LALAL.AI Voice Cloner supporting all languages with good results from short samples is the claim worth testing on your specific language
The all-languages claim is the one worth testing against your specific target language before building a workflow around it. Voice cloning quality varies significantly by language not just because of the model's training data distribution but because of the acoustic characteristics of different languages and dialects. A clone that is convincing in English may have characteristic artifacts in Japanese or Arabic that make it unsuitable for professional use.
The audio quality requirements being well-understood, minimal background noise, consistent microphone distance, clean recording environment, are the input constraints that determine output quality more than the model's capability in many cases. The recommendation of specific sample length and recording conditions is worth following precisely rather than approximating.
The use cases spanning content creation, accessibility tools, multilingual business communications and entertainment are broad enough that the quality threshold varies by application. What is acceptable for a personal podcast is different from what is acceptable for a customer-facing IVR system.
What language did you test LALAL.AI Voice Cloner on and how did the quality compare to your expectations for that specific language?