Skip to content
Dominik Lehner Solutions
Menu

Project in detail

BrandPulse

Content and publishing system with research, evidence checks, design, voice generation, and explicit approvals.

  • Service AI Content Operations / Managed Software
  • Tools / technologies Research agents, source checks, rendering, TTS, Telegram approvals, publishing
  • Outcome / benefit A reviewable production state from initial topic to approved publication.

BrandPulse is my development project for AI-assisted video production. It brings topic research, scripts, visuals, narration and publishing together in one workflow. I use it for my own YouTube channels and develop it further through that practical work.

← DLS Lab

Why I am building BrandPulse

A video involves more than a script. Sources, scenes, narration, subtitles and publishing all need to fit together. BrandPulse grew from the question of how to connect these steps while keeping editorial decisions and the resulting material open to review.

The dashboard keeps brands, episodes and production stages together. Each channel has its own topics, language and visual direction. German and English editions share the underlying workflow, with separate scripts and narration.

From a topic to a video

  1. Research: collect topics and sources, compare evidence and identify a focus for the episode.
  2. Script and review: develop the story, check claims against sources and revise the text for spoken delivery.
  3. Scenes and assets: plan visuals and assemble images, animation or browser and application recordings to suit the format.
  4. Voice and rendering: generate narration with a local Qwen3-TTS service and combine voice, scenes and subtitles into the video.
  5. Approval and publishing: submit the result for human review through Telegram. Publishing follows explicit approval.

What I am working on

The project combines a TypeScript application, a Next.js dashboard, a background worker and Remotion for video rendering. Specialised agents support research, writing and review. Production stages and their assets remain attached to the episode so that a failed step can be investigated and revised.

Development currently focuses on reliable recordings from real applications, natural German and English narration, and synchronisation between speech and what appears on screen. Tutorials, explanatory videos and music clips place different demands on the workflow. The examples below show individual published results; BrandPulse continues to evolve.

Project demo

BrandPulse — AI Content Operations, End to End · A 2:58 overview recorded for OpenAI Build Week, showing the workflow from research to approval and publishing.

YouTube channels

Example videos