This paper focuses on the issue of reducing the bandwidth requirement for FHE ciphertext transmission. While this issue has been extensively studied from the uplink viewpoint (transmission of encrypted inputs towards an FHE calculation), where several approaches exist to essentially cancel FHE ciphertext expansion, the downlink case (transmission of encrypted results towards an end-user) has been the object of much less attention. In this paper, we address this latter issue with a particular focus on the TFHE scheme, for which we revisit a number of folklore methods, including several approaches for switching to more compact linearly homomorphic schemes, reducing the precision of T(R)LWE coefficients (while maintaining acceptable probabilities of decryption errors), and others. We also investigate how to use these methods in combination, depending on the number of encrypted results to transmit. We further perform extensive experiments demonstrating that the downlink TFHE ciphertext expansion factor can be practically reduced to values below 10, depending on the setup, with little additional computational burden.