I'm working on it! The biggest blocker is the reading direction but that should be solvable. Both AssemblyAI and DeepL support it.