Florin Radu

Florin Radu, Vitoriano Ruas, Paulo Trales

This work addresses techniques to solve convection-diffusion problems based on Hermite interpolation. We extend to the case of these equations a Hermite finite element method providing flux continuity across inter-element boundaries, shown to be a well-adapted tool for simulating pure diffusion phenomena (cf. V. Ruas, J. Comput. Appl. Maths., 246 p. 234-242, 2013). We consider two methods that can be viewed as non trivial improved versions of the lowest order Raviart-Thomas mixed method, corresponding to its extensions to convection-diffusion problems proposed by Douglas and Roberts (cf. Computational and Applied Mathematics, 1, p. 91-103, 1982) . A detailed convergence study is carried out for one of the methods, and numerical results illustrate the performance of both of them, as compared to each other and to the corresponding mixed methods.

Oleg Atamanenko, Anna Chalova, Joseph Coombes et al.

We introduce Inworld TTS-1, a set of two Transformer-based autoregressive text-to-speech (TTS) models. Our largest model, TTS-1-Max, has 8.8B parameters and is designed for utmost quality and expressiveness in demanding applications. TTS-1 is our most efficient model, with 1.6B parameters, built for real-time speech synthesis and on-device use cases. By scaling train-time compute and applying a sequential process of pre-training, fine-tuning, and RL-alignment of the speech-language model (SpeechLM) component, both models achieve state-of-the-art performance on a variety of benchmarks, demonstrating exceptional quality relying purely on in-context learning of the speaker's voice. Inworld TTS-1 and TTS-1-Max can generate high-resolution 48 kHz speech with low latency, and support 11 languages with fine-grained emotional control and non-verbal vocalizations through audio markups. We additionally open-source our training and modeling code under an MIT license.