How To Reduce Design Costs By At Least $800 Per Month
First of all, the YTuong team would like to express our gratitude to our customers who have chosen Heyetsy among the myriad options available out there. This is not only a great motivation for the YTuong team to continue developing new features but also to enhance our tool’s capabilities to serve our customers better.
In this document, our team will share tips to help you succeed with Etsy using AI.
You will save a lot of time by using AI and can also reduce the financial burden of design personnel.
Okay,
Let’s get started!
Aligning on Product Selection Mindset
Based on extensive sales experience, I’ve identified two key factors that influence a product’s success.
- Product Keywords
- Product Design Templates
Keywords help customers find the product, while the design template determines whether the product will sell well or not.
Therefore, there are two indicators that influence our work — a product that sells well will definitely:
- Rank on the first page in a short time
- Have high performance metrics (views, conversion rates, like ratio, etc.)
To find well-performing designs, you’re likely already using Heyetsy with this document
Look for products that meet both criteria mentioned above.
Ideas for Re-design Implementation
Step 1: Create Prompts (convert images to text) using AI through these methods:
- Generate Prompts by uploading images to ChatGPT 4o (this version is more optimized than regular ChatGPT)
- Create Prompts using the Describe feature in Midjourney
- Generate Prompts through Playground.com
- …or using any image-to-text conversion tools you’re familiar with
The reason for using multiple tools is that each generates different types of prompts, resulting in varied image generation outputs. While I use Midjourney for image generation, its biggest drawback is that it often has errors in text generation, so you should combine it with other platforms like Ideogram for design generation
Step 2: Input the Prompt into MJ
Create Prompts automatically instead of writing them manually
We will use ChatGPT 4o to generate image Prompts
Usually, users purchase the ChatGPT Plus package for around $20
However, in this article, I’ll share with you a more affordable alternative using Sider.AI (contains YTuong’s Affiliate link). You only need to spend $6.7 per month to access ChatGPT 4o features instead of paying $20 for just ChatGPT Plus.
For example, in the screenshot below, I purchased the highest tier package for personal use, primarily for content creation and data analysis. Sider has many AI tools already integrated into their system.
Now it’s time to execute the product, converting images to Prompts
The above images were generated entirely using GPT-4o
The ideas were taken from several products with very good metrics on Etsy
This ensures that our product meets the criteria of being re-designed from winning designs, while significantly reducing the violation rate since the AI-generated designs are already different.
Another example: during my daily product research, I discovered a design related to the keyword Coquette
It’s a design featuring bows — all t-shirts with bow prints and related designs have good sales potential
So the question is, how can we design based on our niche?
For example, I discovered a design with a bow wrapped in electrical wires as shown in the image below, which has sold around 20 orders just in the past 24 hours.
The successful idea here now needs to be improved to look better.
Using the method above, I created 4 versions that are more beautiful than the original reference version
Tips for separating subjects from backgrounds — you can use the prompt below and adapt it flexibly according to your situation.
I’ve noticed that when generating images to prompts in MJ, there’s always a keyword that causes my designs to be surrounded by a background layer that I can’t separate if that background layer exists.
I made a small change to the prompt structure.
Replace the phrase “The image features” with “The vector t-shirt design of”, then add “The background is a white color, making the characters and text stand out.” at the end of the text above.
The image features a decorative bow with a checkered black-and-white pattern. The edges of the bow have a beaded trim, and it is wrapped with colorful Christmas string lights. The lights come in various colors, including red, blue, green, yellow, and pink, giving it a festive and cheerful holiday look. The black background highlights the details of the bow and the vibrant lights.
the final result becomes
The vector t-shirt design of a decorative bow with a checkered black-and-white pattern. The edges of the bow have a beaded trim, and it is wrapped with colorful Christmas string lights. The lights come in various colors, including red, blue, green, yellow, and pink, giving it a festive and cheerful holiday look. The black background highlights the details of the bow and the vibrant lights. The background is a white color, making the characters and text stand out.
