Customize output with your translation with terminology, style and tone of voice
Customize one brand of machine translation, for example, Google AutoML or Microsoft Custom
The service includes data preparation and model customization for one selected brand of machine translation, typically, Google, Microsoft, ModernMT, or Globalese. An engineer then runs metrics and prepares samples for an evaluation by human linguists.
2-3 days
Best cost
The deliverables are a model and its score before and after fine-tuning.
€1,000 per language
Customize 5 different brands with your data and select the best after training.
2-5 weeks
Best linguistic quality
The competitive fine-tuning and evaluation guarantee the highest linguistic quality possible. The final deliverables are two models per language (the core and the backup), an evaluation report, and support for the model for one year.
€3,500 per language
Translate with OpenAI, Gemini, or another large language model with specific instructions
Large language models with prompts offer an easy way to customize machine translation output for terminology, tone of voice, formality and “do not translate” lists.
In this service, Custom.MT lexicographer develops a prompt for translation with specific instructions, and iteratively improves it to achieve higher quality.
3-5 days
Best flexibility
Unlike machine translation models trained with a large datasets, LLM output may be improved by rewriting prompts in minutes. Iterations may be run daily, allowing for a very flexible approach in customizing MT.
€1,000 per language
Lightweight customization
Adaptive translation denotes lightweight customization without computational investments.
Machine translation follows the style of human translation used to customize, but does not change the underlying model. ModernMT, Amazon ACT, Google Adaptive and some other brands support this approach.
Our service includes data preparation, automated evaluation and optimization of the output.
2-3 days
High flexibility and speed
Adaptive machine translation is best for scenarios with little training data available, for low-budget customization, and for machine translation programs with a high number of language combinations, where engineering for each language pair represents too great of an effort.
€700 per language
