear with screenshots, recompression and sending a photo with compression, and some generators never write them. The bot cannot see an unlabeled image from Midjourney, Flux, Qwen or Stable Diffusion, the picture of a video, or unlabeled music. Read how to read the result before you rely on a verdict.

Synth ID Detector is a MadBots product and is not affiliated with Google. SynthID is the name of Google's watermark technology that the bot tries to find.

Continue to getting started to check your first file.