Local processing versus explicit cloud consent: Boafoc and Nkomo represent two different, honest approaches to privacy-first African-language voice AI, and Nkomo should never be mistaken for an offline assistant.

Boafoc and Nkomo take two distinct privacy-first approaches to African-language voice AI: Boafoc emphasizes local processing, while Nkomo gives people explicit control over cloud use and local history. Nkomo requires cloud processing when consent allows it, so it should never be described as an offline assistant.

Start by separating two privacy models

“Privacy-first” can describe more than one technical design. The important question is what happens to your voice or text after you provide it.

Local processing keeps supported work on the device. This can reduce the amount of personal data sent elsewhere and may suit situations where connectivity is limited or cloud transmission is unacceptable.

Explicit cloud consent takes a different route. The service may use cloud processing, but you decide whether data can leave the device. Nkomo provides three choices:

• Never allow cloud use.
• Ask for permission each time.
• Allow cloud use for the current session.

These options make the boundary visible. They do not turn cloud-dependent functions into offline functions.

That distinction matters when comparing products, writing procurement requirements, or explaining an assistant to a family member. “Private,” “local,” and “offline” answer different questions.

Check what “local” covers

A product may process one part of a voice interaction locally while sending another part to the cloud. For example, recording controls, saved history, speech recognition, response generation, and audio playback can each follow different paths.

Before choosing an assistant, check:

• Which operations happen on the device?
• Which operations require cloud processing?
• Does the app ask before sending content?
• What happens when cloud access is refused?
• Where is conversation history stored?
• Can you delete or export your data yourself?

Boafoc’s local-processing approach demonstrates demand for African-language voice tools that keep processing close to the user. Nkomo addresses the same privacy concern through visible choices about cloud access and on-device retention. Both approaches can be honest, provided their descriptions match their actual behaviour.

Understand what Nkomo controls

Nkomo supports natural Twi and Ghanaian English conversation by voice or text, including the code-switching people use in ordinary speech. In voice mode, you can hold to speak, interrupt naturally, and hear the reply.

Its privacy controls focus on informed choice. You can refuse cloud use, approve it one request at a time, or approve it for the current session. This lets you adjust consent to the sensitivity of the conversation instead of accepting one permanent setting.

History stays on the device under your control. Turn history off and it is purged immediately. You can also export your data or delete your account with one tap.

Errors remain visible. If something fails, Nkomo shows the error instead of silently swallowing it. That matters because a privacy control is useful only when you can tell whether the requested action happened.

For a practical example of session-level consent, read The Cloud Choice Abena Made Before Her Family Voice Note Could Leave Drafts.

Avoid calling Nkomo an offline assistant

Offline normally means the core task can run without contacting remote servers. That description does not accurately represent Nkomo.

Do not infer offline capability from on-device history. Local storage tells you where saved conversations remain. It does not tell you where a request is processed.

Do not infer offline capability from a “never” cloud setting either. Refusing cloud consent controls whether data may be sent. It does not create a local replacement for processing that needs the cloud.

This wording is accurate: “Nkomo provides explicit cloud-consent controls and user-controlled on-device history.”

This wording is inaccurate: “Nkomo is an offline Twi voice assistant.”

Precise language helps people make informed choices, especially when a conversation includes a family matter, financial detail, health concern, or identifying voice recording.

Choose based on the conversation

Start with the data, not the feature list. Ask what the conversation contains and what must happen for you to feel comfortable.

Choose local processing when keeping supported processing on the device is a firm requirement. Confirm exactly which functions work locally and what happens without a connection.

Choose explicit cloud consent when cloud processing is acceptable under conditions you control. For Nkomo, decide whether “never,” “ask each time,” or “this session” matches the sensitivity of the task before speaking or typing.

Then review retention separately. Cloud consent governs whether content may leave the device. History controls govern what remains locally. If you do not want a conversation retained, turn history off and confirm it has been purged. [This guide to voice-note privacy](\/blog/ai-voice-note-privacy-how-efua-understood-twi-and-english-then-purged-the-chat-7d233994/) explains that distinction through a Twi-English interaction.

Take one concrete privacy check

Before your next voice conversation, open the privacy controls and choose a cloud-consent setting deliberately. After the conversation, check whether local history is enabled. Export what you need, purge what you do not, and treat any claim of “offline” as something to verify function by function.