Martina Russo, CEO of The Action Sports Translator (TAST), joins SlatorPod to talk about her journey from being a multilingual outdoor sports enthusiast to leading a language solutions integrator dedicated exclusively to action and outdoor sports.

Martina describes how each action sport has unique subcultures and terminology, from mountain biking’s multiple disciplines to climbing’s variety of techniques and jargon. Even within the same sport, regional language differences present challenges.

For Martina, authentic translation in this industry means more than linguistic quality; it requires a translator who genuinely lives and breathes the sport to capture its culture, humor, and insider tone. She emphasizes the difficulty in sourcing linguists who are both trained translators and passionate sports practitioners, especially for rare languages or specific verticals.

The CEO shares how TAST deploys AI internally for operational efficiency and leverages it in content creation and localization, though she’s cautious about fully replacing human experts. Product copy and FAQs are areas where AI is more applicable, but the authentic, emotional connection critical to sports brands often requires a human touch.

Despite industry challenges amid fluctuating demand, TAST has been experiencing strong growth, even surpassing previous expectations. Martina attributes this success to unwavering niche focus, technological adaptation, and a company culture deeply immersed in the sports it serves.

Looking ahead, Martina remains committed to investing in developing AI solutions, hiring for roles in that department, and staying on top of trends in the outdoor sports and organization industries.

Florian and Esther discuss the language industry news of the week, including the newly released Slator 2025 Language AI 50 Under 50, showcasing fifty of the most innovative and fast-growing language AI startups founded within the past fifty months.

The duo explain how Slator sifted through hundreds of companies, assessing innovation, practical solutions to real buyer problems, and strong market positioning. The final fifty span five categories: multilingual video and audio, live speech translation, transcription and captions, translation and text generation, and accessibility.

The conversation then moves on to language AI and services in the public sector. Esther talks about a new language AI tool, DiploIA, developed and deployed by the French Government for diplomatic agents in sensitive missions.

Turning to the US, Esther reports that SOSi secured a significant USD 260m language services contract with the US Drug Enforcement Administration. Meanwhile, the US Defense Health Agency is looking for providers to deliver large volumes of translation and interpreting services.

Esther also revisits the major acquisition of CyraCom by Propio, calling it one of 2025’s biggest language industry deals. Propio now joins forces with CyraCom’s established presence in healthcare and legal interpreting, creating a combined entity with revenues exceeding half a billion dollars and positioning them strongly in the US interpreting market.

Florian questions AI voice startup ElevenLabs’ plans for an IPO within five years. He then wraps up the pod by exploring large reasoning models (LRMs) and their mixed performance in AI translation. While LRMs outperform traditional LLMs in complex, open-domain translation tasks, research indicates they remain prone to significant weaknesses.

Florian and Esther catch up on a few weeks’ worth of language industry news with a surge of developments in speech translation. Apple’s on-device translation debuts in apps like iMessage and FaceTime, and OpenAI enhances ChatGPT‘s Advanced Voice Mode with more human-like interactions and real-time translation.

Florian unpacks YouTube’s broad rollout of AI dubbing for 80 million creators in 20 languages, where he trials German and finds robotic voices, bad translations, and no editing options, leaving much to be desired.

Esther talks about RWS acquiring Papercup’s IP, aiming to embed AI dubbing into Trados and significantly boost RWS’s capabilities and market reach. RWS’s half-year financials also show slight revenue drops but rising AI-driven revenue, alongside a reorganization into three divisions: Generate, Transform, and Protect

The duo analyzes Meta’s USD 14bn investment for a 49% stake in Scale AI, which raises concerns from other tech giants uncomfortable with a major competitor owning a key data-labeling supplier. They note opportunities arising for competitors like Labelbox, RWS’s TrainAI, Welo Data, and many other LSIs as clients reconsider vendor relationships in light of Meta’s involvement.

In Esther's M&A corner, Propio acquires CyraCom to become a half-billion-dollar language solutions integrator, DigitalTolk buys 24translate to expand into the DACH region, and Powerling boosts its life sciences footprint with the acquisition of Idem.

Rounding out the episode are leadership changes, with XTM appointing Rob Finney as CMO and CQ Fluency naming Tameeka Smith as CEO following the long tenure of Elisabete Miranda.

Davit Baghdasaryan, Co-Founder & CEO of Krisp, joins SlatorPod to talk about the platform’s journey from a noise cancellation tool to a comprehensive voice productivity solution.

Originally built to eliminate background noise during calls, Krisp has expanded into real-time accent conversion, speech translation, agent assistance, and note-taking — technologies being rapidly adopted in enterprise environments.

The company’s accent conversion is now deployed by tens of thousands of call center agents, significantly improving metrics like customer satisfaction and call handling time.

