A cooking assistant that understands Twi-English code-switching should treat English measurements as part of the same recipe conversation. It should follow a spoken instruction in Twi, keep the dish in view, and answer a short English quantity question without asking the cook to start again.
At 6:40 on a Saturday evening in Kumasi, Esi has one hand on a wooden spoon and the other holding her phone above the kitchen counter. Her auntie’s voice note is playing again: instructions in Twi, quick and familiar, moving from the onions to the tomatoes to the point where the stew must sit and reduce.
Esi has already started the base. The pot is hot, and the rice is waiting in a covered bowl beside it. Then the voice note reaches the part she cannot quite pin down: how much stock for the amount of rice she has made.
She does not need a new recipe. She needs one answer inside the recipe already in motion: “For three cups of rice, how many cups of stock?”
If she guesses too low, the rice may stay hard when the food needs to be served. Too much, and the texture changes before she can recover it. There is no spare time for another pot.
With Nkomo, Esi can continue the conversation in the mix of languages that got her there. She can speak the Twi instruction, then ask for the quantity in English. The shift in language carries the same meaning: she is still talking about this stew, this rice, this moment at the stove.
A measurement is part of the instruction
Cooking talk often changes language by task. The memory of how an auntie described the method may arrive in Twi: the timing, the texture, the warning to keep stirring. Measurements may come more easily in English: cups, teaspoons, minutes, oven temperature.
That switch does not signal a new topic. It is one person thinking through one meal.
A voice assistant that makes a cook repeat the dish, the stage of cooking, and the ingredient every time English appears creates friction right when hands are busy. A useful reply holds onto the thread: the cook has mentioned rice, the question is about stock, and the answer belongs in that same exchange.
This is also why short follow-ups matter. “How much?” can mean almost nothing in isolation. After a voice note about rice, it has a clear job.
Natural code-switching lets the conversation sound closer to how instructions travel through families, kitchen counters, group chats, and calls home. The language changes because the thought changes shape, not because the cook has abandoned the recipe.
Voice works best when the pot is already on
Typing while cooking can mean wiping a hand, finding the right words, and watching the oil at the same time. Voice is more practical when the question arrives late.
Nkomo lets people speak or type in Twi and Ghanaian English. In hands-free voice mode, hold to speak, ask the follow-up, interrupt naturally if the answer goes in the wrong direction, and hear the reply.
For Esi, that means she can say the first part in the language that holds the original instruction, then move to English for a quantity. She does not have to translate her auntie’s whole voice note into a formal prompt before asking the one question that matters.
Concise answers matter here. A pot does not need a long lecture while it is catching at the bottom. It needs a clear next action, then room for the cook to ask again if the ingredients or serving size change.
The same principle appears beyond food. A Twi greeting can lead into an English detail, just as a recipe can start with family language and turn to measurements. What Happens When a Twi Greeting Needs English Details? explores why those transitions deserve to be understood as one conversation.
Privacy choices should stay visible
A kitchen question can feel ordinary, yet the way people speak at home still belongs to them. The choice to use cloud processing should be clear at the point it matters.
Nkomo offers explicit cloud-consent controls: never, ask each time, or this session. History stays on the device under the person’s control. Turn history off, and it is purged immediately. People can also export their data or delete their account in one tap.
Those controls give Esi a choice before she starts speaking through a family recipe, a private message, or any other conversation she would rather keep on her terms. The app also shows errors rather than swallowing them silently, so a missed reply does not leave someone wondering whether their Twi caused the problem.
Keep the conversation moving with the cook
A few minutes later, Esi is back at the stove. The rice has gone in. Her phone is on the counter, screen facing up, and she has the next question ready before she needs it: “Should I cover it now?”
The useful assistant response begins with the thread already established. It meets the cook where the recipe is happening, in the language mix that came naturally, with a clear answer before the pot asks for attention again.
Comments
No comments yet.