We edited these videos
through ChatGPT.

The starting point

The expertise was already there.

Allegra Dental Center had recordings of its doctors answering the kinds of questions patients bring to a dental visit. The explanations were useful. The footage already contained edited camera changes. What it needed was a visual presentation that reflected the care behind those explanations.

I worked with Dr. Nana Dickson and the Allegra team to develop a consistent treatment across 17 videos: considered settings, closer framing, supporting visuals, readable graphics, and covers made from professional photography.

The editing directions happened conversationally. We reviewed a sample, described what needed to change, and checked the next version. The examples here show matching excerpts from the starting recordings and the latest review edits.

Begin with one sample, not seventeen.

A representative excerpt makes decisions concrete. Include a camera change, hair or hand movement, and a moment where a graphic might help. Keep the original file and ask for a separate review export.

A room needs to follow the camera.

Replacing a backdrop is more involved than placing a room image behind a person. The foreground needs clean edges around hair, clothing, hands, and the chair. The room needs a believable scale and light direction. If the recording cuts to a different view, the replacement background needs a corresponding perspective.

In this project, tighter cropping alone would have removed part of the presenter. We kept the person intact and worked on the background instead, then checked the result in motion.

Give every graphic a reason to exist.

Our early graphics were too close to ordinary presentation slides. We moved toward a more distinctive visual language: dimensional imagery, deliberate motion, fewer words, and stronger typography. The useful question was always: what does this help the viewer understand?

Supporting imagery must match the subject. A beautiful but unrelated image creates confusion. Text must be large enough to read in the actual website player, not just on an editing monitor. And a collection needs enough variety that viewers are not seeing the same visual repeatedly.

The most important refinement was restraint.

A stronger effect did not always make a better video. Treatments that looked promising in a short sample could feel too processed when the presenter moved or the camera changed. We reduced the correction for Dr. Dickson and Dr. Henderson and kept Dr. Floyd's separately reviewed treatment.

We also explored expression changes, but they did not preserve the natural delivery we wanted. The edits shown here keep the doctors' original facial movement and speech. The lesson is simple: improve the presentation while protecting what makes the person credible.

“3 seconds is too short to assess. Can you create a 20 second clip.”

Actual direction from this project

That feedback shaped this article too. Each comparison gives you 20 seconds to judge movement, transitions, and consistency, rather than asking you to trust a flattering still.

Try it with your own footage

The prompt starts the work. Review makes it useful.

Start with a recording you have permission to use, your brand references, and a clear description of the audience. Provide access to the files in an environment that can run the required editing tools. Ask it to identify missing capabilities before promising an output.

The workflow here involved foreground masking, matched background views, restrained color treatment, generated imagery, coded graphics, and encoded video exports. Some steps needed custom processing and several rounds of feedback. These prompts are reusable starting directions, not a promise that every account or tool setup can carry out every step.

1. Choose one representative excerpt. Test real movement and both camera angles before editing the collection.
2. Set the non-negotiables. Keep the original performance, dialogue, timing, and source files intact.
3. Review the exported video. Watch at normal speed, listen to the audio, and check text on a phone.
4. Let the practice make the final call. The doctors review their appearance and clinical explanations before release.

Our recorded checks covered the 17-video review collection at 1920 × 1080, including playback and preservation of audio and timing. That verifies the deliverables, not a marketing result. We are not claiming that these edits have increased appointments or views.