WhoBuilt

ChatGPT's New Image Model Changes Everything: 5 Incredible Use Cases

Apr 2, 2025· 4 min· 2.9K views·Watch on YouTube

Peter Yang, a host on the Peter Yang channel, demonstrates five ways to use ChatGPT’s new image model; no guests are identified.

Turn personal photos into anime art

Yang shows family photos transformed into Studio Ghibli-inspired art and says the process uses ChatGPT Plus, the mobile app, and GPT-4o. He also suggests trying the approach on memes and movie scenes. If a style request is blocked, he recommends describing the visual traits instead of naming the style.

Create marketing assets from references

Yang demonstrates adapting an existing social ad for a new product, creating an image of Rick Rubin holding Sahil Bloom’s book, and combining a drink ad with Liquid Death branding. He says the model can generate marketing imagery and copy from visual references.

Rework websites and infographics

Yang shows a Lovable landing page converted to an anime look and an infographic redesigned with cuter illustrations. He then asks for illustrations that better fit the product management content.

Build a thumbnail through iterative edits

Yang starts with his own thumbnail and asks for an anime-inspired version, then removes the text so he can position it himself. He requests a separate transparent image of the hand and finger, adjusts his avatar’s expression using Greg Eisenberg’s thumbnail as a reference, and asks for missing hands, sky, and clouds before adding the text.

Combine image and video tools

Yang says the image model can be paired with a video model such as Luma to make anime short films. He shows a short movie made for Severance as an example.

Takeaways

  • Start image generation with a visual reference when you need a design to follow an existing ad or brand direction.
  • Make edits in stages, separating artwork, transparent elements, expression, and text so each change is easier to control.
  • Describe visual characteristics when a named style prompt does not work.
  • Review generated illustrations for relevance to the content and request specific corrections.
  • Pair image generation with a video model such as Luma to explore short animated sequences.

Mentioned in this talk

Search inside this talk

What was claimed 9

  • feature
    ChatGPT: Generates and edits images from prompts and reference images, including styled photos, marketing assets, infographics, and thumbnails.
    “what it can do”
    YouTube
    Unverified1 source· Talk
  • uses
    “pay $20 for ChatGPT Plus”
    YouTube
    Unverified1 source· Talk
  • uses
    “pick GPT-4o as your model”
    YouTube
    Unverified1 source· Talk
  • metric
    “pay $20 for ChatGPT Plus”
    YouTube
    Unverified1 source· Talk
  • feature
    OpenAI: Adding more moderation checks, according to the host.
    “OpenAI is adding more moderation checks”
    YouTube
    Unverified1 source· Talk
  • feature
    ChatGPT: Can create marketing assets by mimicking a reference image and adapting it for a product.
    “amazing at creating marketing assets if you give it the right image to mimic”
    YouTube
    Unverified1 source· Talk
  • feature
    ChatGPT: Can create websites and infographics and revise illustrations to better fit their content.
    “great at creating websites and infographics”
    YouTube
    Unverified1 source· Talk
  • feature
    ChatGPT: Supports iterative thumbnail editing, including removing text and exporting image elements with transparent backgrounds.
    “export them as a separate transparent image”
    YouTube
    Unverified1 source· Talk
  • integrates with
    ChatGPT integrates with Luma
    “use this model along with a video model like Luma”
    YouTube
    Unverified1 source· Talk

Related talks