Generative AI MT Engines represent a significant evolution in machine translation by leveraging large language models (LLMs) and advanced AI technologies. These engines redefine translation quality and adaptability, offering businesses unprecedented flexibility and scalability.
Advantages of generative translation engines
-
Contextual Understanding
Generative models can process entire documents or conversations, preserving nuanced meaning and adapting tone better than neural MT, which focuses on sentence-level translations.
-
Multi-functionality
Generative AI models can handle translation, content generation, summarization, and localization in a single framework, reducing the need for separate tools.
-
Rapid Adaptation
Generative AI can be adapted to a specific industry, use case and brand style without retraining, enabling faster updates to meet domain-specific translation needs.
-
Enhanced Multimodal Capabilities
Many generative models integrate multimodal learning, meaning they can process text, images and other inputs together to enhance translation quality in specific contexts.
-
Improved Learning from Data
Current generative AI models leverage unsupervised and self-supervised learning techniques, allowing them to learn from a broader range of unstructured data.
-
Scalability
Generative models often operate in cloud-based systems optimized for handling large-scale translation demands, which improves cost-effectiveness for high-volume applications.
Available engines
Phrase offers the following translation engines, powered by generative AI. These are fully managed engines, meaning:
-
Phrase bills their usage in MTUs. Customers do not require provider-specific credentials
-
Glossary support and tag handling is managed by Phrase on behalf of customers.
-
These engines are available across the Phrase Platform (in TMS, Phrase Portal, API, etc.)
By combining the creative adaptability of Generative AI with brand-specific precision, Next GenMT bridges the gap between traditional translation systems and state-of-the-art AI capabilities. It empowers businesses to achieve faster, more reliable, and on-brand localization outcomes, setting a new standard for machine translation in global markets.
Is powered by a highly specialized, in-house LLM technology built for and uniquely tuned for language. The specialized LLMs ensure more human-like, accurate translations with significantly fewer post-editing requirements, translating to dramatic productivity gains.
DeepL Next Gen is especially recommended for use cases requiring highest quality and security standards such as legal translations, banking and financial data.
DeepL Next Gen is particularly strong in working with the larger European and Asian languages.
Phrase customers are able to configure DeepL to automatically switch between the available model with the highest quality or highest latency or to fall back to the earlier model if the language pair is not yet supported.
DeepL Next Gen offers Enterprise-Grade Security and meets the highest standards with ISO 27001, GDPR compliance, and SOC 2 Type 2 certification. Additionally DeepL commits that customer data is never used to train models.
The Widn.AI engine is a GenAI engine that is powered by Unbabel’s award-winning TowerLLM model which often outperforms other Gen AI models such as GPT-4.
The Widn.AI engine is integrated across the Phrase Suite as a fully managed engine and accessible for Phrase customers directly through Phrase without the need for a Widn or Unbabel subscription.
Widn is particularly strong for Customer Support use cases.
Widn predominantly supports European languages as well as others.
Wind is GDPR and CCPA compliant, and ISO 27001 and ISO 9001 certified. Customer data is anonymised and de-personalized.