Can Tilde.ai become the solution for automated translation in TYPO3?

If you're making a strategic decision about language control in the enterprise CMS space, don't underestimate Tilde, which is filling a niche that the big US players have neglected.

Open source and flexibility: the end of one-size-fits-all AI

TildeOpen-30b is an open-source base model (not a finished chatbot) and therefore ideal for corporations.

• Adaptability: The model can be tailored to the specific jargon and terminology.
• Unlike closed systems, we can integrate Tilde deeply into TYPO3 backend processes.
• Local hosting or private cloud: Deploy the model where the data is.
• A distinctive USP of TildeOpen is the targeted filtering of the training data against disinformation.
• Technological independence from the pricing policies or availability of California tech giants.

Standard AI models (like GPT-4) are generalists. They know a lot about everything, but often not enough about the specifics. In highly regulated or technically complex industries, “almost correct” terms lead to misunderstandings, legal risks, or reputational damage.

By accessing the model weights of TildeOpen-30b, so-called adapter layers (LoRA/QLoRA) can be trained. The model retains its world knowledge, but learns the specific domain language as a new “primary language”.

1. Manufacturing & Engineering (Mechanical Engineering)
Technical terms like “axial play”, “cutting rate” or internal part numbers are often foreign to standard AI.

The Tilde model learns the technical documentation of your system. In TYPO3, this means that editors only enter basic data, and the AI creates precise product data sheets or maintenance instructions that are technically 100% correct in all target languages at the same time.

2. Healthcare & Life Sciences (Pharma/Medicine)
The AI ensures that patient portals use medical nomenclature (active ingredient names) and comply with strict regulatory requirements to avoid misleading medical advice ("hallucinations").

3. Finance, Legal & Insurance
A misplaced comma or an imprecise term in legal terminology or an imprecise interpretation of clauses can be costly. To ensure that the automated creation of legally secure FAQ areas or summaries of annual reports works directly in the CMS, the model can be fed with contract archives and current legislations. This way, Tilde understands the difference between "liability" and "transfer of risk" in every context.

4. Public Sector & Government
Authorities, research centres or universities need to make complex matters understandable for all citizens. TildeOpen-30b can be trained specifically to translate official German into easy language. As the model natively understands European language structures, this works much better in European languages than with US models.

PLUS Corporate Identity & Brand Voice
Beyond the technical terms, a translation should fit the company’s “personality”:

• You (informal) or you (formal)?
• Innovative-progressive or conservative-safe?
• Short, concise sentences or detailed explanations?

Sharpening up on best-practice content (blog posts, white papers, press releases) and thus training to become the “perfect editor”. This not only saves time in the TYPO3 workflow when creating text but also ensures consistent brand management across language and country borders.

In many companies, AI support ends with the text field in the CMS. But the reality in the editorial department is different: Product managers upload technical manuals as PDFs, marketing teams provide presentations (PPTX), and the graphics department delivers banners with fixed text.

The vision: Asset-first workflows in TYPO3

We want to integrate Tilde AI deeply into the TYPO3 File Abstraction Layer (FAL). So that the CMS no longer just manages files but also “understands” their content.

The planned workflow is as simple as possible:

1. Upload: An editor uploads a document (PDF, DOCX, PPTX) or an image (JPG, PNG).
2. AI analysis (OCR): Tilde recognizes text in images and documents.
3. Contextual translation: Tilde translates the text, taking into account the terminology of the field.
4. Layout Reconstruction: The AI puts the translated text back exactly where the original was.
5. Review & Go: The editor checks the result directly in the TYPO3 preview and releases it.

The technological basis and vision are finalized: With TildeOpen, we are opening up the fully automated, context-sensitive translation world – from the simple news article to the complex, laid-out product manual.

But the best AI is only perfected by real enterprise challenges: We are currently in the pilot phase of integration and are looking for strategic partners who want to define the standard for AI in the TYPO3 ecosystem together with us.

Now become part of that vision:

• Tailored solution: Early adopters can directly influence the prioritisation of features
• Competitive advantage: Reduce editorial and localization costs by automating asset translation
• Digital independence from US giants through a sovereign solution that puts data security at the center

We are looking for organisations with a high degree of internationalisation, complex document workflows, or a need for highly specialised terminology who do not want to compromise on quality.

Do you have a specific scenario, such as a large library of PDFs that needs to be available in ten languages, or a highly specific glossary that has driven every AI system to despair?

Book a strategy meeting