Writing Prompts

Hey Deepmoders, 

I want to share a little about how I write my prompts in Deepmode. This is definitely not a definitive method and I am sure you'll have your own opinions, but hopefully, my method might give you a few ideas.

Watch the video or read on to find out more.

Generally, I like to keep things simple. I see some people use very detailed, almost story-like descriptions but I tend to be a bit more technical. Things like "The atmosphere is calm and inviting capturing a moment of quiet reflection" will not add much and won't be interpreted by AI the same way a human would understand it

I usually divide my prompt into two parts. Part one is a general overview of the style and technical aspects of my desired image and Part two is specifics about what I want in terms of subject and theme.

Usually, the first part of my prompt that focuses on the overall theme is something that will stay the same and can often be copied and pasted through many different prompts. Keeping in mind I want to keep things simple this part of the prompt is usually a list of film or photography techniques. As mentioned in my Lighting tutorial video, using technical terms from film and photography is a quick way to set a tone. Commas should separate these keywords and there is no need to include proper sentences, this will improve the chances of getting the result you want.

Include information on your framing, for example, closeup or wide shot. Describe your lighting. Is the image greyscale or color? Do you want a specific camera lens or an analog film effect? Set the time of day and list any effects such as depth-of-field or bokeh. Just make sure these are all general terms that could be applied to almost any situation or scene.

The second part of the prompt is going to describe what you want your image to be about. Deepmode has a focus on a model so this is usually “man” or “woman”. These are the trigger words for Deepmode specifically and need to be included.

I then like to describe my model’s hairstyle, clothing, and if they have any specific style of makeup. Following that I list clothing, pose and the location and this is where being more literal and using basic language is going to help. Currently, Deepmode's prompt adherence is limited, and getting overly descriptive or very specific is going to complicate things.

Keeping all of this in mind let's build a simple prompt...

closeup photo, Depth of field, cinematic lighting,

woman, green t-shirt, black bob haircut, sitting in a bedroom,

This is a good base to work from and from here you can choose to change hairstyle or the color of the clothes. And you can do this either using remix or by generating a new image.

The main advantage of prompting in this way is that, usually, the things you are more likely to edit or tweak are all listed clearly in the second section and adding new things or adjusting colors etc. is a little more clear. 

Why not try these techniques yourself and make your own Farah images at Deepmode.

Also check out my Farah X page or Farah Films Youtube channel for more AI tips and content.