AI for video editing: improvement of videos

Some clients contact us just for editing services. They’ve recorded video material by themselves with sub-par quality, sometimes on mobile phones or through platforms like Zoom. Their expectation is almost always the same: To convert these less-than-stellar footage into engaging, high-quality videos.

AI has become a vital cog in our video editing process. We leverage AI to boost our efficiency, speed, and quality of our productions. However, our video editing work remains predominantly manual and AI serves as a supportive tool.

Deepfake technology

We recently faced a unique challenge. When a client’s webinar presenter forgot to turn on their camera for the first 15 minutes of the virtual event, and we were faced with fixing this footage in post-production and showing him. To solve this, we utilized generative AI to create a deepfake of the presenter, which was integrated seamlessly into the webinar footage, giving the impression that the camera was on throughout.

AI-assisted audio enhancement

It is fascinating how far AI has advanced in refining audio quality. With AI, we can transform any recording into studio-like quality sound. Regardless of external noises such as passing cars or building sounds, we are confident that our AI tools can filter these out to give a clean and crisp audio output. The utilization of AI to improve audio quality has become a fundamental aspect of our video production – and no video leaves our production house without it.

Automated subtitle generation

In 2022, we started creating short videos to showcase our thought leadership, which we publish on LinkedIn. To cater to a wider audience, we decided to include subtitles. We developed a process where we run AI over the final video to generate subtitles, which are then reviewed and corrected by a native speaker. This is followed by a final review by an AI spelling correction software. The corrected subtitles are then integrated into the video for customer review and feedback.

For international audiences, we provide translated subtitles, for instance, translating German video content into English subtitles. Here, we employ ChatGPT to translate SRT files (standard for subtitle management) while maintaining the timestamps.

Improving visual aesthetics

The introduction of the “Generate Fill” feature in Adobe Photoshop revolutionized our editing process. This feature allows us to rectify distracting elements in the background of video shots. For instance, during a project where we followed a tourist group around Switzerland, we used this feature to remove non-group individuals who were accidentally captured in the footage.

Custom background music

Background music plays a crucial role in setting the tone of a video. We use AI services that generate music based on predefined parameters and the intensity of the video. These AI-powered services create a unique soundtrack tailored specifically for each video.

However, it’s worth mentioning that not all our experiments with AI in video editing have yielded satisfactory results.

Rotoscoping, or isolating a foreground element from the background, is still a challenge for AI. The results, especially in videos, are currently not up to our quality standards. We also tried using AI to link camera movements to a person’s movements, but the results fell short of our expectations. Furthermore, AI-generated voice-overs, particularly when mixed with human voices, did not deliver the desired quality and landed directly in the uncanny valley.

In conclusion, while AI presents some limitations, its potential in transforming video editing workflows is undeniable. It’s a tool that improves our efficiency, enables faster production, and raises the bar for video quality. As we continue to explore new AI capabilities and refine our processes, our aim remains to deliver engaging and high-quality content to our clients.