Check for a Deepgram watermark
Check audio for the watermark used by Flux TTS.
Drop one audio file
One file • 50 MB • 10 minutes
No account needed. Uploaded audio is deleted after decoding. We do not use uploads to train our models. We use Cloudflare Turnstile to check for automated use.
FAQ
Deepgram watermark detected. Some or all of the audio was likely generated using Deepgram Flux TTS. This does not confirm that the speech is true or accurate.
No Deepgram watermark detected. The audio may still be AI-generated. It may use another model, or its watermark may have been changed or removed.
Inconclusive. There is not enough reliable evidence. Try a longer or clearer recording.
Choose one audio file. Wait for its result before you upload another file. Folders and bulk uploads are not supported.
Use at least 3 seconds of speech per file. Very short clips return an inconclusive result. Each file must be no larger than 50 MB and no longer than 10 minutes.
Common audio formats are supported, including WAV, MP3, M4A, FLAC, OGG, and Opus. Audio from MP4 and WebM files is also supported.
Each network can make up to 3 requests per minute, 10 per hour, and 25 per day. Requests that fail also count. People on a shared network share these limits.
There is also a service-wide limit. If a limit is reached, the page shows how long to wait.
This check looks for the watermark used by Deepgram Flux TTS. It does not check other Deepgram models or other companies’ models. Detection for Aura models is planned for later this year.
Deepgram Flux TTS uses DeepMark technology to add an inaudible signal to generated speech. This check looks for that signal in the audio. It does not verify file metadata.
Editing, poor recording quality, or processing can affect the result. Use the result with other information about the source.