Hide a seed in audio or video

A photo has thirty-odd million low bits. A three-minute WAV recording has about sixteen million samples, each with low bits of its own, and a minute of video has thousands of frames, each a photo. If the image method has room for a seed, these have room for a library. People sometimes read that as "more room means more hidden". It doesn't. The seed needs 132 bits either way. What changes is the carrier's habits, and audio and video have worse habits than photos.

Audio

The uncompressed formats, WAV and AU, store each sample as a plain number, and a tool can nudge the lowest bit of each sample without an audible change. Steghide does exactly this for WAV and AU. The result plays back as before.

Then the file meets a music player, a phone, or a messaging app, and gets converted to MP3, AAC or Opus. Lossy audio compression throws away what the ear can't hear, and the low bits are the first to go. The message doesn't survive the conversion, and the conversion is the default everywhere. A WAV file that stays a WAV file on a drive is fine. A WAV file that gets "added to your library" is gone.

There is a second audio method that survives compression: hide the message in the sound itself, not in the bits. Spread-spectrum and echo-hiding techniques put a signal in the audible band at a level under the music, and a lossy encoder keeps it because it can't tell it from the music. These are research methods and watermarking products, not seed tools. A seed phrase is too small a payload to justify the engineering, and you would be trusting a technique you can't check by hand.

Video

Video is a stack of photos plus audio, so both methods above apply, with every frame as a carrier. Tools exist that spread a message across frames, which makes it harder to find and harder to lose to one damaged frame.

Video also has the worst compression habits of the three. Nearly every video file you have ever handled is lossy (H.264, H.265, VP9, AV1), and every re-encode, which includes uploading, sharing, trimming, and most "export" buttons, generates new frames. A raw, uncompressed video file is enormous and conspicuous. A compressed one won't hold the bits. Video is the carrier with the biggest capacity and the least chance of keeping anything.

The question photos don't get asked

A folder of family photos needs no explanation. A WAV file of a voice memo is odd: phones record memos in a compressed format. A 4 GB uncompressed video of nothing in particular is odder. Steganography depends on the carrier being unremarkable, and audio and video carriers are only unremarkable in the formats that destroy the message.

Ask what story the file tells to someone opening your drive. "Here is the raw stem of a song I recorded" works if you record music. "Here is an uncompressed clip from a wedding" works if you edit video. If you can't finish the sentence, the file is a flag, not a hiding place.

If you do it anyway

The same rules as for images, stricter:

• Keep the carrier in its uncompressed format, and never play it from a phone or a library app that might "optimise" it.
• Keep three copies made by file copy, and check them against each other with a hash.
• Record the tool and version. Extraction depends on both.
• Test extraction on a fresh copy before you fund anything.
• Hide numbers from a cipher, not the words. A file that yields twelve numbers is a second puzzle; a file that yields twelve words is a wallet.

We will never ask for a seed word or a passphrase, and you should never type one into this site or any tool it links to. Try any of this with the public example phrase (pizza educate fat kiss domain venue …) and nothing else.

Where this leaves audio and video

Below photos, and photos are already below paper. The physical page and the storage pages cover the methods that don't depend on a file format staying unchanged for ten years. Audio and video belong to the border-crossing case, where the goal is a copy on a device that looks like nothing, and even there a photo is the simpler carrier.

Sources