ScreenRun's answer

ScreenRun is used by product teams and designers that want to share engaging videos of their apps or screens to colleagues, followers or clients.

Xinity's answer:

Regulated European enterprises where data sovereignty and compliance are non-negotiable: finance, healthcare, legal, public sector, etc. These are organizations currently unable to adopt cloud AI because doing so would breach sovereignty requirements.

ScreenRun's answer

Many screen recording apps are either too simple (just record your screen) or complex, and none of them offered a way to quickly create engaging screen recordings by placing a few points over images or video clips to zoom and engage the person watching.

Xinity's answer:

Xinity was founded in 2025 in Vienna by Alexander Zehetmaier (CEO) and Jonas Vander (CTO), who have built AI systems together for over a decade and studied AI at Radboud University in the Netherlands. They saw European companies forced into an impossible choice between powerful cloud AI that violated data sovereignty and open-source tools that were too complex to run without dedicated teams. Xinity was built to eliminate that tradeoff. On April 1, 2026, the company open-sourced its core Runtime under Apache License 2.0, making sovereign AI infrastructure freely available to developers across Europe. The mission: a compute-independent Europe.

ScreenRun's answer

Everything is web-based using modern browser APIs for video encoding and decoding. ScreenRun also uses AI from OpenAI to create automatic transcripts from your audio and video assets. But everything runs in your browser, no cloud involved for total privacy.

Xinity's answer:

Xinity is built on Bun and TypeScript. The core packages are an OpenAI-compatible API gateway, a model runtime daemon that runs on the GPU hardware, an operator CLI, a model registry (infoserver), and a SvelteKit admin dashboard. vLLM serves as the inference backend, with the data layer on Drizzle ORM, environment validation via Zod, and logging via Pino. It deploys through Docker Compose, with NixOS support. The proprietary R&D layer is Distributed Split Inference using a Mixture-of-Experts architecture, where expert sub-networks run across separate compute nodes and embedding encoding prevents any single node from reconstructing the output. The engine (gateway, daemon, CLI, infoserver, DB layer) is Apache 2.0; the dashboard is source-available under Elastic License 2.0.

ScreenRun's answer

Entirely web-based video creation too with zooms, webcam and automatic captions. And unlike many web-based products, ScreenRun runs entirely inside your web browser without sending any data to a server. This aspect is critical for total privacy in work contexts.

Xinity's answer:

Existing solutions force a binary choice: cloud APIs that violate data sovereignty requirements, or raw open-source tools that require dedicated MLOps teams to operate. Xinity eliminates this tradeoff. Its Scalable On-Premise LLM Management Automation System lets enterprises deploy production-grade generative AI on their own hardware, with OpenAI-compatible APIs, automated orchestration, and deployment in days rather than months. Existing applications can be redirected to on-premise inference with a single line of code. It is sovereign by architecture, not by contract.

ScreenRun's answer

ScreenRun is so simple to use! And because it's web-based, you can access it from anywhere, including iPhone or Android.

Xinity's answer:

Most competitors sell contractual sovereignty. EU-region hyperscaler offerings and European sovereign cloud operators still process your data on infrastructure they operate, so sovereignty rests on a jurisdiction clause, not physics. That clause does not override CLOUD Act reach, and your data still leaves your perimeter. Xinity is sovereign by architecture: the model runs on hardware inside your perimeter, so no data leaves and no third party can access it. Against raw open-source tooling, which needs a dedicated MLOps team, Xinity adds production-grade orchestration, one-line migration, and a fully auditable Apache 2.0 codebase.