DEV Community: Kahunyuro Godfrey

Your Video Has English, Swahili and Kikuyu Sections - Here's How to Get It Right

Kahunyuro Godfrey — Tue, 29 Sep 2026 11:10:25 +0000

A documentary that's mostly English with a Kikuyu-language testimonial. An NGO training video in Swahili with an English-language expert interview cut in. A news package with an English anchor intro and a Swahili interview clip. None of that is unusual - and it quietly breaks most automated video transcription and translation tools, including a lot of dedicated Swahili-to-English translators built for exactly this language pair.

Key takeaways

Most transcription pipelines - generic tools and Kauli's own AI-drafting step alike - assume one language for the whole file, not a different language per section.
A section that doesn't match the assumed language gets transcribed as if it were that language, producing garbled or nonsensical text for exactly that part.
Kauli's real fix isn't automated per-segment detection (we don't have that yet, and say so honestly) - it's that a human editor listens to the actual audio and corrects whatever language each part really is, on every order, regardless.

Why this trips up automated tools specifically

Automatic speech recognition models are usually run with a single language parameter for an entire audio file - it's how Whisper and most other ASR engines are actually called in a typical pipeline, Kauli's included.

That single parameter tells the model what language to expect for every second of the recording, not just the dominant one.

The industry term for a recording that doesn't hold to one language is code-switching - most often used for a speaker mixing languages mid-sentence, but the same underlying problem shows up at the larger scale of a whole section in a different language too.

Either way, a pipeline that assumes one language throughout will mishandle whatever doesn't match it.

What actually goes wrong if nobody accounts for it

If you submit a mostly-English video and tell the tool it's English, the English portions transcribe fine.

The Swahili or Kikuyu section gets forced through the same English-language decoder anyway.

The result is usually garbled, nonsensical text for exactly that section - not a clear error message, just quietly wrong output that looks like a transcript until you actually read it against the audio.

A translation or dub built on top of that mistranscription inherits the same error, now harder to catch because nobody's comparing it back to the source anymore.

What most video-translation tools actually offer today

Searching for how other services handle this turns up mostly single-language-pair tools - an "English to Swahili" translator, a "Swahili to English" one, built the same way, one fixed language assumption per file.

Some larger transcription platforms now advertise dedicated code-switching detection, usually aimed at the mid-sentence version of the problem rather than a whole separate section in a different language.

None of this is a knock on those tools specifically - it's a real, structural limitation of how automatic speech recognition is normally deployed, not a bug unique to any one vendor.

How Kauli actually handles this today - honestly

We don't have automated per-section language detection either, and we're not going to pretend otherwise.

What we do have, on every single order regardless of plan: a real human editor who listens to the actual source audio and corrects the transcript against what's really being said, in whatever language it's really in.

If the AI's first pass mishandled a Swahili or Kikuyu section because the order was set to English, that's exactly the kind of mistake the human review step exists to catch - not a special case, just the same real review every order already gets.

The one thing worth doing on your end

Tell us. When you submit an order with a section in a different language, add a note - "there's a Swahili section around the 4-minute mark" is enough.

The notes field on every order goes straight to the editor working on it, visible on the same brief they check before they start.

It doesn't change what we do - a human is already checking the whole file against the audio either way - but it means they know to expect the switch instead of discovering it mid-review, which usually means a faster, cleaner first pass.

What you actually get back

One transcript, one translation, one set of captions or a dub - covering the whole video, each section in the language it's actually in, checked by a person against the real audio before anything ships.

Not a fully automated per-segment miracle - a real editor doing the part that automation still genuinely can't do reliably yet.

See our formatting standards page for how a transcript is formatted by default, or start with a real clip that has this exact problem - the first few minutes are free to try.

Kenya's New YouTube Withholding Tax Changes the Math on Where Your Audience Lives

Kahunyuro Godfrey — Tue, 29 Sep 2026 11:10:22 +0000

If you're a Kenyan YouTube creator, you've probably already seen this: starting with September 2026 earnings paid out in October, Google withholds 5% of your AdSense-for-YouTube earnings for Kenya-based accounts.

You also need a valid Kenyan KRA PIN on file in AdSense by October 1, 2026, or your payments can be held.

Key takeaways

Google now withholds 5% of AdSense-for-YouTube earnings for Kenya-based creators starting with September 2026 earnings, on top of Kenya's existing digital-content withholding tax rules.
Creators need a valid 11-character KRA PIN submitted in AdSense by October 1, 2026, or risk held payments.
This isn't tax advice - but it's a real reason to look at where your views actually come from, since audience geography now touches more than just growth.

