We recently had two retrospectives where parties joined from two continents. It wasn't that different than having the whole team in one room.
The key is where you store the information. It is common that teams write a shared documentation during the retrospective and that kills the dynamics of the event. Select one site and do the writing and drawing part there. The cameras we have now can nicely transfer what is going on, and both parties can see the whiteboards or whatever you have. Of course, the others won't be able to write or draw, but they can explain what they want. If you use a laptop camera that is even better, because you can walk that around the room in case you have multiple whiteboards.
Create photos and distribute them after the meeting, or write a memo afterwards (not during the event). Some other small things that can improve the experience:
- No phones and no laptops besides the camera; based on my experience people use electronic devices more heavily during a video conference and that is not good for the dynamics of the meeting.
- Detailed agenda in advance, what people should prepare, etc; conference calls still start with setting up the equipment, waiting for others, and therefore the meetings are actually shorter. The data gathering part is the longest part, but can be done alone. Do it before the session.
- You can work in groups as well (we did). Have one dedicated person who connects the other sides by explaining them what is on the board and what is happening right now, and who will write or draw what the others would like to share.