Inaudible watermarks for generated audio

Embed a watermark in the samples of WAV and MP3 files, then sign them with C2PA Content Credentials. A copy that lost its metadata can still be matched to your record.

Get an API key Read the watermark guide

Studio microphone emitting a sound wave, drawn in terminal characters

CR · Content Credentials

MANIFEST: Embedded · valid
WATERMARK: In the pixels or samples
RESULT: Declared origin readable

Recovery is a lead back to your signed record, not cryptographic proof. Picture: AI-generated illustration.

Audio algorithms

Algorithm	`watermark_algorithm`
Meta AudioSeal, streaming, quality-first (default)	`audioseal-mm02-4`
Meta AudioSeal, streaming	`audioseal`

Neither is on the C2PA soft-binding list today, so the watermark is a server-side recovery signal for your account rather than a binding written into the manifest.

Formats and limits

Format	Upload limit	Watermark
WAV	200 MiB	Yes
MP3	100 MiB	Yes
AAC in MP4 (M4A)	256 MiB	No: sign and fingerprint
FLAC	200 MiB	No: sign only

Fingerprints for audio

Add fingerprints with "fingerprint": true next to "watermark": true. The default audio fingerprint is Chromaprint; audfprint and ISCC Audio-Code are selectable. See content fingerprinting.

Billing

One watermark unit per started second of audio, so 12.5 seconds is 13 units. Units are graduated per month, from $0.015 down to $0.0025. Signing is billed once per file.

Watermark and sign an audio file

# After the upload of a WAV or MP3 file is complete:
curl -sS -X POST "$OP_API/v1/sign" \
  -H "Authorization: Bearer $OP_API_KEY" \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d "{\"asset_id\": \"$ASSET_ID\", \"watermark\": true}"
# 202 returns a job. Poll GET /v1/jobs/{job_id}, or use webhooks.

Mark the audio you generate

Create an API key and send a WAV or MP3 file with watermark set to true.

Get an API key Book a demo