Thomas Pornin's last paragraph is right on, but this concept is still so frequently misunderstood that I wanted to add my 2 cents.
From a layman's perspective, If you really had something that implemented a one-time-pad, what would it look like? It would necessarily have to involve a physical machine that generates a truly random key stream. That machine would not take as input any passwords, keys, or seeds -- instead it would generate truly random bits using properties of physics assuming physics is correct in the existence of true randomness. Note that because these bits are truly randomly generated, they cannot be derived again, which implies that the machine would also have to output those key stream bits so you can physically deliver them to the intended recipient, allowing him to decrypt the ciphertext. By the way, when I say "physically deliver", that means getting on plane or in a car, going to your destination, and handing over the bits to the recipient. Why not use public key exchange instead? Well, you could, but that comes at the cost of losing perfect secrecy, so doing so would defeat the entire purpose of using a one-time-pad in the first place.
At this point, you may be thinking this is crazy, and if so, congratulations: now you understand why it is a theoretical construct only, and unlikely to be realised in practice. Instead, what we use today is cryptography that depends upon complexity theory, which is practical.
A stream cipher, from a layman's perspective is an attempt at a practical construct that "feels like" a one-time-pad, except lacking in the perfect secrecy property. It does not need a physical machine to realise it. Instead of true randomness, it takes input a password, key, or seed, and expands that to a long string of bits, which are derived from that input. Those same bits can be derived over and over again by using the same password/key/seed, so in order for the recipient to decrypt, they only need to get that input. And because it does not have the perfect secrecy property and instead only depends upon complexity theory, you are losing nothing when you exchange that key using public key cryptography.
The take-away: just about everything you will ever see in practice that "feels like" a one-time-pad is not really a one-time-pad. It is a stream cipher. People should not label an arbitrary stream cipher as a one-time-pad because they are entirely missing the point. One-time-pads are theoretical concepts, stream ciphers are practical imitations but lacking in the security properties that make one-time-pads what they are.