There are roughly two ways to start.
One is from the stems of the stereo mix. Stems are usually stereo files of groups of instruments or vocals, like drums, percussion, guitars, keys, backing vocals, vocals and so on.
Usually these stems already contain eq, compression and even reverbs and delays. Reverbs and delays affect your Atmos mix a lot, so if you can get them without and have them printed seperately, that would give you a much bigger freedom.
Preferrably the stems come with a stereo master file. This can be seen as the reference file, as the producer as well as the artist both agreed on the mix.
So, when I start a session, I import the stereo file and the stems, and make sure the sum of these stems sound very similar if not almost identical to the stereo master.
Keep in mind that the stereo master has been mixed in a totally different way than you would do in Dolby Atmos: busses, bus compression, sidechaining to create space, and ofcourse the masterbus treatment with limiters and what not. So, you might want to bring the volume of the stereofile down in order to match the maximum level protocol from Dolby Atmos, which is -18db LUFS integrated.
Once you are close enough to the stereo mix its time to place the different elements in your 3d space. This is where the fun begins, but dont lose your head over the space and always be aware of the song. Actually that doesnt differ much from regular mixing 🙂
The more stems you have, the more precise you can place different elements of the track in the room.

