Toku launches Makimoto for voice AI with local data processing in APAC

Singapore-based Toku launched Makimoto on April 27, an open-source conversational AI initiative built around data residency and local data processing in Asia-Pacific. Its first product, Makimoto Kawa, is planned for July 1 as a managed transcription API hosted in Singapore. Country-specific APIs and a containerised self-hosted version are set to follow.

The important part is the architecture. Toku is opening the orchestration layer under an MIT licence, while components such as voice activity detection, diarisation, noise handling, speech-to-text and post-processing remain managed by Toku for now. The pipeline is modular, so enterprises can swap individual components without rebuilding the whole system.

For leaders in regulated industries, this is a useful example of where voice agents hit reality. Accuracy is rarely the only blocker. The harder question is where audio, transcripts and customer data are processed, and whether the architecture can handle different local processing rules in markets such as Indonesia, Vietnam and Singapore.

Makimoto is still early, and the first API is not available until July. But the direction is right: customer-facing AI infrastructure has to be built for jurisdiction, not just for demos.

Source: Toku via PR Newswire, “Toku launches Makimoto: open-source conversational AI built for Asia-Pacific data residency”, published April 27, 2026.