Bark is an open-source text-to-audio model by Suno AI that generates realistic speech music sound effects and even non-verbal sounds like laughter and crying from text prompts. It can produce complete audio scenes not just voiceovers. Run locally or via API with no cost for personal use.