A voice assistant for Twi and Ghanaian English has to follow the thought as it moves, including the switch from English into Twi. In the kitchen, it also has to answer aloud quickly enough to be useful before the next task demands your hands.
The tap is running. A cook has wet hands, a spoon in one hand, and a pot that needs watching. The question begins in English because that is where the thought starts: “How long should I leave this on?” Then the real concern arrives in Twi: “na ɛrenhye anaa?” Will it not burn?
That switch carries the instruction that matters. The cook needs a spoken answer now, in the same conversation, without stopping to type every word into one approved language.
A thought does not wait for a clean language boundary
People who speak Twi and English together do not always separate them neatly for a device. A question can begin with a measurement in English, move into Twi for the concern, then return to English for a detail. That is ordinary conversation in Accra, Kumasi, London, Toronto, and anywhere Ghanaian families are making food, settling a family question, or trying to remember the next step.
A voice tool that treats the switch as a failure creates extra work at exactly the wrong moment. The speaker has to repeat themselves, simplify the sentence, or choose a language that does not carry the thought as naturally.
Nkomo is built for typed and spoken conversation in natural Twi and Ghanaian English. You can hold to speak, hear the reply, and interrupt naturally when the answer needs correcting or the situation changes. That matters when the question has a clock attached to it, even if that clock is only a pot on the stove.
The useful measure is simple: can you say the thing as you would say it to someone beside you? “Mepa wo kyɛw, should I add the salt now, anaa mentwɛn kakra?” The language mix is part of the request, not noise around it.
Apollo 13 had a problem that could not wait for perfect conditions
In 1970, the Apollo 13 crew faced a crisis after an oxygen tank exploded on the way to the Moon. Jim Lovell, Jack Swigert, and Fred Haise had to preserve the spacecraft’s limited resources while Mission Control in Houston worked through problems whose solutions had to fit the equipment already available.
One of those problems involved carbon dioxide. The lunar module used square lithium hydroxide canisters, while the command module had round ones. The crew needed an adapter. Engineers on the ground developed a way to make one using materials that were on board, and the crew followed the procedure. The stakes were far beyond a kitchen, of course, but the working pattern is familiar: a need arrives in the middle of another task, and an answer that comes too late or in the wrong form does not help.
NASA History documents the Apollo 13 mission and the improvised carbon dioxide solution. The point is not that everyday conversation is an emergency. The point is that useful help must meet the situation people are actually in, with the tools and language available at that moment.
For a cook, the available materials are a wet hand, a boiling pot, and a spoken question that may move between English and Twi. The answer needs to fit that reality.
Spoken help should remain under your control
Speed should not make privacy vague. A spoken question can include a family recipe, a health concern, an address, or a detail someone would never put in a group chat. People deserve a visible choice about what reaches the cloud and what stays on the device.
Nkomo makes that choice explicit. You can set cloud consent to never, ask each time, or this session. Conversation history stays on your device under your control. Turn history off and it is purged immediately. You can also export your data or delete your account in one tap.
Those controls belong close to the conversation, because that is when the decision is real. A person asking about “how much salt” may be happy to send the request to the cloud. The same person may want a different setting before discussing a private family matter. Ghana AI privacy controls: Why Adwoa Needs a Visible Choice Before She Speaks explores why that visible choice matters before speech begins.
Clear replies and clear failures protect the moment
A useful assistant should tell you when it cannot complete a request. Silence leaves the speaker guessing: Did it hear the Twi? Did the connection fail? Should they repeat the whole sentence while the food keeps cooking?
Nkomo shows errors rather than swallowing them. That does not remove every interruption, but it gives the speaker an honest next step. They can retry, change the request, or decide to handle it themselves.
Apollo 13’s crew could act because the problem was made concrete: a mismatch had to be solved with what was on board. Everyday tools earn trust the same way. Say what you understood. Say when something failed. Let the person keep moving.
For English-Twi cooking instructions, the wording can carry practical judgment as well as translation. “Fa Gu So Kakra” Means Start Small is a useful reminder that a short Twi phrase can contain the caution a recipe needs.
Comments
No comments yet.