Hi Product Hunt! 👋

We’re the team behind Makimoto, and we’re excited to show you what we’ve been building.

We’ve worked closely with contact-centre and engineering teams across Asia-Pacific, and the same problems kept coming up. Quality teams replay calls one by one to find out what happened. Transcripts fall apart when a caller switches language mid-sentence. And when an AI pipeline gets something wrong, nobody can tell which step failed.

Most speech AI arrives as a single box. Audio goes in, an answer comes out, and every decision in between belongs to someone else. We built Makimoto so nothing in the pipeline is hidden. 🔍

It turns calls into data you can build on, with speaker-separated transcripts, summaries and tags. There are three ways in:

🔌 REST API and Python SDK for building it into your own product
🧪 Playground for trying it with nothing to install
🧩 Makimoto Flows for designing your own pipeline block by block, then exporting it as code

Makimoto is an open-source initiative. We use and adapt open models rather than training our own, and put our effort into the software around them. Kawa, our orchestration layer, is MIT-licensed on GitHub alongside our Python SDK: github.com/makimoto-ai/kawa

🌏 The pipeline is built for how people in Asia-Pacific speak, with regional accents, multilingual calls and code-switching. Audio and transcripts are stored and processed in Singapore.

🎁 Every account comes with 1,000 free transcription minutes, more than 16 hours of audio. That’s plenty to try it on your own calls.

If you work with multilingual or code-switched calls, we’d especially like to hear from you. What kind of calls would you run through it first? 💬