HOME | DD

Description
I often get requests for tips and details about my prompts, so here is a quick reference guide!

Software: Bing app (mobile or desktop), which utilizes DALLE-3. Won’t create real people, will create fictional characters, even if they are copyrighted, unlike ChatGPT-4, which is painfully more restrictive in my opinion. I also tried DreamUp here but didn’t care for the results; however, the options for different aspect ratios is nice.

Exact prompt used for the image here: “happy confident young woman Disney Princess Rapunzel cozy pink winter clothes warming feet soles fireplace, light makeup, low angle full body pov, barefoot, hyperrealistic Pixar 3D animation art style, realistic hourglass body proportions, colorful festive tree in background, log cabin interior, blurred background, soft lighting, music earbuds, normal medium length straight golden hair, sitting cross-legged on soft carpet floor, f/8 ISO 90 high saturation, hands under chin”

General tips:

1 - separate details with commas. In my experience the first set of words has the strongest impact on the final result, so be detailed about the main aspects of your image there.

2 - be very descriptive and use lots of adjectives to describe the scene in your imagination! You can even use AI LLMs to help you construct detailed structure with only a few words.

3 - build prompts from a more general and less restrictive set at first, then progress into more details. This will make it easier to identify phrases or words that might be triggering content filters.

4 - use words that you think could be in the tags of images with elements that you want. Sentence fragments are fine for descriptive elements, but you should include your main subject and primary verb in your opening string to your prompt.

5 - if a particular phrase is triggering the filters, ask for help from an AI to try rephrasing it with a more innocuous description! Aso, note that Bing has two possible reasons for prompt failures: inappropriate prompt and inappropriate content detected in search results when constructing the image from its training data. For the first error, do NOT try the same prompt again as it will definitely fail again and could even lead to account suspension. For the second type of error, sometimes trying a few times results in some image creation, so you can either keep trying or adjust the prompt. Just be aware that you get ~500 prompt attempts each day, failures included.

6 - try searching your own prompt strings. The search results will show you a similar set of images that the generator will reference for that string. Sometimes this can help identify whether or not a particular string is effectively generating useful reference images.

That’s all for now, but I will probably edit and add to this post, as AI image creation is still in its early days! Happy creating, artists!

Description
I often get requests for tips and details about my prompts, so here is a quick reference guide!

Software: Bing app (mobile or desktop), which utilizes DALLE-3. Won’t create real people, will create fictional characters, even if they are copyrighted, unlike ChatGPT-4, which is painfully more restrictive in my opinion. I also tried DreamUp here but didn’t care for the results; however, the options for different aspect ratios is nice.

Exact prompt used for the image here: “happy confident young woman Disney Princess Rapunzel cozy pink winter clothes warming feet soles fireplace, light makeup, low angle full body pov, barefoot, hyperrealistic Pixar 3D animation art style, realistic hourglass body proportions, colorful festive tree in background, log cabin interior, blurred background, soft lighting, music earbuds, normal medium length straight golden hair, sitting cross-legged on soft carpet floor, f/8 ISO 90 high saturation, hands under chin”

General tips:

1 - separate details with commas. In my experience the first set of words has the strongest impact on the final result, so be detailed about the main aspects of your image there.

2 - be very descriptive and use lots of adjectives to describe the scene in your imagination! You can even use AI LLMs to help you construct detailed structure with only a few words.

3 - build prompts from a more general and less restrictive set at first, then progress into more details. This will make it easier to identify phrases or words that might be triggering content filters.

4 - use words that you think could be in the tags of images with elements that you want. Sentence fragments are fine for descriptive elements, but you should include your main subject and primary verb in your opening string to your prompt.

5 - if a particular phrase is triggering the filters, ask for help from an AI to try rephrasing it with a more innocuous description! Aso, note that Bing has two possible reasons for prompt failures: inappropriate prompt and inappropriate content detected in search results when constructing the image from its training data. For the first error, do NOT try the same prompt again as it will definitely fail again and could even lead to account suspension. For the second type of error, sometimes trying a few times results in some image creation, so you can either keep trying or adjust the prompt. Just be aware that you get ~500 prompt attempts each day, failures included.

6 - try searching your own prompt strings. The search results will show you a similar set of images that the generator will reference for that string. Sometimes this can help identify whether or not a particular string is effectively generating useful reference images.

That’s all for now, but I will probably edit and add to this post, as AI image creation is still in its early days! Happy creating, artists!