The field of text-to-speech synthesis is witnessing significant advancements, driven by the development of more sophisticated models and evaluation frameworks. One of the key areas of focus is the improvement of expressive speech synthesis, particularly in languages with complex pitch-accent systems. Researchers are exploring the use of linguistic cues, such as syntactic boundaries and sentence stress, to enhance the naturalness and intelligibility of synthesized speech. Another important trend is the creation of comprehensive benchmarks that can accurately assess the performance of text-to-speech models in various scenarios, including emotions, paralinguistics, and linguistic complexity. These benchmarks often employ innovative approaches, such as model-as-a-judge or synthetic data generation, to evaluate the capabilities of state-of-the-art models. Noteworthy papers in this area include the introduction of StressTest, a benchmark designed to evaluate a model's ability to distinguish between interpretations of spoken sentences based on stress patterns, and EmergentTTS-Eval, a comprehensive benchmark that covers six challenging TTS scenarios and employs a model-as-a-judge approach. These developments are expected to pave the way for more advanced and realistic text-to-speech systems.