For Temi Babs, founder and CEO of Lagos-based voice AI startup Spitch, the threat of well-capitalized global players like OpenAI or ElevenLabs turning their attention to African languages is not a cause for alarm. When asked what would happen if those companies decided to make a serious push into the continent's languages, his response was unequivocal: they would have to go through him first.

It is a striking claim from a founder who acknowledges he has "raised very little money" and has yet to reach profitability. Yet Babs insists that in building voice infrastructure for African languages, capital alone is not the competitive advantage many assume. The stakes are considerable: Africa is home to over 2,000 languages, and major global speech and translation platforms have historically offered limited or unreliable support for them, leaving a communications gap that affects everything from customer service to civic participation.

"I understand this so well because we've bootstrapped to get this far ahead," Babs told Technext in an exclusive interview. "I do not think that capital alone is what is going to win this. You could have all the capital, but if you're unable to build…" He paused before making his central argument: any well-funded global giant attempting to serve African markets would need the relationships, the data, and the understanding of the mindset of the average Nigerian, the average West African or East African — assets Spitch has spent two years accumulating. "You can't beat that with billions of dollars."

From Classroom Failure to Infrastructure Play

Spitch's origins trace back to a university lecture transcription project that exposed a fundamental gap in existing speech technology. Lecturers switched between Yoruba, Igbo, and English mid-sentence. Some spoke Pidgin. Others blended two languages in the same breath, against a backdrop of classroom noise that no clean training set had been designed to handle.

"It was impossible for that use case for that product to actually work in the African context," Babs recalled.

Rather than abandon the effort, the team interpreted the failure as a signal. If transcription broke down that severely in lecture halls, the core challenge — enabling machines to reliably hear, understand, and speak African languages — extended well beyond education. It was an infrastructure problem.

The pivot came quickly. Spitch signed its first paying customer a week before its official product launched in October 2024, a milestone Babs still highlights with evident satisfaction. "That gave us validation for the products we're building."

Current Footprint

Two years later, Babs reports that Spitch serves approximately 1,800 developers, four enterprise clients, and fifteen mid-sized businesses that route customer communications through the platform. Named enterprise clients include Safaricom Ethiopia — part of the Safaricom group, one of East Africa's largest telecommunications providers — and the Nigerian news publication Leadership, along with what Babs described as several smaller business-to-business partnerships.

The use cases, he readily acknowledged, are not particularly unconventional: inbound and outbound customer support and content delivery to customers in their native languages.

"Nothing is really surprising from what we have seen," he said, "as long as you tie it to the core fundamental needs of the business."

The Data Challenge

When asked what the most difficult aspect of the work is, Babs pointed not to model architecture but to data.

"Compute is to a large extent limited by how much money you have," he explained. "But data is a different story. There are policies and regulatory frameworks around how you handle and manage data." This regulatory layer, he argued, makes data acquisition harder than the compute problem by "one or two levels." The challenge is compounded by the scarcity of high-quality, labelled speech datasets for African languages — a gap that academic and open-source initiatives have only begun to address in recent years.

Spitch's data strategy has evolved through three distinct phases. The founding team initially recorded their own voices to train early models. They subsequently licensed labelled datasets from partners covering individual languages such as Hausa, Igbo, Yoruba, and Amharic. The current approach involves partnering directly with enterprise clients, using their existing voice data to fine-tune models that are then deployed within the clients' own environments.

The platform can technically support more than 400 languages, inherited from the broader infrastructure it builds upon, but Spitch actively trains and maintains strong support for approximately fifteen to twenty languages across West and East Africa.

Ethics and Provenance

With Nigeria's 2027 elections on the horizon, and following a recent incident in which a presidential spokesperson circulated an AI-generated video, the ethics of voice cloning are far from abstract for the company. Nigeria, Africa's most populous nation with over 220 million people, has seen increasing scrutiny of digital misinformation, and the 2023 passage of the Nigeria Data Protection Act has added a formal compliance dimension to how companies handle personal data, including voice.

Babs stated that every piece of audio the platform generates is cryptographically signed, enabling provenance verification for any file in question. He also emphasized the role of public awareness: "Work with the assumption that anything you see online could be fake; don't believe it immediately unless it's coming from that person directly."

On regulation, Babs struck a measured tone rather than a combative one. Nigeria's National AI Strategy and emerging data protection frameworks are, in his assessment, "quite inclusive" of local builders thus far. His concern is one of timing — that governments risk "over-engineer things too early," imposing compliance requirements on companies before either the technology or its potential harms are sufficiently understood.

Positioning Beyond Geography

Babs is notably averse to being labeled "the ElevenLabs of Africa." He contends that defining a voice AI company primarily by its geographic market understates its actual function.

"Africa is communicating with the rest of the world. Africa is not in isolation," he said. In his framing, Spitch is not a regional replica pursuing a global original; it is infrastructure built for languages the rest of the world has overlooked.

Whether that positioning can withstand a well-capitalized competitor entering the market remains to be seen. But Babs is unequivocal about his five-year ambition: to be involved in at least one-third of all voice communication across Africa.

"Voice is not just a way to communicate," he said. "It is our identity, our culture. It is how we live."