Companies

ElevenLabs to Acquire Kotoba for Japanese Voice AI

ElevenLabs plans to acquire Kotoba Technologies in Q4 2026, adding the Koto team’s Japanese voice research, Salesforce deployment experience and plans for local expansion.

ElevenLabs to Acquire Kotoba for Japanese Voice AI Expansion
ElevenLabs to Acquire Kotoba for Japanese Voice AI ExpansionRobinson Recalde / unsplash
By James Calloway6 min read

Why it matters

  • ElevenLabs announced the Kotoba acquisition on November 16, 2026 and expects it to close later in Q4 2026, subject to approvals and conditions.
  • The November 16, 2026 announcement disclosed no purchase price, financial terms or closing date.
  • Koto supports Japanese text-to-speech, speech-to-text and speech-to-speech conversion, and already powers Salesforce Agentforce Voice conversations.
  • Kotoba was founded in July 2023; Kotoba Technologies Japan K.K. was incorporated in October 2023.
  • Post-close plans prioritize Japanese, then Korean and other East Asian languages, alongside sales and infrastructure investment in Japan.

By James Calloway

ElevenLabs announced on November 16, 2026, that it plans to acquire Kotoba Technologies, the Japanese voice AI lab behind the Koto model. It expects the transaction to close later in the fourth quarter of 2026, depending on customary regulatory approvals and closing conditions (requirements that must be met before completion). The announcement did not disclose the purchase price, financial terms or a closing date.

ElevenLabs said the combination would join its research resources and global business infrastructure with a Japanese team already developing and deploying speech systems. Kotoba's work covers voice agents (software that converses with people), simultaneous interpretation (live translation of speech) and its own Kotoba interpretation app.

The stakes are practical: Japanese conversation can turn on context, timing and indirect meaning. ElevenLabs calls this expectation kūki o yomu, or “reading the air.” Enterprise agents therefore need more than accurate transcription.

What does Koto add to ElevenLabs?

ElevenLabs describes Koto as supporting Japanese text-to-speech (generating spoken audio from text), speech-to-text (transcribing spoken Japanese) and speech-to-speech conversion (converting speech for simultaneous interpretation). Together, those functions cover input, output and live interpretation. The announcement presents them as a foundation for agents that can handle complex Japanese conversations naturally.

Koto already powers Japanese-language conversations in Salesforce Agentforce Voice, according to ElevenLabs. That named deployment gives the Kotoba team an existing route into enterprise use. The announcement does not say when the deployment began, how many conversations it handles, whether the Salesforce arrangement is exclusive or what commercial terms attach to it.

Kotoba's announcement also says its voice AI powers products at unnamed companies in Japan and the United States. Its interpretation technology supports artists such as YOASOBI, and the company offers a simultaneous interpretation app called Kotoba. That app shares the company name, while Koto is the model.

Takeaway: ElevenLabs gains an existing Japanese enterprise deployment, but the announcement provides too little evidence to judge its scale.

Why is Japanese more than another language option?

ElevenLabs identifies four linguistic features that complicate voice-agent design. Each changes how an agent must interpret timing, intent and relationship.

  • Japanese sentences can omit subjects and objects.
  • Vocabulary can change with the relationship between speakers.
  • Pauses can signal thought rather than awkwardness.
  • People often avoid direct refusals.

Kotoba turns those observations into a product requirement, according to its announcement. A natural system must understand context, follow the flow of a conversation and decide when to respond, in addition to recognizing and generating speech accurately. That requirement helps explain why the team's planned post-closing role spans research, product development and deployment support.

For a voice agent, recognizing every word does not guarantee a useful reply. A system can mistime its response or miss an implied refusal even when the transcript itself is accurate. The failure would occur in the conversation's meaning, not merely in its audio processing.

The announcement offers no accuracy, response-time, interruption-handling or comparative performance data for Koto. Without those measures, buyers cannot independently assess the claim that Koto can support agents handling complex Japanese conversations naturally. It describes a combined research-and-deployment effort, but supplies no benchmark showing that the model outperforms alternatives.

