About Clarus News

Multilingual news platform with editorial selection, AI-assisted tools, and transparent multi-model architecture.

Our Mission: Editorial Selection with AI Support

The digital media landscape has changed the rules. In many cases, it's no longer newsrooms that decide which topics are visible, but algorithms. These prioritize content by engagement, not relevance. What works well gets reach – what's important can fade into the background.

Clarus News follows a different approach: human journalistic selection combined with AI-assisted tools. The goal is coverage that is reliable, understandable, and free from algorithmic bias.

Our Principles

1. Editorial Selection Instead of Engagement Optimization

The newsroom decides which topics are socially relevant – based on importance, impact, and context. Short-term trend curves or social media dynamics are not decisive, but journalistic criteria are.

2. Transparency: Multi-Model Approach

Unlike many platforms that obscure their AI usage, Clarus News transparently documents the use of various AI models. The concept is based on three principles:

• Specialization: Each model is used for its specific strengths
• Flexibility: Models can be switched at any time – no vendor lock-in
• Transparency: Disclosure of the technology used

Models are selected according to task, cost, quality, and data protection. AI supports editorial work but does not replace it.

3. Multilingualism

Content is prepared in German, French, and English. Translation and multilingual availability are core functions of the platform.

Audio Functions: From Podcast to Press Conference

The platform has been extended with audio processing functions. The developed audio pipeline enables:

• Transcription (Speech-to-Text API): Automatic audio-to-text conversion powered by Whisper v3. Fast and reliable conversion with automatic speaker detection (Speaker Diarization). Support for over 100 languages with minimal latency – 30 minutes of audio transcribed in under one minute. Cost-effective at just CHF 0.65 per 3 hours of speech.
• Swiss German processing: Reliable processing of regional dialects
• Translation: Cross-language audio processing
• Multilingualism: Transcription in DE, FR, EN

Typical Use Cases

• 📻 Podcast transcription: Automatic conversion of audio content to text
• 💬 Interview processing: Multilingual availability
• 🌐 Content distribution: Multilingual audio content for international audiences
• ♿ Accessibility: Automatic subtitles and accessibility

Data Protection and Digital Sovereignty

Depending on data sensitivity, different processing options can be chosen:

• Global optimization: International models for non-sensitive content
• European options: EU-based models for higher privacy requirements
• Swiss solutions: Locally hosted alternatives for maximum data protection compliance

The flexible architecture enables the use of privacy-friendly variants without restricting functionality.

Technical Framework

Clarus News is based on a custom-developed, modular architecture:

• Backend: C# .NET Core on Linux basis
• Database: PostgreSQL
• Model integration: Orchestrated multi-model architecture via Model Context Protocol (MCP)
• Audio pipeline: Standalone module for speech-to-text, translation, and real-time processing
• Frontend: .NET CMS with multilingual support

The modular design makes it possible to expand individual components independently or integrate them into other systems.

Service Option: Technology for Organizations

The developed technology can also be used as a service for other organizations:

• For media houses: Multilingual availability of audio content
• For podcasters: Automatic transcription and translation
• For institutions: Transcription of press conferences, parliamentary debates, or public events
• For authorities: Multilingual content distribution with privacy compliance

Frequently Asked Questions (FAQ)

❓ How does editorial selection work?

The newsroom selects topics according to journalistic criteria: importance, social relevance, and context are priorities. AI tools support source validation and fact-checking, but topic selection remains an editorial decision.

🤖 Which AI models are used?

Clarus News uses a multi-model approach with various language models dynamically selected according to requirements (cost, quality, privacy, speed). Details on our team page.

🗣️ How does Swiss German processing work?

The audio pipeline can transcribe Swiss German language variants and translate them into other languages. Linguistic complexity and regional variants are processed by specialized models.

🔒 What privacy options exist?

Depending on sensitivity, various processing options can be chosen – from global solutions to EU-based models to local Swiss alternatives. The modular architecture enables flexible compliance.

🔧 Can the technology be used as a service?

The developed infrastructure is modularly designed and can also be offered as a service for media, podcasters, and institutions. The flexible architecture allows adaptations for different requirement profiles.

Last updated: January 7, 2026