15 Mind-Blowing New ChatGPT Image Generator Prompts (That Actually Worked, Or Totally Flopped)

Ben Milsom
6 Apr 202513:20

TLDRThe speaker shares their experience experimenting with the GPT40 image generator, highlighting both successes and failures. They explore its potential for creative workflows in product marketing, branding, and concepting. Examples include branded photoshoots, logo creation, and personalized imagery. While noting limitations like slow generation speed, they emphasize the tool's usefulness and potential for future development. They also discuss practical applications like lighting advice and product macro shots, showcasing the possibilities of integrating AI into creative processes.

Takeaways

  • 👨‍🎨 The speaker tested various prompts for GPT40's image generation, finding some successful and others not so much.
  • 🚀 Some prompts led to paid client work, showing the tool's potential in product marketing, branding, and concepting.
  • 💡 The new GPT40 image model is accessible through a 'create image' button, and users can test it with unique prompts.
  • 🎥 The speaker explored using GPT for branded photoshoots, though found it slower and less realistic compared to other tools like Replicate.
  • 👨‍👩‍👧‍👦 GPT was used to create personalized, Disney-like images for Mother's Day, with mixed results but overall success.
  • 👕 The tool was tested for creating custom t-shirt designs, showing potential for quick concept visualization.
  • 📈 Structured information (e.g., JSON) was found to be effective for generating consistent logos and icons.
  • 🌐 GPT was used to create an April Fool's prank for a client's social media, demonstrating its versatility.
  • 💡 The speaker found that GPT provided better character consistency in image generation compared to other models.
  • 💡 The tool was tested for various creative applications, including lighting setups and macro shots, with mixed success.
  • 🤖 The speaker emphasizes the importance of integrating AI tools into real workflows for practical benefits.

Q & A

  • What is the main focus of the video?

    -The video focuses on exploring various image generation prompts using GPT40, discussing what worked and what didn't, and how this technology can be applied in creative workflows such as product marketing, branding, and concepting.

  • How did the speaker test the GPT40 image generation capabilities?

    -The speaker spent a week experimenting with different prompts, pushing the tool to its limits, and observing the results. They also compared the outcomes with other models like Flux Pro Ultra in Replicate.

  • What are some examples of successful image generation prompts mentioned in the video?

    -Successful prompts included creating branded photoshoots for companies like an electrical contractors company and a landscape gardener, generating personalized Disney-style images for Mother's Day, and creating stylized logos using JSON prompts.

  • Why did some image generation attempts fail or produce unsatisfactory results?

    -Some attempts failed due to issues like lack of realism, inconsistency in character portrayal, or the tool's inability to accurately interpret the prompt. For example, the lighting setup prompt resulted in images that were too dark and gloomy.

  • What are the limitations of GPT40 image generation as discussed in the video?

    -The limitations include slow generation speed, which makes it impractical for commercial batch processing, and occasional inconsistencies in character portrayal. It also struggles with certain complex prompts, like accurately generating macro shots of products.

  • How does the speaker suggest using JSON prompts with GPT40?

    -The speaker suggests using JSON to structure information in a way that language models can easily understand. This helps in generating consistent and stylized outputs, such as logos and icons, which can be used for branding purposes.

  • What is the speaker's opinion on using GPT40 for commercial purposes?

    -The speaker believes that while GPT40 is incredibly useful for certain creative tasks, it is currently too slow for commercial batch processing. They suggest waiting for API access and faster speeds to make it viable for commercial use.

  • How did the speaker use GPT40 for a Mother's Day project?

    -The speaker used GPT40 to create personalized Disney-style images of their family. The tool was able to generate appealing and beautiful images, adding a new level of customization for personalized messages and cards.

  • What is the difference between using GPT40 and other models like Flux Pro Ultra for image generation?

    -GPT40 provides better consistency in character portrayal and art direction compared to Flux Pro Ultra. However, Flux Pro Ultra generates images faster but lacks the consistency needed for certain creative workflows.

  • What are the speaker's future plans regarding GPT40 and image generation?

    -The speaker plans to continue exploring the tool's potential and integrate it into real-world workflows. They are also excited about the possibilities of API access, which they believe will change the game for image generation and automation.

Outlines

00:00

🎨 Exploring the Limits of GPT40 Image Generation

The speaker describes their experience over the past week experimenting with GPT40's image generation capabilities. They share their successes and failures, highlighting how some prompts led to paid client work and changed their approach to creative workflows in product marketing, branding, and concepting. They explain how to access the new GPT40 image model and demonstrate its capabilities by generating an image of a man in a sad duck costume. The speaker then discusses their experiments with creating branded photoshoots for different companies, comparing the results with those from other models like Replicate's Flux Pro Ultra. They also mention using GPT40 to create personalized images for Mother's Day and generating stylized logos using structured JSON prompts. Finally, they show how GPT40 can be used to create consistent character imagery and provide art direction for various scenarios.

05:01

🖼️ Comparing GPT40 with Other Models and Practical Applications

The speaker continues their exploration of GPT40's image generation by comparing it with other models like Flux Pro Ultra in terms of consistency and speed. They demonstrate how GPT40 can generate consistent character imagery, which is crucial for branding and influencer marketing. They also discuss using GPT40 for practical applications such as providing lighting advice for a photography setup, generating macro shots of products, and creating stylized logos. The speaker highlights the potential for using structured information like JSON to achieve consistent and high-quality results. They show examples of GPT40-generated images, including a rugged character called the Maverick in various cinematic settings, and compare these with less consistent results from other models. They also mention using GPT40 to create an April Fool's prank for a client's social media channels.

