In this post, we will see how ChatGPT responds when generating a prompt for the system itself, with the DALL-E alternative also available for testing, and with a final version in Leonardo AI, also with the same prompt. As a starting point, we used this image posted in the Facebook group AI Prompt Hub.
This is the image we will call the original image, or reference image:
This is a serene night landscape with a full moon illuminating a calm river flowing through a lush forest, featuring a solitary tree and distant mountains under a starry sky.
Table of Contents
Prompt By ChatGPT
Based on the reference image, this is the prompt generated by ChatGPT, which we will use by applying it to different text-to-image AI models.
A surreal, highly detailed fantasy artwork of a glowing full moon illuminating a peaceful river at night. The moon appears large and radiant in the sky, casting a magical golden reflection across the flowing water. A tall, graceful tree stands on the riverbank, its leaves glowing softly in the moonlight. The river flows gently over small rocky cascades, surrounded by lush green grass, bushes, and flowering plants that sparkle with soft light. In the background, dark cliffs rise beneath a sky filled with glowing clouds and a flock of birds flying across. The entire scene is bathed in shades of deep blue, turquoise, and silver, creating a serene, dreamlike, and mystical atmosphere. 9:16 Ratio with 16K.
Images Generated With DALL·E 4o
With that same prompt and also in ChatGPT but using the specific DALL-E chat, with model 4o, I get these results:
#1
We can see an AI-generated artwork showcasing a mystical river landscape lit by an enormous luminous moon. The scene features cascading water over mossy rocks, a glowing tree, colourful flowers, and a cosmic sky filled with stars and planets. This enchanted night view beautifully blends elements of nature and fantasy, perfect for AI art enthusiasts and lovers of surreal digital landscapes.
#2
Now we can see an enchanting AI-generated night scene displaying a radiant full moon casting golden light over a serene river valley. The lush greenery is dotted with glowing trees and magical fireflies, while the river flows smoothly amidst mist and rocky cliffs. This captivating blend of nature and celestial beauty is ideal for fans of AI art and fantastical landscapes.
Images Generated With DALL·E 5
We are now using model GPT-5 instead of 4o, although the results will be very similar since we are using the same prompt.
#1
The image invites us to immerse ourselves in a fantasy scene where a radiant golden moon illuminates a peaceful river flowing through lush valleys and sparkling flowers, under a cosmic sky filled with planets and stars. A dreamlike artwork that blends nature, magic and serenity.
#2
We can see a digital fantasy illustration showing a luminous full moon over a river flowing through rocky valleys, with a tree in the foreground and multiple planets in the night sky. High-resolution artwork with detailed lighting and textures.
Images Generated With ChatGPT
ChatGPT also generates images directly, without leaving the chat, i.e. without having to click on the DALL·E GPTs.
When we enter the prompt, this is the result we get:
We can see a high-resolution fantasy artwork showing a luminous full moon above a peaceful river flowing over small rocky cascades. A single tree stands on the riverbank surrounded by vegetation, while distant mountains rise under a cloudy, star-filled night sky. The moonlight creates golden reflections on the water, highlighting the contrast between deep blue tones and bright details.
And How About Sora?
Without leaving ChatGPT, we decided to try the video and, using exactly the same prompt, we tested Sora’s ability to see what it could generate. This is text-to-image, with no initial image or style reference.
Let’s See How Leonardo AI Responds.
For Leonardo, we are going to use a prompt very similar to the previous one, although with some changes in terms of the visual reference. I’m going to use LUCID ORIGIN as a model to generate the image, because it handles details such as water, sky, and stars very well. I could also use LUCID REALISM, although I think it works better in cinematic scenes with characters but doesn’t give such good results with natural environments.
LUCID ORIGIN
🎯 Prompt (Text-to-Image)
A fantasy night landscape with a gigantic glowing full moon in the sky, casting a golden reflection across a calm river. A single tall tree stands on the riverbank with a simple silhouette and sparse leaves. The river flows gently with small cascades over rocks, surrounded by soft grass. In the background, subtle mountain silhouettes and a few scattered clouds in a dark blue night sky. Cool blue tones dominate the scene, with golden light only in the moon and its reflection. Minimal composition, clean atmosphere, soft textures, vertical 9:16.
🚫 Negative Prompt
No humans, no extra trees, no multiple moons, no vivid green, no over-saturated colours, no heavy cinematic effects, no clutter, no text, no logos.
This is the result:
This is a high-resolution fantasy illustration of a glowing full moon rising over a calm river with golden reflections, a single tree in the foreground, and distant mountain silhouettes under a night sky.
My Final Assessment
Leonardo is a safe bet, interpreting prompts in detail and offering much more comprehensive customisation options than DALL·E and ChatGPT, although I already knew this before I started. What caught my attention here is that ChatGPT generates much more interesting images than DALL·E, i.e., the global module performs better than the specific module for generating images, and I suppose that is in line with a business strategy.
I understand that OpenAI is investing its money in improving ChatGPT’s performance and needs this AI to meet all the expectations of users who decide to rely on it, especially paying users, of course. But even though I’m not surprised by the present,
I’m convinced that DALL·E will be one of the cornerstones of the creative workflow integration offered by OpenAI with ChatGPT. This means that they will improve the editing and control of image editing, making it more intuitive and consistent in terms of characters and styles, something that all AI users have been demanding for a long time. DALL·E will maintain and improve its symbiosis with ChatGPT. Video and animation are likely to become the main features of the model, generating video from text—something OpenAI is now doing with Sora—but also offering animation of the images it generates.
All the advantages that DALL·E offered as GPTs are now integrated into GPT5, but there is something that caught my attention: with the same prompt, I get better results with GPT5 than with DALL·E5. Or at least the generated image is closer to what I expected before I started. DALL·E5 has given it a more artistic look, while GPT5 has generated an image with a more cinematic photographic style.
That’s all about this moon river prompt. I would love to hear your thoughts on this, so please leave a comment and I will respond.
💜 Thank you for taking the time to read this.
Related Content
Amazing Leonardo AI Prompt With 15 Faces
Google Gemini’s Picture Me. Much better than you can imagine