Given a script, the challenge in Movie Dubbing (Visual Voice Cloning, V2C) is to generate speech that aligns well with the video in both time and emotion, based on the tone of a reference audio track. Existing state-of-the-art V2C models break the phonemes in the script according to the divisions between video frames, which solves the temporal alignment problem but leads to incomplete phoneme pronunciation and poor identity stability. To address this problem, we propose StyleDubber, which switches dubbing learning from the frame level to phoneme level. It contains three main components: (1) A multimodal style adaptor operating at the phoneme level to learn pronunciation style from the reference audio, and generate intermediate representations informed by the facial emotion presented in the video; (2) An utterance-level style learning module, which guides both the mel-spectrogram decoding and the refining processes from the intermediate embeddings to improve the overall style expression; And (3) a phoneme-guided lip aligner to maintain lip sync. Extensive experiments on two of the primary benchmarks, V2C and Grid, demonstrate the favorable performance of the proposed method as compared to the current state-of-the-art. The code will be made available at https://github.com/GalaxyCong/StyleDubber.

Summarisation of long financial documents is a challenging task due to the lack of large-scale datasets and the need for domain knowledge experts to create human-written summaries. Traditional summarisation approaches that generate a summary based on the content cannot produce summaries comparable to human-written ones and thus are rarely used in practice. In this work, we use the Longformer-Encoder-Decoder (LED) model to handle long financial reports. We describe our experiments and participating systems in the financial narrative summarisation shared task. Multi-stage fine-tuning helps the model generalise better on niche domains and avoids the problem of catastrophic forgetting. We further investigate the effect of the staged fine-tuning approach on the FNS dataset. Our systems achieved promising results in terms of ROUGE scores on the validation dataset.