Davit details Krisp’s live AI speech translation feature which uses a multi-step pipeline of speech-to-text, text translation, and text-to-speech. The majority of use cases are in high-resource language pairs such as English, Spanish, French, and German, although Davit recognizes the importance of expanding to lower-resource languages.

He shares how Krisp is also seeing increased demand in human-to-AI voice communication. AI voice agents are highly sensitive to background noise, which disrupts turn-taking and response accuracy. Krisp’s noise isolation tech plays a foundational role in enabling smoother voice AI interactions, with billions of minutes processed monthly by major AI labs and startups.

The CEO discusses LLMs' impact, noting how they enable advanced features like note-taking, call summaries, and real-time agent guidance. He sees Krisp as a platform combining proprietary technologies with third-party AI to serve both B2B and B2C markets.

Davit advises startups to explore the vast and still underdeveloped voice AI landscape, noting the current era is ripe for innovation due to AI advancements. He highlights Krisp’s roadmap priorities: expanding accent packs, refining voice translation, and building more AI Agent Assist tools focused on voice workflows.

Stefaan van der Jeught, Professor of EU Constitutional Law at Vrije Universiteit Brussel, and a Press Officer at the Court of Justice of the European Union, joins SlatorPod to talk about the complex relationship between language and law in the EU.

Stefaan outlines the historical evolution of EU language policy, from French-only founding treaties to the gradual inclusion of all member state languages. Despite formal equality, institutions largely define their own internal language regimes, leading to fragmented and often English-centric practices. 

Stefaan’s book EU Language Law, now in its second edition, examines these issues in depth. Updated with new case law, legislation, and developments in AI and governance, it includes a 10-point roadmap for reform. Stefaan advocates for greater transparency, legal protection of linguistic diversity, and a constitutional debate on the role of language in EU integration.

AI, Stefaan believes, is a tool that can enhance multilingual access and consistency across EU communications. However, he cautions against using AI as a cost-cutting measure that replaces linguistic expertise. Instead, AI should serve as a support tool, with human revision, especially in legal contexts.

On regional languages like Catalan, Basque, and Galician, Stefaan notes they face legal hurdles at the EU level because they lack full legislative status in their home countries. He argues for a more transparent and constitutional debate on language policy, drawing inspiration from multilingual countries like Switzerland and Belgium.

Stefaan concludes by advising universities to train future legal linguists by going beyond technical instruction to foster critical thinking, comparative law expertise, and cultural literacy.

Tom Elias Hanna, COO of Hanna Interpreting Services (Hanna), joins SlatorPod to talk about how a family-run interpreting initiative grew into a national language solutions integrator (LSI).

Tom explained that the company was born out of necessity during an influx of Arabic-speaking refugees in San Diego, with his mother providing interpretation and him leveraging his legal background to establish a compliant, scalable business that now serves healthcare, government, education, and social service sectors across California.

Tom described how their inclusion-focused work led to partnerships like one with San Diego FC, where they provide ASL at every home game. He emphasized that while AI holds potential in sign language interpretation, it must evolve with and for the deaf community to be truly effective, due to cultural, emotional, and experiential nuances that current technologies cannot replicate.

On-site language interpretation remains the core service, especially across healthcare, education, and social services. Though remote interpreting and AI are on the rise, Tom emphasizes the irreplaceable value of human interpreters, particularly in high-stakes, emotionally nuanced settings like hospitals.

Tom explained that their recent rebrand emphasizes both human connection — central to their on-site interpreting services — and technological growth.

Initially, without a sales team, Hanna grew through referrals, client satisfaction, and high service quality. Only after COVID did Tom begin to formalize a sales strategy and identify account management as a natural extension of their client-first approach.

Tom expressed that despite considerable internal and industry-wide discussions, the Trump Executive Order designating English as the only official language of the US had no tangible effect on Hanna, so far. He noted that no clients inquired about it or changed their behavior.

Looking ahead, Tom aims to scale geographically, explore strategic acquisitions, and develop proprietary technologies to improve experiences for clients, staff, and linguists alike.

Florian and Esther welcome Slator’s Anna Wyndham and Alex Edwards to SlatorPod to explain the rationale behind the new industry framework introduced in the Slator 2025 Language Industry Market Report.

Drawing from the flagship report and echoing the buzz of SlatorCon London, the team explains why the traditional labels, Language Service Providers (LSPs) and Translation Management Systems (TMSs), no longer capture the scope and complexity of the evolving market. Instead, Slator has introduced two new terms: Language Solutions Integrators (LSIs) and Language Technology Platforms (LTPs).

Anna defines LTPs as pure-play technology providers that develop language tools, applications, orchestration platforms, and AI models. LSIs, she explains, are organizations whose core offering is to deliver fit-for-purpose multilingual content solutions by integrating language technology and AI with human experts as part of a fully managed solution.

