en
Русский- Images: labels, generator, SynthID watermark estimate
- Audio and voice messages: labels and AI voice estimate
- Video: soundtrack and file labels
- The generator name when the file carries it
- Plain verdicts with the signs found
- Links to the official checkers for a manual check
Find out if an image or audio was made by AI
Send a file to the bot. It reads the AI labels that generators write into files, looks for the invisible Google SynthID watermark in images and for an AI voice in speech, and tells you what it found and what it could not see.
- Images, audio and video
- Labels, watermark, AI voice
- Free
Three steps, no app to install
Everything happens in your chat with the bot.
1
Send the file
Send an image as a file (paperclip, then File), not as a photo. Audio, voice messages and video can be sent as they are.
2
The bot checks it
It reads the labels inside the file, looks for the SynthID watermark in an image and for an AI voice in speech. For a video it checks the soundtrack.
3
Read the verdict
AI-generated, AI-edited, Content Credentials without AI, or no labels found, with the generator and the signs it found. Every answer says what the check cannot prove.
Telegram strips all metadata from a photo sent with compression, so there would be nothing to check. That is why the bot asks for images as files.
What the bot checks
Three kinds of signs, and honesty about what the bot cannot see.
AI labels in the file
C2PA Content Credentials, IPTC and XMP markers, audio tags and the generation parameters that Stable Diffusion and ComfyUI write into PNG files.
Who made it
When the file says so, the bot names the generator: Google AI, OpenAI, Adobe Firefly, Midjourney, Stable Diffusion, Flux, Suno, Udio, ElevenLabs, Runway and more.
SynthID watermark in images
A neural network looks for the invisible Google SynthID watermark even when there are no labels. It finds about 9 in 10 Gemini images and the bot calls the result likely, not certain.
AI voice in audio
A neural network listens for synthesized speech in audio and voice messages, even without labels. It finds about half of it and does not check music or singing.
Video by its sound
For a video the bot checks the soundtrack and the labels of the file itself. It does not analyse the picture, and it says so.
Clear verdicts
AI-generated, AI-edited, Content Credentials without AI, or no labels found, with the signs behind each one.
Honest about limits
A missing label is not proof that a file is real. The bot says what it cannot see and links the official checkers for a manual check.
Free and fast
The answer comes in a couple of seconds, and the bot has no paid plan.
Who this is for
A viral picture
Check an image that went around the chat or the news feed before you share it.
A suspicious track
See whether a song carries labels from Suno, Udio, ElevenLabs or another music generator.
Submissions and portfolios
Check images and audio sent by contractors or contest entrants for labels that name an AI generator.
A voice message or a recording
Check a voice recording for signs of a synthesized voice, then confirm with the official checker if it matters.
Pricing
Synth ID Detector is free. There is no paid plan and no credits: send an image as a file, or an audio or video, and get the answer.
Questions people ask before the first check
Is it free?
Yes. The bot has no paid plan and no credits: you send a file and get the answer.
What can I send?
Images as a file (JPG, PNG, WebP, HEIC), audio files, music, voice messages, video files and video notes. A photo sent with compression is not checked: the bot asks you to send it as a file.
Why should I send the image as a file?
When Telegram compresses a photo it strips all metadata, and the AI labels live in the metadata. As a file, the image arrives untouched.
How does the bot know it is AI?
Three ways. It reads the labels the generator writes into the file (C2PA, IPTC, XMP, PNG parameters, audio tags). For images it estimates the invisible SynthID watermark with a neural network. For audio and video sound it estimates an AI voice with another neural network.
If the bot finds no labels, is the file real?
Not necessarily. Labels disappear with screenshots, recompression and sending as a photo, 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. Every answer without labels says so.
How accurate is the SynthID check?
It is an estimate by an open approximation model, not Google's official detector. It finds about 9 in 10 Gemini images, and occasionally fires on other generators, so the bot calls the result likely. To check for sure, use synthid.com.
How accurate is the AI voice check?
It finds about half of synthesized speech (ElevenLabs, TTS voice-overs, voice clones) and occasionally fires on a real voice. It does not check music or singing, and it does not judge recordings shorter than 4 seconds. For an ElevenLabs voice, use their AI Speech Classifier.
Does it check video?
By the soundtrack and the labels of the file only. The bot does not analyse the picture and says so in the answer.
Does the bot keep my files?
The bot downloads a working copy to run the check and deletes it afterwards. The check itself is stored as a short record of the file type, size and the result. See the privacy policy.
Is it a Google product?
No. 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.
Check your first file
Open the bot, send an image as a file, or an audio or video, and read the verdict.