What happened
ElevenLabs announced that it is acquiring Kotoba, a Japan-based voice AI lab founded in 2023 by AI researchers Noriyuki Kojima and Jungo Kasai. The companies describe the transaction as a way to accelerate adoption of AI agents and voice interfaces in Japan, which ElevenLabs calls the world’s fourth-largest economy. The acquisition has not yet closed: ElevenLabs says completion is expected later in Q4 2026, subject to customary regulatory approvals and closing conditions.
Kotoba’s voice model, Koto, supports Japanese text-to-speech, speech-to-text, and speech-to-speech for simultaneous interpretation. ElevenLabs says several major global enterprises already use Koto. One named example is Salesforce, where Koto powers Japanese-language conversations in Agentforce Voice. The announcement does not describe changes to Koto’s product, APIs, availability, or pricing as a result of the deal.
ElevenLabs frames the acquisition around the difficulty of making voice interactions feel natural across languages and cultures. Its announcement points to characteristics of Japanese communication, including context-dependent vocabulary, sentences that may omit subjects or objects, indirect ways of declining, and pauses that can signal thought. ElevenLabs says Kotoba’s research and market knowledge make the team a fit for its goal of building AI interactions that feel natural across cultural contexts.
The companies’ leaders describe the intended combination as research and product expansion. ElevenLabs CEO Mati Staniszewski says Kotoba’s work reflects the importance of embedding language and social norms in technology architecture from the outset. ElevenLabs CTO Piotr Dąbkowski says combining the teams’ research should improve Japanese across the company’s stack. Kotoba co-founder and CEO Noriyuki Kojima says the team wants to develop voice AI that benefits people and businesses in Japan and beyond.
ElevenLabs says the acquisition is an initial step in a longer-term plan for everyday AI interactions in Japan. In the coming months, it plans to build out local sales leadership and continue investing in infrastructure in Japan. It also says it expects to partner more closely with Japanese enterprises and multinationals operating or expanding there. The announcement does not give investment amounts, staffing targets, or a product integration schedule.
What the archive adds
New Runtime’s earlier coverage on ElevenLabs’ Campaign Manager (2026-11-09) described a capability for running outbound campaigns across calls and messaging. The Kotoba announcement is a new event, not a restatement: it concerns an acquisition intended to strengthen Japanese voice research and local operations, rather than a campaign-management feature.
Trend signal
This announcement is consistent with agent-ready software exposing capabilities: voice and language technology can become part of an agent product’s usable interaction surface. The evidence here is strategic intent and one stated deployment at Salesforce, not a new capability or proof that the acquisition will improve agent outcomes. The pattern is relevant, but this deal alone does not establish how broadly culturally adapted voice interfaces will be adopted.
Why now
ElevenLabs presents the deal as a step toward broader adoption of agents and voice interfaces in Japan. It says it plans to add local sales leadership and continue infrastructure investment in the coming months, while the acquisition itself remains conditional and is expected to close later in Q4 2026. The announcement gives no more specific schedule for integration or launch.
Connections to other events
New Runtime’s earlier coverage of ElevenLabs’ Campaign Manager (2026-11-09) described a system for outbound campaigns across calls and messaging. The new announcement extends the company’s stated agent ambitions toward Japanese-language interaction, but does not say that Kotoba’s technology will be integrated into Campaign Manager.
Why it matters
A voice agent can perform well on transcription or speech synthesis benchmarks and still feel unnatural in a real conversation. The behaviors ElevenLabs highlights—omitted sentence elements, indirect refusals, and pauses with conversational meaning—point to product requirements that ordinary language coverage may not capture. For teams deploying voice systems across markets, the practical question is whether an agent understands conversational context and responds appropriately, not only whether it recognizes words accurately.
The deal could give ElevenLabs stronger research and enterprise relationships in Japan, but its announcement is a statement of intent rather than a measured result. Builders should treat the Salesforce deployment as an example of existing use, not proof that the acquisition will improve quality, expand access, or reduce integration work. Until those changes are demonstrated, the clearest lesson is to assess localization in the full interaction: recognition, turn-taking, interpretation, and culturally appropriate responses.
Limits and uncertainty
The acquisition is pending regulatory approvals and other closing conditions. ElevenLabs does not disclose the purchase price, the terms of the transaction, staffing plans for Kotoba, or how the companies’ research and products will be integrated. It does not provide comparative evaluations of Koto, measures of conversational naturalness, customer outcomes, or an expanded list of deployments. The announcement’s cultural examples explain the company’s motivation but do not establish how any system handles those situations in practice. The expected Q4 closing window is not a confirmed completion date.
Skill to build now
For a voice agent intended for Japanese users, test more than transcription accuracy. Build evaluation conversations that include omitted sentence elements, indirect refusals, context-dependent politeness, and pauses that may signal a speaker is thinking. Record whether the system interprets intent correctly, waits appropriately, and responds without forcing an overly literal exchange. This is a concrete way to test the kinds of conversational nuance ElevenLabs says matter, before treating language coverage as equivalent to localization.
Reader takeaway
ElevenLabs is buying research and a team it believes can strengthen Japanese voice AI and its local enterprise business. The strategic direction is clear, but the operational result is not yet: the transaction remains pending, and the announcement offers no integration schedule or measured improvement. For builders, the useful distinction is between supporting a language and handling the norms that shape how people actually speak.