You can add more variations if you want different results!
I’m using this simple example to help you understand.
In reality, when you generate images, they often look like this, with a background surrounding the design that’s very difficult to separate.
However, with the prompt adjustment method above, I have separated the background out; now I can easily handle it in Photoshop as shown in the image below.
The disadvantage of MidJourney is that the text generation is very poor, so to solve this, you can flexibly use Ideogram as well.
Okay, now here’s another example.
The cat above with wires thrown around it is in high demand for purchases from users. Now I will do the following:
- I will detach the cat
- Put it into the list of tools I mentioned at the beginning of this document
- Take all the prompts generated from that through MJ or Ideogram to create images
The result is as below, all these Ideas have the potential to win.
Prompt: A t-shirt design of an illustration of a cartoon raccoon wearing sunglasses and holding up a glass of wine, wrapped in Christmas lights. The text says “Overstimulated & Festive AF”.
I’ll adjust the structure a bit.
The vector t-shirt design of an illustration of a cartoon raccoon wearing sunglasses and holding up a glass of wine, wrapped in Christmas lights. The text says “Overstimulated & Festive AF”. The background is a white color, making the characters and text stand out.
Return result
Note: To achieve the best results, you should try to crop the image to a large size so that ChatGPT-4o can create the most accurate prompt possible. The prompt that describes the image usually requires us to add a few more elements to make the image look the best; typically, this involves separating the background, converting to vector, and editing the text.
In some cases, I have to extract each part of the image to describe it as accurately as possible!
Up-Scale Images
I spent quite a lot of time understanding these tools, you can learn more.
Below is the mechanism for image enlargement.
The operating principles of current upscale tools:
Increase the resolution of the image
- Increase the resolution of the image
- Reduce noise
- Enhance sharpness
- Reconstruct and predict details
Illustrative image
On the left is the before image, and on the right is the after image when Up-Scale and sharpened.
With a series of interventions like those bullet points above, enlarging the image will make it not look real anymore. Therefore, the first principle when enlarging an image is the image must not be too blurry or too detailed.
I will list 3 tools considered the most optimal, ranked in descending order for you to choose from:
dgb.lol (perfect)
pixelcut.ai (perfect)
upscayl.org (occasionally has artifacts in the image)
The advantage of upscayl.org is that it can be installed on a computer and can process multiple designs in a folder. If you are not a perfectionist due to the drawbacks I’ve listed above, then upscayl is definitely a good choice.
There is an older tool called Gigal Pixel that I used before, but I no longer use it because the output files are quite large, and the computer installation requires high configuration.
Reducing Image Size
Generally, when reducing image size, except for images with too many colors, I recommend using Illustrator.
After 77 trials, I found that Illustrator is the most optimal.
It helps to merge color areas that are slightly off-color into one.
For example, the image below weighs up to 4MB; when upscaled, it can double the size (this can create difficulties for sellers when printing, uploading products, work speed, etc.)
The solution for images with excessively large file sizes will be to use the Trace feature in Illustrator to merge colors.
As you can see in the image above, there is not much difference in color quality, and the size has been reduced by 2/3.
Hope these tips will help you save a lot of design time.
Thank you to our valued customers for trusting and supporting the YTuong team for the project Heyety.com and all products in the ecosystem.
In addition, sortpin.com and gsctool.com still have promotions until December 3rd.
Below are some feedback from customers.
About the author
I am John, selling products on Etsy, and I am also a writer.
I hope that my sharing can inspire you to have more motivation in your work.
My shares have achieved a high reading rate, so I can say that the experiences I share can definitely help you achieve similar results
Success comes from both big achievements and small contributions!
👏 Bonus: Get my Free PDF guide on:
❤️🔥 Subscribe HERE to receive an email each time I post my next article.
😻 Tools I Use to become successful on Etsy. And most of them are free to use.
Do you like my article? Don’t forget to subscribe to receive email updates on all my publications.
