Perplexity's summary: ## Effects and Trade-offs 1. **Latency Reduction**: Going from level 0 to 5, you can expect to reduce the latency by 300-400ms. 2. **Quality Impact**: Higher optimization levels may affect the accuracy and quality of the generated speech. 3. **Text Normalization**: At the highest optimization level (4), the text normalizer is turned off, which can lead to mispronunciations of numbers and dates but provides the best latency.