#ad — In partnership with @Descript
Get 50% off your Descript subscription: https://linke.to/descript (Affiliate link for which I may earn a commission if you subscribe)
My Descript course: https://www.descriptmastery.com/course
Can ChatGPT actually edit a video end-to-end through Descript?
I filmed a quick, unscripted video while walking my dog, then asked ChatGPT to edit it through its Descript connection—arranging clips, cutting pauses, removing filler, and adding text, B-roll, and effects.
Here’s what worked, what went wrong, and where a human editor can still contribute.
1. The AI first pass works incredibly well
ChatGPT transcribed and assembled the raw clips, swapped scenes on request, cut pauses of 2.5 seconds or more, removed a retake, and applied Studio Sound at 100% to reduce wind noise. It also added a quote-layout hook, a red arrow call-out, and even a crying-face emoji I never asked for.
It didn’t get everything right. It missed the intended clip order and chose some....interesting B-roll including a timeline from a different editor. All of which I corrected with follow-up prompts.
2. Hybrid editing is the real workflow
Some things are faster by hand. Resizing text directly beats describing the change in a prompt. My approach: let AI handle the heavy lifting and the tasks you dislike, then make manual adjustments where needed.
I also share my review process: watch the full cut, take notes, and send one consolidated prompt with the fixes.
3. Cost, model choice, and why I used ChatGPT over Underlord
During this test, ChatGPT’s usage indicator showed that I’d used about 7% of my five-hour allowance on the $20/month plan. The edit also consumed 216 Descript AI credits—roughly a quarter of the Creator plan’s 800 credits.
I estimate a similar edit without the extra experimentation could use around 100 Descript credits, allowing roughly eight comparable videos per month. Actual usage will depend on the edits you request.
Why use ChatGPT over Underlord? Model choice and existing context. ChatGPT already knew about my audience, my brand, and even my dog, Elvis—it suggested featuring him in the thumbnails without being asked. It also generated 25 title-and-thumbnail concepts that helped me choose the options for this video’s thumbnail test.
00:00 Can AI Edit This Video
01:14 Setup and Goals in the Field
03:01 Importing Clips Into Descript
03:47 Connecting ChatGPT to Project
04:45 First Instructions and Clip Ordering
06:42 Building the Editing Prompt
12:40 Results Review and Fixes
15:54 Brightness Font Music and B-Roll Pass
19:00 Note Taking and Iteration Workflow
22:01 Accidental Send and Final Tweaks
24:15 Hybrid Editing Takeaways
25:36 Costs Credits and Model Choices
27:29 ChatGPT vs Underlord - Why? Just why?
29:33 ChatGPT Made the thumbnails for this video
Get 50% off your Descript subscription: https://linke.to/descript (Affiliate link for which I may earn a commission if you subscribe)
My Descript course: https://www.descriptmastery.com/course
Can ChatGPT actually edit a video end-to-end through Descript?
I filmed a quick, unscripted video while walking my dog, then asked ChatGPT to edit it through its Descript connection—arranging clips, cutting pauses, removing filler, and adding text, B-roll, and effects.
Here’s what worked, what went wrong, and where a human editor can still contribute.
1. The AI first pass works incredibly well
ChatGPT transcribed and assembled the raw clips, swapped scenes on request, cut pauses of 2.5 seconds or more, removed a retake, and applied Studio Sound at 100% to reduce wind noise. It also added a quote-layout hook, a red arrow call-out, and even a crying-face emoji I never asked for.
It didn’t get everything right. It missed the intended clip order and chose some....interesting B-roll including a timeline from a different editor. All of which I corrected with follow-up prompts.
2. Hybrid editing is the real workflow
Some things are faster by hand. Resizing text directly beats describing the change in a prompt. My approach: let AI handle the heavy lifting and the tasks you dislike, then make manual adjustments where needed.
I also share my review process: watch the full cut, take notes, and send one consolidated prompt with the fixes.
3. Cost, model choice, and why I used ChatGPT over Underlord
During this test, ChatGPT’s usage indicator showed that I’d used about 7% of my five-hour allowance on the $20/month plan. The edit also consumed 216 Descript AI credits—roughly a quarter of the Creator plan’s 800 credits.
I estimate a similar edit without the extra experimentation could use around 100 Descript credits, allowing roughly eight comparable videos per month. Actual usage will depend on the edits you request.
Why use ChatGPT over Underlord? Model choice and existing context. ChatGPT already knew about my audience, my brand, and even my dog, Elvis—it suggested featuring him in the thumbnails without being asked. It also generated 25 title-and-thumbnail concepts that helped me choose the options for this video’s thumbnail test.
00:00 Can AI Edit This Video
01:14 Setup and Goals in the Field
03:01 Importing Clips Into Descript
03:47 Connecting ChatGPT to Project
04:45 First Instructions and Clip Ordering
06:42 Building the Editing Prompt
12:40 Results Review and Fixes
15:54 Brightness Font Music and B-Roll Pass
19:00 Note Taking and Iteration Workflow
22:01 Accidental Send and Final Tweaks
24:15 Hybrid Editing Takeaways
25:36 Costs Credits and Model Choices
27:29 ChatGPT vs Underlord - Why? Just why?
29:33 ChatGPT Made the thumbnails for this video
- Категория
- Бесплатные курсы
Комментариев нет.
