Weekly YouTuber makes thumbnails in ChatGPT from a detailed style brief, then fixes them in Canva

What they did

They wrote ChatGPT a very detailed description of how their thumbnails should look, down to the color palette, so they don't read as obviously AI-made. They chose ChatGPT over Claude and Gemini because they find it better at this task, and sometimes move the result into a Canva template for small fixes. After filming and editing, they get a transcript from Whisper Transcription and give it to a Claude project with formatting rules to get title options, a description, timestamps and tags capped at 500 characters. A Gemini notebook loaded with their video archive, view counts and idea list helps them pick which ideas are worth filming.

How it went

Thumbnails are the one task they've handed off entirely, and they've never hired a designer. Hand-timestamping a 17-minute video is no longer their job.

Worth knowing

• AI doesn't come up with their ideas, outlines, scripts, filming or editing. The post includes plugs for the author's live workshop and Notion template.
• Price check: ChatGPT: free plan; Go: US$8/month (checked Nov 1, 2026); Canva: free plan; Canva Pro: US$18/month, or US$12/month on a yearly plan (US$144/year) (checked Oct 26, 2026); Claude: free plan; Pro: US$20/month, or US$17/month on a yearly plan (US$200/year) (checked Nov 1, 2026); Google Gemini: free plan; Google AI Plus: US$4.99/month (checked Nov 1, 2026); Notion: free plan; Plus: US$12/month, or US$10/month on a yearly plan (US$120/year) (checked Oct 26, 2026); CapCut: free plan; Pro: US$19.99/month, or US$15/month on a yearly plan (US$179.99/year) (checked Oct 26, 2026).

Paying for ChatGPT? See cheaper ChatGPT alternatives.

Try it yourself

Paste this into a chatbot, then add your own details underneath.

Write a thumbnail style guide I can reuse: describe my face placement, background, color palette [hex codes], font style, maximum word count and what to avoid so it doesn't look AI-made. Then use it to make a 16:9 thumbnail for a video titled [title], with the text [3 to 4 words].

Original post

See Jenna Redfield (Red Thread Toolbox)'s original post on Substack → (Aug 13, 2026)