I Got DALL-E 3 Access: My First Impressions and Tests

TL;DR

DALL-E 3’s strongest first-impression feature is natural-language image iteration, though its photorealism and consistency are uneven. Tests produced four initial images, moved a model through New York and Norway scenes, modified a chair image, and created Michael Scott-style memes, blog illustrations, blueprints, and YouTube thumbnails. Read on to see which edits worked, where details changed unexpectedly, and why the workflow still impressed the creator.

Transcript

okay so I just got access to doll E3 as you can see here so I thought this video we can just test a few things out so I thought we can just start with a professional photo of a stunning American born female in a 16 9 aspect ratio so what I want to do here now in the beginning is just to try to generate an image I like then I want to iterate on it m... Read More

Key Insights

• ✋ Doll E3 has the potential to generate high-quality and realistic images, although it may not reach the level of photorealism achieved by other platforms.
• 👤 The platform is capable of generating memes in the style of popular comedy quotes, providing users with creative options for humor.
• 🔨 Doll E3 can be a valuable tool for illustrators, as it can generate illustrations and blueprints for various purposes quickly and efficiently.
• 🎨 It is possible to experiment with different designs and colors using Doll E3 to create visually appealing YouTube thumbnails.
• 🍧 The natural language iterations within Doll E3 make it convenient for generating variations without having to rewrite the entire prompt.
• ☠️ The wait time for generating images may be subject to a rate limiter, requiring users to be patient before generating additional images.
• 🛄 Doll E3's image upload feature, aimed for future development, has the potential to enhance its capabilities further.

Questions & Answers

Q: What were the creator’s first impressions of DALL-E 3?

The creator found DALL-E 3’s images impressive and especially liked making follow-up changes with natural language. However, some photorealistic scenes shifted toward a painted look, and requested details were not always preserved.

Q: How realistic were DALL-E 3’s generated photos?

The first set of four professional photos looked realistic to the creator, though not at Midjourney’s level. A later Norway hiking image looked more like a painting, while the model herself still looked quite good.

Q: Can DALL-E 3 keep the same person while changing scenes and outfits?

It could create related variations, including placing the woman on a New York street and dressing her for hiking in Norway. The woman was not exactly the same across images, and the visual style also changed during later iterations.

Q: How well does natural-language image editing work in DALL-E 3?

The creator could request changes conversationally without rewriting the full prompt each time. DALL-E 3 changed a chair to pink, replaced a monkey with a Pokémon, and made the Pokémon larger and more realistic, although it also altered the chair unexpectedly.

Q: Did DALL-E 3 follow small editing instructions accurately?

Results were mixed. It made the chair pink as requested, but the banana did not appear smaller, and replacing the monkey with a Pokémon changed the chair more than expected.

Q: Can DALL-E 3 generate Michael Scott-style memes?

The creator prompted it to make a Michael Scott-style “that’s what she said” meme and then asked for funnier, edgier versions. Some outputs did not make sense, but a few were funny and the generated text was close rather than perfect.

Q: What other uses of DALL-E 3 were tested?

The tests included illustrations for a blog post, blueprints for an imagined futuristic gaming console and the human brain, and YouTube thumbnail concepts. The creator also experimented with different thumbnail designs and colors.

Q: What limitations appeared during the DALL-E 3 tests?

Character identity and scene details were not always consistent between iterations, and some photorealistic results became more painting-like. Requested object changes could also be missed, while unrelated elements such as the chair sometimes changed substantially.

Summary & Key Takeaways

• The video demonstrates the process of generating photorealistic images of a female model in different scenes, showcasing the varying levels of realism achieved.

• The creator also explores the ability to generate memes in the style of Michael Scott's iconic "that's what she said" joke, experimenting with different prompts to create humorous variations.

• Doll E3's capabilities are further tested by generating illustrations for a blog post and exploring its potential in creating blueprints for an imagined futuristic gaming console and the human brain.

• The creator also explores the possibility of using Doll E3 to create YouTube thumbnails, experimenting with different designs and colors.