This paper presents design requirements for expressive, avatar-based multi-modal messaging on mobile platforms. It is argued that expressive messaging needs to exploit context of peers, embodied appearance and behaviour, in combination with text. Our approach allows strong expressiveness and yet simple, on the fly message compositions required in a mobile, noisy setting. Technical challenges for a user ready prototype are sketched. It is argued that the context of usage between work-related stationary terminals and mobile ones is radically different.