The first key to preparing a good talk is taking into account the backgrounds and interests of your audience. This preparation begins with the title and abstract. You need to motivate people to come!
You should NOT give the same talk to experimentalists as to theorists, or the same talk at a specialised conference and a departmental colloquium. This may seem obvious but it is amazing how much it happens. It actually does require significant work, experience, and discipline to give relevant, appropriate, and enjoyable talks.
Next week I am giving the UQ Quantum science seminar. It is attended by physicists, mostly theorists, working in condensed matter, BECs, quantum information, and quantum optics. Below is the abstract I prepared.
A quantum physicist's view of hydrogen bonding
Hydrogen bonding plays a central and diverse role in chemistry and biology. It is key to the unique properties of water, the double helix structure of DNA, and the unique folding of proteins. Yet it is arguably the most poorly understood form of chemical bonding. Indeed the International Union of Pure and Applied Chemistry (IUPAC) recently gave a new definition of hydrogen bonding.
As a physicist I recently investigated hydrogen bonding with three goals:
i. to use the simplest possible model
ii. to describe a wide range of phenomena
iii. to elucidate the role of quantum physics in hydrogen bonding.
I consider a model which describes hydrogen bonding and proton transfer between two molecules due to the quantum mechanical interaction between the orbitals of the H-atom and of the donor (D) and acceptor (A) atoms in the molecules .
The model is based on a effective Hamiltonian which acts on two diabatic states and has a simple chemically motivated form for its matrix elements.
The model gives insight into the "H-bond puzzle" , describes different
classes of bonds (weak, low-barrier, and strong), and gives a quantitative description of empirical correlations between the donor-acceptor distance and binding energies, D-H bond lengths, the softening (hardening) of D-H stretch (bend) vibrational frequencies.
A key testable prediction of the model is the UV photo-dissociation of symmetric
H-bonded complexes via an excited electronic state with an exalted vibrational frequency.
 R.H. McKenzie, Chemical Physics Letters 535, 196 (2012).
 G. Gilli and P. Gilli, The Nature of the Hydrogen Bond (Oxford UP, 2009).