Takeaway: The goal is to understand meaning and timing, not just words, but the announcement provides no measurements showing how well Koto performs.

Who is behind Kotoba?

AI researchers Noriyuki Kojima and Jungo Kasai founded Kotoba in July 2023, according to ElevenLabs. Kotoba's corporate profile lists a San Francisco headquarters and a Tokyo operation, Kotoba Technologies Japan K.K. The profile records the Japanese entity's incorporation in October 2023.

Kojima holds a computer science PhD from Cornell and worked as a research scientist at Hugging Face. He co-built Fugaku-LLM, an open-source model trained from scratch on the RIKEN Fugaku supercomputer.

Kasai holds a computer science and engineering PhD from the University of Washington and served as an assistant professor at Toyota Technological Institute at the University of Chicago. ElevenLabs cites the founders' research backgrounds as an anchor for its East Asian language ambitions.

Still, the announcement gives no Kotoba funding history, employee count, revenue figure, customer count, model size, training-data size or third-party benchmark. The credentials establish relevant experience; the release does not establish the company's commercial scale.

Takeaway: Kotoba brings credible research credentials, but no public operating metrics accompany the acquisition announcement.

What changes after the deal closes?

ElevenLabs expects the Kotoba team to continue across three areas after closing. Closing remains conditional, and the announcement gives the continuation no end date or staffing target.

  • Research into Japanese and other East Asian voice models.
  • Product development informed by real-world Japanese deployments.
  • Deployment support for enterprise customers.

Kotoba plans to combine its East Asian language expertise with ElevenLabs' research capabilities and computing resources. The stated sequence is to improve voice AI in Japanese first, then in Korean and other East Asian languages. Insights from real-world Japanese deployments are intended to feed models and products ElevenLabs delivers worldwide.

ElevenLabs also plans to build local sales leadership and continue investing in infrastructure in Japan. The company plans closer work with Japanese enterprises and multinationals operating or expanding in the country.

Under ElevenLabs' framing, Japan will have a dual role. It will supply deployment-driven research for worldwide products and serve as a market for ElevenLabs' commercial and technical expansion. The announcement gives no hiring target, infrastructure budget, product schedule or revenue forecast, and the deal still depends on approvals and closing conditions.

Takeaway: Japan becomes both a research source and a sales market, though the announcement sets no measurable targets.

Can ElevenLabs turn Koto into a broader business?

The clearest concrete evidence is Koto's presence in Salesforce Agentforce Voice. The central strategic claim is that lessons from Japanese deployments can improve ElevenLabs products worldwide. Neither point establishes near-term revenue, and the announcement does not disclose the price or terms that would let outsiders test the deal's financial logic.

The announcement also leaves several operating questions unanswered. It does not describe the Kotoba app's roadmap, the treatment of the Koto brand, or how acquired research will enter ElevenLabs product release schedules.

The announcement also provides no usage figures, commercial customer names beyond those cited, or comparisons with other Japanese voice models. Those omissions do not negate the combination, but they leave its scale and execution uncertain.

For ElevenLabs, the test is whether production lessons from Japanese agents improve its wider model portfolio while the Kotoba team turns local deployments into a durable Japanese business. The announcement establishes the strategic logic, but leaves the technical and financial evidence for later.

Takeaway: ElevenLabs must turn Japanese deployment experience into measurable model gains and a durable local business.

Source: unite.ai

Share this article:

More from James Calloway

James Calloway

Show full bio

News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.

722 articles

Related articles

  1. Alibaba's Qwen launches Audio 3.1 with five models, cuts prices up to 95%
  2. LG Uplus launches K-Voice AI Alliance with three universities
  3. ElevenLabs expands into enterprise agents, tuning and partnerships
  4. ElevenLabs launches Eleven v4 with tighter voice control
  5. Nvidia's open Nemotron 3.5 Transcribe tops speech benchmarks on one GPU

« Previous articleNext article »