Translating technical documentation – reliable and consistent 

Yes, but not for every type of content. For internal emails, quickly understanding foreign-language texts or content with a low risk of error, a machine or AI-assisted translation may be sufficient.

Higher requir­em­ents apply to technical docum­ent­ation, safety-relevant information and public content. Errors here can lead to misund­erstand­ings, incorrect operation, complaints or liability risks. Specialist terminology, target market, cultural part­ic­ul­ar­ities and formal requir­em­ents also need to be taken into account.

A specialist translation does not always have to be produced entirely manually. A combination of machine translation and prof­ess­ional post-editing is often econ­om­ic­ally sensible. Our native-speaker specialist translators review and revise the machine output according to the required quality level.

Whether a specialist translation is necessary does therefore not only depend on the source text. The decisive factors are the intended use, the target audience and the cons­eq­uences of any possible translation error. The higher the risk and quality requir­em­ents, the more important a prof­ess­ional review remains.

Machine translation refers to systems spec­if­ic­ally developed and trained for the automatic translation of texts. Modern solutions already work with artificial intell­ig­ence. They usually translate reliably, quickly and cost-effectively, especially for clearly worded and recurring specialised content.

AI translation generally means generative language models nowadays. In addition to the actual translation, these can also take additional instruct­ions into account, such as a particular linguistic style, a target audience or the context of a text. As a result, translat­ions often sound more natural and can be adapted more flexibly to different requir­em­ents.

However, the boundaries are fluid. Machine translation is also a form of AI. The decisive difference lies less in the concept than in the model used, the data basis and the integration of company terminology and existing translat­ions.

For companies, it is therefore above all relevant whether the solution protects conf­id­ential content, reliably uses approved technical terms and achieves the required quality for the respective purpose. For publication-ready or safety-relevant content, the result should be prof­ess­ion­ally reviewed, regardless of the technology used.

At kothes, we charge for specialist translat­ions based on the number of words in the source text. The price per word depends above all on the combination of languages. Frequently requested language comb­in­at­ions are generally cheaper than rare languages. A translation from German into English is, for example, around 0.16 euros per word.

However, this price does not autom­at­ic­ally apply to every word. For each customer, we create a translation memory in which already approved translat­ions are stored. Repetitions and similar text passages can be reused for later jobs and processed at a reduced price. As a result, translation costs decrease over time.

That is why the pure per-word rate from the price list is often less meaningful than the actual average price charged. For many of our customers, this real per-word price is in the low single-digit cent range thanks to repetitions and existing translat­ions, often around 0.06 to 0.07 euros per word.

Additional costs may arise for DTP, graphic translat­ions, terminology work, special file formats or very short processing times. For reliable costing, we therefore first analyse the source files and compare them with the existing translation memory. This allows you to see which content actually needs to be translated from scratch and what price you will actually pay before placing the order.

At kothes, a machine translation with prof­ess­ional post-editing is generally about 30 percent cheaper than a pure specialist translation done by humans. The actual savings potential depends on how good the machine result is and how extensively it needs to be revised.

For AI translat­ions, the costs consist of the platform fee and the usage-based character or token consumption. If you convert these costs to the translated word count, pure translation costs of about 0.01 euros per word are possible.

However, this price is not directly comparable to a publication-ready specialist translation. It applies to automated translation without prof­ess­ional review. For internal emails, pres­ent­at­ions or quick text compreh­ension, this quality may be sufficient. For technical docum­ent­ation, safety-relevant information and externally published content, you should addit­ion­ally check whether prof­ess­ional post-editing or a specialist translation is required.