Kikuyu Dubbing: What's Actually Possible Today

We get asked this often enough that it's worth answering honestly and in public: can Kauli dub Kikuyu (Gĩkũyũ) content into English? The short answer is yes, with one real limitation worth understanding before you send us a job - and no, we're not going to pretend that limitation doesn't exist.

The actual blocker: no speech-recognition model understands Kikuyu

Every AI dubbing or transcription pipeline - ours included - starts with automatic speech recognition (ASR): software that listens to audio and turns it into text.

We checked this carefully before saying anything to clients.

OpenAI's Whisper, the model family most local and cloud transcription tools are built on, supports 99 languages - Swahili is one of them.

Kikuyu is not.

Google Cloud's Speech-to-Text doesn't support it either, though it does support Kenyan Luo and Kamba.

This isn't a small gap or an oversight anyone is likely to close soon.

Kikuyu, spoken by roughly 7 million people in Kenya's central region, is what's called a low-resource language in speech technology - there simply isn't enough transcribed audio data publicly available yet for a general-purpose model to have learned it.

(There is promising academic research on this - a Kikuyu-specific Whisper fine-tune from Dedan Kimathi University reports real progress - but at a 36% word error rate, it's a research result, not something we'd put in front of a paying client without a lot more validation.)

What we do instead

Rather than either refusing Kikuyu jobs outright or quietly running a model we know doesn't work and hoping nobody notices the errors, we do the honest thing: a Kikuyu-speaking transcriber types the source transcript by hand, directly into the same editor our staff use to correct every other job.

From that point on, the pipeline is identical to a Swahili or English order - AI-assisted translation into English, a human editor checking every line, and (if you ordered a full dub) voice synthesis for the English output.

Two things are worth knowing if you're considering this for a real project:

• It costs more per minute. Real human transcription time is real labor cost, in a way that AI transcription for Swahili or English isn't. We priced this against the actual professional-transcription industry rate rather than guessing.
• It takes a bit longer. A human typing a transcript by hand is not instant the way an AI draft is - budget extra turnaround time versus a Swahili or English job of the same length.

Why we're telling you this instead of just saying "yes, we do Kikuyu"

Because the alternative - quietly running an ASR model we know doesn't understand the language and shipping whatever garbage it produces - would be worse than not offering the service at all.

We'd rather you know exactly what you're getting: real human transcription quality, on the same review pipeline as everything else we deliver, at a fair price that reflects the actual work involved.

If your project needs Kikuyu, get in touch and we'll walk you through it.