You can set up transactional replication in two ways. In one way you declare that the subscriber already has the data, and a snapshot is not necessary. This is usually used for topologies that include very large sets of data, or may have limited network bandwidth to push the initial snapshot through. If you take this option, you will need to settle how the initial snapshot makes its way to the subscriber. i suspect it will involve at least a few transaction log backups, and if you have high update volumes, maybe even a short bit of downtime.
The other option (the default option, actually) is to let the snapshot agent bulk copy the data from the publisher to the subscriber. I would advise you to try that out on tables of a few sizes to see what the limits of your setup may be.