AI Agent for Multilingual Real Estate Video Dubbing & Automated EditingBeta

About the role
DeSH Real Estate GmbH is an international real estate brokerage company connecting German real estate with international buyers and investors.
In this Project Study, students will develop and validate a multimodal AI Agent that transforms raw Chinese-language property-tour videos into professionally edited English-language real estate marketing videos with German subtitles.
Our real estate agents regularly record complete property tours in Chinese. These videos combine spoken explanations with continuous walkthrough footage of living rooms, kitchens, bedrooms, bathrooms, gardens, balconies and other property features.
Today, converting this material into high-quality international marketing content requires substantial manual work. Editors need to transcribe the narration, remove repetitions and pauses, rewrite and translate the content, create a new voice-over, generate subtitles, select suitable visual sequences, match the spoken content to the corresponding rooms and property features, edit the footage and perform final quality control.
The goal of this project is to investigate how speech AI, multimodal AI, large language models and AI agents can automate this workflow end-to-end.

Concrete Use Case
The target workflow is:
Raw Chinese property-tour video
→ Chinese speech transcription
→ semantic understanding and content restructuring
→ natural English real-estate script
→ AI-generated English voice-over
→ German subtitle generation
→ video / room / scene understanding
→ matching narration with relevant footage
→ intelligent clip selection
→ AI-generated editing plan
→ automated first cut
→ quality control
→ editable 16:9 / 9:16 marketing video

A successful prototype should be able to take a several-minute raw Chinese property-tour video and automatically produce a professional first cut requiring only limited human corrections before publication.

The project covers

• Chinese speech-to-text and timestamp alignment
• detection and removal of pauses, repetitions and irrelevant speech
• semantic restructuring of spontaneous Chinese narration
• generation of natural English real-estate language rather than literal translation
• English text-to-speech / AI voice-over generation
• automatic German subtitle generation and synchronization
• room and property-feature recognition
• multimodal matching between narration and corresponding visual footage
• detection of blurred, unstable, repetitive or unsuitable scenes
• intelligent clip ranking and selection
• AI-based storyline and editing-plan generation
• creation of structured editing instructions including in/out points, shot order, duration, crop, speed, transitions and overlays
• programmatic execution of the editing plan
• automated technical, linguistic and editorial quality control
• editable video timelines
• export for horizontal 16:9 and vertical 9:16 formats

High-End Real Estate Editing Style
A further objective is to formalize a premium international real estate editing language inspired by leading luxury real estate marketing content.
Students should investigate how characteristics such as the following can be translated into explicit and configurable editing rules:

• strong opening hooks
• concise storytelling
• removal of unnecessary walking sequences
• close synchronization between narration and relevant B-roll
• alternation between wide shots and detail shots
• purposeful camera movements
• dynamic but controlled pacing
• predominantly clean cuts
• selective use of transitions
• music-aware editing and beat synchronization
• premium typography and overlays
• visually strong opening and closing shots