Streamlining ‘Talking Heads’ Video Editing with ChatGPT

Colm Moore
3 min readJul 3, 2024

--

In our quest to enhance our video editing workflows and embrace the power of AI at Tiny Ark, we turned to ChatGPT and other large language models to assist in creating scripts for our corporate videos. Traditionally, each corporate video begins with a substantial amount of interview footage. This footage is transcribed, usually into a Google Doc, and then reviewed by the director or client. They highlight specific quotes and assemble these into a “paper edit,” a laborious process of manually reading transcripts, marking key points, and reassembling them into a cohesive story from potentially hours of interview footage, aiming for a final product of just three to five minutes.

The Challenge of Manual Transcription

The initial manual process involved:
1. Transcription: Getting interview footage transcribed.
2. Highlighting Quotes: Reviewing the transcription to highlight key quotes.
3. Creating a Paper Edit: Manually pulling quotes and reassembling them into a narrative.

This was time-consuming and prone to human error, making it a prime candidate for automation with AI.

Introducing AI into the Workflow

We began experimenting with uploading these transcripts to large language models like ChatGPT. These models excel at processing and generating language, but our early attempts were hit or miss. For instance, when prompted to create a three-minute story focusing on specific topics, ChatGPT would sometimes fabricate quotes or fail to maintain accurate timestamps, which are crucial for video editors.

Refining the Prompt

Through trial and error, we honed our prompt to be very specific. For example, we instructed ChatGPT to:
- Use Exact Snippets: Pull exact quotes from the transcript.
- Maintain Timestamps: Include the corresponding timestamps from the original transcript.

These refinements required detailed instructions and continual adjustments.

Custom GPTs: A Breakthrough

The introduction of custom GPTs was a turning point. We developed a specialised model named “Tiny Ark Video Editor GPT.” This custom model allowed us to upload a transcript and provide basic instructions, such as:

“Create a three-minute video based on this transcript using only quotes from Alex, focusing on finance.”

Within seconds, ChatGPT would deliver a list of relevant quotes, complete with a suggested story arc, providing a start, middle, and end. This capability dramatically reduced the time spent on initial edits from days to mere minutes.

ChatGPT hard at work

Benefits and Impact

The key advantages we observed included:
- Speed: Generating the first version of the interview edit within 15 to 20 minutes.
- Efficiency: Allowing editors to bypass the tedious groundwork and focus on creative tasks.

By automating the initial assembly of interview footage into a preliminary edit, ChatGPT freed up significant time for our editors, enabling them to concentrate on more creative aspects of video production. This innovation has not only accelerated our workflow but also enhanced the quality and coherence of our final products due to the editors gaining back a lot of time.

We’d consider the GPT version of the script 75% good. It’s a good starting point but often needs a bit of work to get it to reach our standards.

Conclusion

Using ChatGPT as a video script editor has revolutionised our process, demonstrating the potential of AI to transform traditional workflows. The ability to quickly generate structured narratives from raw transcripts means our editors can now deliver high-quality videos faster and more efficiently than ever before.

Embracing AI tools like ChatGPT can significantly enhance productivity and creativity in video editing, making it an invaluable asset in modern content creation.

--

--

Colm Moore

Head of Post Production at Tiny Ark, Dublin. Past lives include VFX Artist, Video Editor and Photographer. www.colmmoore.com //