When looking at the first issue in information hiding, determining levels of visibility can be done by asking one question: Does the embedding process distort the cover to the point where it is visually noticeable? If the image is unacceptably distorted, the carrier is not sufficient for the payload; if it is not distorted, the carrier is enough.
Most things in life require trade-offs of some sort. Something for nothing is not the way the universe works, and information hiding is no different. To have a robust method of embedding the message means you must have redundancy to resist changes made to the cover. This redundancy subsequently lowers the payload.
The exact opposite is also true. With little or no redundancy (robustness) the payload of the secret message can be larger.
More robust = lower payload
Less robust = higher payload
Some image and sound files are either lossy or lossless; the conversion of lossless information to the compressed "lossy" information can destroy the hidden information in the cover. A good example is to think of the conversion of an uncompressed bitmap to a compressed, estimated JPEG. The compression and estimation change the bits to include the bits that might be the embedded message.