Esther confirms early advisory adoption of the terms, noting investor interest in clearer tech-service distinctions. Alex adds that automatic dubbing startups tend to fit LTPs better than LSPs, as they often operate self-serve AI platforms.

Anna clarifies that big tech players like OpenAI and Google are excluded from the market sizing as they are foundational enablers, not language-focused businesses. The team also discusses why the term “AI” was excluded from the new categories as it may become as ubiquitous as “Cloud”.

To close, Anna points out that LSIs currently capture the bigger portion of the total addressable market (TAM). The team sees a strong demand for expert-in-the-loop services and growing LTP–LSI partnerships.

In a SlatorPod first, four guests — Scott Cooper, CEO of Language Services Associates (LSA), Pablo Tercero, COO of LSA, Jerry Song, CEO of Lingolet, and Edward Varela, VP of Business Development at Lingolet — shared their insights into a newly forged partnership.

Scott explained that the decision to partner instead of building an in-house AI solution allowed LSA to stay focused on its core strengths while leveraging Lingolet’s technical edge. The alliance goes beyond a vendor agreement, with LSA taking an equity stake in Lingolet, anchoring the relationship in shared strategy and long-term commitment. 

For Jerry and Edward, the partnership is a natural progression, rooted in their deep knowledge of both software development and the language services market.

In a standout use case, the two companies helped a Major League Soccer team enable multilingual, AI-assisted communication in coaching sessions — bridging language gaps in real time for international players.

On the tech front, Jerry and Edward explained that Lingolet doesn’t build large language models from scratch but instead acts as an orchestrator, offering clients the ability to customize and deploy AI tools within secure, dedicated environments. 

Pablo underscored LSA's HIPAA and PCI compliance, as security, compliance, and privacy are essential pillars, especially in regulated sectors like healthcare.

Looking ahead, LSA and Lingolet are preparing to offer hybrid solutions where AI can bridge the gap during interpreter unavailability, even for brief moments, potentially transforming session wait times. 

Their roadmap includes broader language coverage, deeper integration with healthcare platforms, and innovative pricing models that will reshape how interpretation services are sold and scaled.

Rong Yan, CTO of HeyGen, joins SlatorPod to recount the company’s transformation from a Metaverse-focused startup to leading the emerging field of AI video generation.

Rong recounts HeyGen’s beginnings and the pivot to its current avatar model, which saw ARR go from zero to USD 1m within six months.

Rong attributes HeyGen’s success to its emphasis on three key elements: quality, consistency, and controllability. The company’s newest model, Avatar IV, enables full-body video generation with natural gestures, synchronized audio, and emotion to speech.

While some of the platform’s growth has been viral, Rong believes sustained success comes from building something users truly value, with a focus on pushing video quality from 70% to 95%.

The platform extends beyond avatars, offering translation, voice cloning, and real-time interactivity. Its dynamic duration feature adjusts translated speech to fit original video timing, preserving realism. Rather than build everything from scratch, HeyGen integrates external models with its own orchestration and user data, optimizing output across languages and contexts.

Rong emphasized that HeyGen’s long-term vision is not entertainment or Hollywood, but helping everyday professionals, especially marketers and educators, who lack traditional video production skills.

Looking ahead, Rong sees video agents, tools that generate complete videos from simple prompts, as the next frontier, driving accessibility and transforming storytelling through AI.

Véronique Özkaya, Co-CEO of DATAmundi, returns to SlatorPod for round 2 to talk about the company’s strategic rebrand and how it is positioning itself as a key player in the data-for-AI space.

Véronique details her journey to leading DATAmundi, formerly known as Summa Linguae, where she now drives a strategic shift from traditional language services to AI-focused data enablement.

The Co-CEO explains that their LSP background makes them well-suited to offer fine-tuning services for AI, especially in multilingual and domain-specific contexts. However, she cautions that language expertise alone isn’t enough; deep tech infrastructure, data science capabilities, and the ability to quickly build custom workflows are also essential.

While many companies still rely on crowd-sourced, basic annotation, DATAmundi targets higher-complexity projects requiring domain experts and linguists. Véronique notes the market for data-for-AI is growing significantly faster than traditional LSP work and sees a second wave of demand from enterprises needing to adapt pre-trained models.

Véronique highlights data scarcity, hallucination, and bias as core AI challenges that DATAmundi tackles through technical solutions and expert guidance, helping enterprises as they face pressure to implement AI despite legacy systems and unclear strategies.

Looking ahead, DATAmundi plans to expand its consultative services through further acquisitions, focusing not on tech per se, but on organizations that deepen its expertise in data application and AI deployment.