Editing a YouTube video is only half the job. The thumbnail still has to communicate the topic in one image, make the face recognisable, convey the right emotion, and compete with dozens of other images in the feed. Sacha Dumay got stuck there while running his own channel. Describing himself as inexperienced in marketing and design but stronger in backend and AI development, he built AIThumbnail to help non-designers make YouTube thumbnails.[1] Interview[2]
AIThumbnail starts by asking for a video idea. Its website says no design experience is needed and a thumbnail is generated within 30 seconds. Enter a video's topic, for example, and the AI suggests improvements to context, colours, text and elements, rather than simply producing an image.Website[3] Users can also upload a photo of themselves each time for the AI to incorporate; the website says no special setup or training is needed.
Image: official product page[3], obtained on 2026-11-18; the interface is a website demonstration.
Replica Mode is what distinguishes it from “another AI image generator.” Upload a thumbnail you like, have the AI immediately reproduce its style, then add your own face and change the text.[3]
That feature deserves a pause. People making thumbnails often do not want something unprecedented. They have just seen an arresting image and want that feeling. Making “imitation” an explicit feature is more candid than asking users to describe a style indirectly through prompts.
To understand why the product took this form, go back to version one. Sacha gave V1 to a few YouTuber friends. He described their feedback as good but brutal, and ultimately rebuilt everything. After finishing V2, he said his motivation actually fell. OpenAI then released image-1; he upgraded the app that day and subsequently launched on Product Hunt.[2]
The changelog[4] dates both image-1 integration and the Product Hunt launch to April 26. The first MVP launched on April 3, integrating Flux Pro/Ultra and AI training for generating real faces. Downloads in 16:9 arrived only on May 18, with AI automatically optimising composition for that ratio. For a tool primarily making YouTube thumbnails, 16:9 did not arrive particularly early.
In late April 2025, with no preparation or paid marketing, he reached Product Hunt's daily Top 8, securing homepage exposure and traffic. A $49 Stripe notification followed, then a few more, and he decided to commit fully.[2]
The website's pricing cards list Starter at $9 a month with 80 monthly credits and Pro at $19 with 160. Each image generation or edit uses 1 credit. There is no free trial. The refund policy[5] instead offers a 3-day trial for US$1, with refunds for dissatisfaction within 3 days of purchase, citing high AI training and usage costs. The homepage FAQ also says all plans include annual billing with significant savings, but supplies neither actual annual prices nor charging intervals, so an annual total cannot be calculated from that claim.[3]
On revenue, Sacha reported in the July 18, 2025 interview that when he passed $500, he had 3,000 unique visitors and slightly above 10 paying users.Interview[2] He did not call it MRR or give the measurement period for those figures, so neither a conversion rate nor revenue per user can be calculated. Subscriptions are the main business model, while the earliest customers bought a $49 lifetime deal with credit limits. Cost amounts, profit, retention and customer distribution were not disclosed. Profit therefore remains unknown.
He said in that interview that steady traffic was his challenge. Attempts included building in public on X and YouTube, where he said the audience was not targeted enough; posting on Medium every few weeks; making YouTube Shorts and TikTok reels; investing in SEO; posting two free tools on Hacker News to speed up Google indexing; and running an automated blog publishing daily.Interview[2] These actions show experimentation across channels, not which channel actually produced revenue.
Someone else has packaged a similar need as an API. An Actor on Apify[6] generates thumbnails from a title or story, adds an expressive AI close-up face and an all-caps hook, and outputs 9:16, 16:9 and 1:1 PNGs using the user's own OpenAI key. The same page lists “from $144.00 / 1,000 thumbnails,” while its description says $150.00 per 1,000, or $0.15 each. The prices conflict. It is therefore only a candidate for the specific job of generating thumbnails in bulk from titles. The page does not list the full Replica Mode and single-person-photo workflow, and the page offers no like-for-like comparison.
What AIThumbnail sells is “skipping the designer step”: compressing finding someone, waiting for a draft and repeatedly changing text into an upload. Its most consequential design is not 30-second generation. It is letting users upload someone else's style and put their own face into it. Much of a thumbnail's persuasion comes from style and from “this person is actually in this video.” The product turns those two things into two upload boxes.
[2] www.fakemayo.com
[6] apify.com