Case Study: Lenovo achieves 1,000+ hours of high-quality multilingual voice data at 99% acceptance with LILT

Lenovo delivers 1,000+ hours of multilingual voice data at 99% acceptance with Lilt

A leading e-commerce technology company, Lenovo, needed to scale its multilingual voice-enabled AI systems but found that generic crowd-sourcing could not provide the volume of high-quality, consistent training data required. Their challenge was to grow a voice corpus across five languages without compromising on speaker consistency, accent integrity, or allowing synthetic voices into the training set. They partnered with Lilt to solve this.

Lilt applied a production-grade approach with expert casting, anchored accents, and AI-assisted human review to build the corpus. The solution delivered over 1,000 hours of voice data across five languages with a 99% acceptance rate, successfully screening out all synthetic samples. The results were so trusted that Lenovo licensed six voices from Lilt for use in their shipping product.

Lenovo