Improve quality of transcriptions and translations

There are some recent developments that might be of interest to us concerning transcriptions and translations:

Meta's Seemless m4t (transcription and TTT-translations for ~100 languages in one massive language model)

• https://huggingface.co/docs/transformers/main/en/model_doc/seamless_m4t
• https://about.fb.com/news/2023/08/seamlessm4t-ai-translation-model/

Whisper Implementation that reduces GPU load 3x times+:

• https://github.com/bbc/brave

Wav2Vec

• https://github.com/facebookresearch/fairseq/blob/main/examples/wav2vec/README.md