What's actually changing

Google withholds the 5% automatically once it applies - there's no separate form to fill out for the withholding itself.

The KRA PIN deadline is the part that needs action from you: submit it in your AdSense account's Kenya tax information section before October 1.

This is genuinely worth checking your own AdSense account for, and talking to a real tax professional about - the specifics of how this nets against your own filings aren't something a blog post can tell you.

Why audience geography matters more now

A view is a view on your analytics dashboard, but it isn't a view for tax purposes - where the AdSense account is based is what matters here, not where any individual viewer sits.

Still, there's a real, separate reason geography matters beyond this one deduction: international ad rates (US, UK, EU audiences in particular) are typically higher than rates on Kenya-based views, withholding tax aside.

A creator whose audience is entirely Kenyan is now dealing with this withholding on effectively all of their AdSense income - not because of where viewers are, but because that's simply where all the views currently come from.

The real bottleneck is usually language, not content quality

Swahili or Sheng-only content has a hard ceiling: it only reaches people who understand Swahili or Sheng, no matter how good the content actually is.

An English version - subtitles or a full dub - is the concrete lever that moves that ceiling, because it lets a non-Swahili-speaking viewer watch and understand content that was never accessible to them before.

That's a growth story on its own, and it's now also a story about not having every dollar of ad revenue running through one country's tax treatment.

This isn't a reason to panic, and nothing here is tax advice - check your own AdSense account and a real tax professional for what this actually means for your numbers.

But if growing past a Kenya-only audience is already part of your plan, checking whether language is the thing quietly capping that growth is worth doing this week, not eventually - Kauli's free 2-minute sample is a real, no-cost way to see what an English version of your own content actually looks like before committing to anything.

Real Human Voice Actors vs. AI Dubbing Voices: When the Difference Actually Matters

Kahunyuro Godfrey — Sat, 26 Sep 2026 07:42:22 +0000

Modern AI dubbing voices are good.

That's not a marketing claim - for a lot of content, a synthesized voice reading a human-checked translation is genuinely close to indistinguishable from a person, especially at normal viewing speed and attention.

The honest question isn't "is AI dubbing good enough" in the abstract - it's "is this specific piece of content one where the difference actually matters to the people watching it."

Key takeaways

An AI voice, backed by a human-reviewed translation, is genuinely the right call for explainers, training material, and fast-turnaround content.
A real human actor earns the extra cost for flagship pieces, emotionally sensitive content, or a specific recognizable voice.
Adding a human voice actor runs alongside the automated pipeline, not instead of it - the AI dub still ships on the normal timeline.

Where an AI voice is genuinely the right call

Explainers, internal training material, social clips, a fast turnaround on a piece that will be watched once and not re-watched - an AI voice, backed by a human-reviewed translation, delivers real quality at a fraction of the cost and turnaround of booking talent.

There's no credibility reason to pay more here; the content does its job either way.

Where a real human actor earns the extra cost

A few situations consistently come up where clients ask for a real voice, and the reasoning holds up:

A flagship or brand-defining piece - a launch video, a documentary, something with your organization's name on it that's meant to last and be re-watched, not scroll past once.
Emotionally sensitive content - a testimony, a survivor's story, anything where a slightly flat or mechanical read would feel wrong even if every word is accurate.
A specific, recognizable voice matters - a known presenter, a spokesperson whose actual voice is part of the brand, where a generic (even excellent) synthesized voice would feel like a downgrade regardless of accuracy.

How this actually works, practically

Adding a real voice actor doesn't replace the automated pipeline - it runs alongside it.

The AI dub still gets made and delivered right away on the normal timeline, so nothing is held hostage waiting on casting.

The human version is cast and delivered separately once it's ready, which takes longer than an automated dub by definition - a person has to actually record it.

For a multi-speaker video, the same consistency principle applies here too: each character can get their own cast voice, not one narrator reading every part.

What we're not promising

This isn't a same-day service, and it isn't free - a real person's time costs real money, and pretending otherwise would just mean underpaying the person doing the work.

It's priced as a clear, published add-on on top of the standard dub rate, not a custom quote you have to negotiate for.

If your timeline can't accommodate real casting and recording, the AI dub - human-reviewed, same as everything else we deliver - is a legitimately good default, not a consolation prize.