10:01

🎥 Future Possibilities and Shifting Focus

The speaker concludes their exploration of GPT40's image generation capabilities by discussing its potential for video generation and animation. They demonstrate how GPT40 can generate images from different angles and perspectives, which can be used to create smooth transitions in video content. The speaker reflects on the rapid advancements in AI image generation and the endless possibilities it opens up for creative industries. They then shift focus back to their core work of building custom tools and automations for clients, emphasizing their commitment to improving ad conversion rates and scaling revenue-generating tools. They mention their ongoing work with Lovable, a client, and the potential for further innovation with API access. The speaker encourages viewers to experiment with GPT40 themselves and integrate it into real workflows to unlock its full potential.

Mindmap

Keywords

💡ChatGPT Image Generator

The ChatGPT Image Generator is a tool that creates images based on text prompts. In the video, the speaker explores the capabilities of this tool by testing various prompts to see what kind of images it can produce. For example, they tried to generate images of a man in a sad duck costume eating a hot dog on the beach in the rain, which demonstrates the tool's ability to interpret and visualize unique and specific scenarios.

💡Art Direction

Art direction refers to the overall visual style and look of a project, including elements like color, composition, and lighting. In the context of the video, the speaker discusses how they used ChatGPT to get art direction prompts for branded photo shoots, such as for an electrical contractors company and a landscape gardener. They wanted to achieve consistency in the characters and lifelike imagery, and the tool helped them come up with ideas for how to achieve that.

💡Branded Photo Shoots

A branded photo shoot involves creating images that align with a specific brand's identity and style. The speaker experimented with using ChatGPT to generate branded photo shoots for different companies. For instance, they asked for a series of images for an electrical contractors company, where the workers were depicted in branded clothing and dealing with electrical contractor tasks. This shows how the tool can be used to create visual content that reflects a brand's image.

💡Realism

Realism in the context of image generation refers to how lifelike and realistic the generated images appear. The speaker notes that while some of the images generated by ChatGPT looked fine, they still appeared a bit artificial or 'dead-eyed.' They compared this to images generated by another tool called Replicate, which they felt looked more realistic due to the use of RAW output. This highlights the ongoing challenge in image generation to achieve high levels of realism.

💡API Access

API access refers to the ability to programmatically interact with a tool or service using Application Programming Interfaces. The speaker mentions that they are waiting for API access to the ChatGPT image generation tool, which would allow for faster and more automated use of the tool. This is important for commercial applications where speed and efficiency are crucial, such as batch processing images for a client.

💡Personalized Messages

Personalized messages are communications tailored to a specific individual or group. In the video, the speaker talks about using ChatGPT to create personalized Mother's Day imagery for their family. They were able to generate Disney-like or Pixar-style images of their family members, which shows how the tool can be used to create unique and personalized content for special occasions.

💡JSON Prompt

A JSON prompt is a way of structuring information in a format that language models can easily understand. The speaker mentions that a user named Jack created an amazing JSON prompt that helped them generate stylized logos. By using JSON, they were able to ensure that the outputs were consistent and well-structured, which is important for creating cohesive visual elements like logos and icons.

💡Character Consistency

Character consistency refers to maintaining the same visual appearance and characteristics of a character across multiple images. The speaker highlights how ChatGPT was able to generate a series of images of a rugged character called the Maverick, where the character looked the same throughout. This is a significant advantage over other models like Flux Pro Ultra, which did not achieve the same level of consistency.

💡Macro Shots

Macro shots are close-up photographs that capture fine details of a subject. The speaker experimented with feeding low-resolution product images to ChatGPT and asking for macro shots. For example, they used a low-res image of a track pump and asked for high-res macro shots, which the tool generated, although with some inaccuracies. This demonstrates the potential of the tool for creating detailed product imagery.

💡Video Generation

Video generation involves creating video content, often by interpolating between images. The speaker mentions that the ability to manipulate images generated by ChatGPT, such as changing the camera angle or adding subjects, opens up possibilities for video generation. They suggest that by tweening between different generated images, the tool could create frames for a video, which is an exciting application for the future.

Highlights

The speaker spent a week testing GPT40's image generation capabilities, exploring both successful and failed prompts.

Some prompts led to paid client work almost instantly, demonstrating the tool's practical applications.

The new GPT40 image model is now accessible to all users, identifiable by the 'create image' button.

The speaker successfully generated consistent character imagery for a branded photoshoot for an electrical contractors company.

Using GPT for branded photoshoots showed potential but still felt artificial compared to other models.

The Replicate model generated more realistic images faster than GPT40, highlighting the need for speed improvements.

GPT40 successfully created personalized Disney-style imagery for Mother's Day cards.

The tool was used to generate skateboarding-themed t-shirt designs and visualize them on models.

Structured information using JSON prompts can enhance logo and icon creation with GPT40.

GPT40 was used to create an April Fool's prank by transforming a truck image into an ice cream truck.

The speaker demonstrated consistent character generation across different scenes using GPT40.

GPT40 provided interesting but imperfect lighting setup advice for a photo shoot.

The tool generated conceptual macro shots of products, though with some inaccuracies.

GPT40's ability to manipulate camera angles and subjects in generated images opens possibilities for video generation.

The speaker emphasizes the need to integrate GPT40 into real workflows for maximum value.