This project uses Adobe Premiere's auto-generated transcripts and ChatGPT to create timecoded scripts for quick video rough cuts.
An explanation of how the basic technique works is here: [https://www.moreyummy.com/ai-video-editing]
Also on YouTube here: [https://youtu.be/1S2o2clDvRU]
This folder contains working examples to demonstrate how the AI Rough Cut Editor workflow operates using Adobe Premiere transcripts and ChatGPT prompts.
This is a sample transcript exported from Adobe Premiere Pro using the Text-Based Editing workspace. It includes speaker labels and timecodes, formatted exactly as Premiere outputs them.
This file contains a structured prompt designed to guide ChatGPT (or another LLM) in transforming the transcript into a themed, timecoded script. The prompt is focused on extracting moments from Churchill’s speech that reflect the theme of resilience—even though the word itself is never spoken.
- Open ChatGPT or another LLM interface.
- Copy the contents of the prompt into the chat.
- Upload the full contents of the transcript when prompted.
- Let the model generate a new rough cut script based on the theme.
- Use the resulting timecodes to quickly assemble a timeline in Premiere Pro.
This demo shows how an LLM can understand themes and narrative context, going beyond literal keyword searches. It's a powerful example of how AI can save editors hours of manual work.
Want to contribute a new example? Just add your own transcript and prompt using this same structure.
The goal of this project is to integrate these tools into a professional editing NLE such as Adobe Premiere.
If you're familiar with Premiere's XML structure, scripting, or AI workflows—join us! More soon.