The right question isn't "AI or human" as a blanket policy.

It's asking, per project, whether this is content where the difference is worth paying and waiting for - and building a pipeline flexible enough to give a straight answer either way, instead of forcing every order down the same path.

How to Translate a Video From English to Swahili (or Swahili to English)

Kahunyuro Godfrey — Sat, 26 Sep 2026 07:41:00 +0000

Translating a video between English and Swahili sounds simple until you actually try it: get the words right, sure, but also get the timing right, the tone right, and - if it's going to be dubbed rather than just subtitled - fit a translation into however long the original speaker actually took to say it.

Here's what the process actually involves, and where it tends to go wrong.

Key takeaways

Three real paths exist - do it yourself, hire a freelancer, or use a localization service - each with real tradeoffs in cost, quality control, and effort.
Transcribe accurately against the source audio before translating - a common mistake is translating from an unchecked ASR draft.
If the content is being dubbed, timing is an ongoing problem, not a one-time setting - Swahili and English don't always take the same number of syllables to say the same thing.

Your three real options

Do it yourself. Workable for a short, low-stakes clip if you (or someone on your team) are genuinely fluent in both languages - free, but slow, and easy to get subtly wrong in ways a non-fluent reviewer won't catch.

Hire a freelance translator. Good quality if you find the right person, but you're managing the timing, the subtitle formatting, and often the dubbing separately yourself - and quality varies a lot from one freelancer to the next with no easy way to check before you've paid.

Use a localization service. Handles transcription, translation, subtitle timing and (if you need it) dubbing as one pipeline, usually with some form of quality review built in - worth it once the content matters enough, or there's enough of it, that doing it piecemeal stops making sense.

Transcribe accurately first - don't translate from a rough guess

The most common mistake in a rushed video translation isn't the translation itself - it's translating from a transcript nobody actually checked against the audio.

An automated speech-to-text draft is fast, but it makes real mistakes: a misheard name, a dropped negative (didn't heard as did), a number transcribed wrong.

Translate from that without checking it first, and the translation faithfully reproduces the original mistake in a second language, now harder to catch because nobody's comparing it back to the audio anymore.

Translate for meaning, not word-for-word

A literal, word-for-word translation between English and Swahili often reads stiffly, or misses an idiom entirely - it's raining cats and dogs translated literally means nothing in Swahili.

A good translation carries the actual meaning and tone into natural Swahili phrasing, which is also exactly why a purely automated translation - even a good one - benefits from a human pass checking it reads the way an actual Swahili speaker would say it, not the way a translation engine assembled it.

If it's being dubbed, timing is its own real problem

Subtitles just need to be readable in the time they're on screen.

A dub is harder: the translated line needs to fit roughly the same span of time the original speaker took to say it, or the voice track drifts out of sync with what's happening on screen.

Swahili and English don't always take the same number of syllables to say the same thing, so a literal translation that's accurate can still be the wrong length for dubbing - sometimes the fix is a shorter phrasing that keeps the same meaning, sometimes it's a small, natural stretch to the audio itself.

This is a real, ongoing part of the process, not a one-time setting.

Decide on speaker labels and transcript style up front

Two decisions are much easier to make before translation starts than after: whether speakers should be labeled in the transcript (useful for an interview or panel, unnecessary for a single narrator), and whether you want a clean, readable transcript or a full verbatim one that keeps every false start and filler word - the second matters for a legal or research record, and gets in the way of almost everything else.

Voice cloning needs real, explicit consent - not an assumption

If a dub is meant to sound like the original speaker's actual voice, rather than a standard voice, that's voice cloning - and it should never happen without that person's explicit, on-record consent, checked before the clone runs, not assumed because the source audio was uploaded.

If you're evaluating a service for this, it's worth asking exactly when and how they get that consent, not just whether they offer voice cloning as a feature.

How Kauli approaches this

Every order goes through an AI-drafted transcript and translation first, then a real human editor checks it against the source audio - line by line - before anything's marked ready for delivery.

Timing is handled automatically (real per-word timestamps on the transcript, automatic time-fitting on a dub), speaker labels and transcript style are choices you make when you submit a file, and voice cloning only ever runs after explicit, logged consent.

Pricing is transparent and per-minute, and the first few minutes each month are free to test the actual quality on your own material before spending anything.

See our formatting standards page for exactly how a transcript, dub or caption file is formatted by default, or the video translation service page for the full details.