Perchance Inage maker
-
- Posts: 2
- Joined: Mon Feb 01, 2016 3:33 pm
Perchance Inage maker
Hi to everybody. I have a question about Perchance image maker. I can type in what i want, all images are not as wished. When i type in "woman till her shoulders in mud" it makes only images where i see a upper body or only with the legs in mud. Have tryed to type different things, but makes not the images i wish. Has anybody a good idea for it.
- Mudbogz
- Posts: 209
- Joined: Sun Apr 26, 2009 3:27 am
Re: Perchance Inage maker
You can try modifying your parameters... make sure you separate your parameters with a comma. Try something like this:
Woman, blonde hair, sinking shoulder deep in quicksand
You can add other parameters such as clothing or body shape, skin color etc. just make sure you have a comma between parameters.
Hope that helps!
Mudbogz
Woman, blonde hair, sinking shoulder deep in quicksand
You can add other parameters such as clothing or body shape, skin color etc. just make sure you have a comma between parameters.
Hope that helps!
Mudbogz
- cerberus
- Posts: 552
- Joined: Thu Apr 16, 2009 7:34 pm
Re: Perchance Inage maker
Perchance uses Stable Diffusion as its AI generator, so I'd have a look on the internet for advice on writing Stable Diffusion prompts, there's quite a lot of it available. However, the interpretation of prompts is a bit dependent on the checkpoint models used, so not all advice may apply to Perchance.
There is a recommended structure for SD prompts:
- framing instructions e.g. close-up shot from in front
- subject detail e.g. blonde haired woman aged 25 wearing a bikini
- situation detail e.g. sinking in quicksand on a beach
- other details e.g. sunny day clear skies
I've found that prompt terms like waist deep, shoulder deep etc are generally ignored and the AI derives the depth from the framing instructions and image aspect ratio. By this I mean, you have a better chance of getting a shoulder deep subject by choosing a landscape format image and specifying "close-up shot" or even "extreme close-up shot". Perchance allows batches, if I remember correctly, I used to use these most of the time, usually 4 batches of 4 images per batch (i.e. 16 images a generate), then I picked the best ones.
Another thing about prompts is that they are prioritised from the front, so if a term it getting ignored putting it closer to the front of the prompt can help.
Prompt terms can be weighted to increase their effect, this is done by putting the term in parentheses and adding a weight factor e.g. "(sinking in quicksand:1.4). Note 1: this does not necessarily make them sink deeper but may make them muddier etc. Note2: too high a weighting factor cause weird effects, with Perchance keep it below 1.8.
You can strengthen the effect of prompt terms by using multiple terms with similar meanings, results can be a little unpredictable.
Some final points, make use of the negative prompt to suppress things you don't want, I always find, when doing quicksand/bog pictures that "water" is a good thing to suppress. Also, Perchance, like most other AIs, has restrictions on violent, hateful and harmful images, so you tend not to be able to get the subjects in over their shoulders, though it will occasionally generate this kind of image of it's own accord.
Hope this helps, post some pics so we can see how you are doing and feel free to ask questions.
There is a recommended structure for SD prompts:
- framing instructions e.g. close-up shot from in front
- subject detail e.g. blonde haired woman aged 25 wearing a bikini
- situation detail e.g. sinking in quicksand on a beach
- other details e.g. sunny day clear skies
I've found that prompt terms like waist deep, shoulder deep etc are generally ignored and the AI derives the depth from the framing instructions and image aspect ratio. By this I mean, you have a better chance of getting a shoulder deep subject by choosing a landscape format image and specifying "close-up shot" or even "extreme close-up shot". Perchance allows batches, if I remember correctly, I used to use these most of the time, usually 4 batches of 4 images per batch (i.e. 16 images a generate), then I picked the best ones.
Another thing about prompts is that they are prioritised from the front, so if a term it getting ignored putting it closer to the front of the prompt can help.
Prompt terms can be weighted to increase their effect, this is done by putting the term in parentheses and adding a weight factor e.g. "(sinking in quicksand:1.4). Note 1: this does not necessarily make them sink deeper but may make them muddier etc. Note2: too high a weighting factor cause weird effects, with Perchance keep it below 1.8.
You can strengthen the effect of prompt terms by using multiple terms with similar meanings, results can be a little unpredictable.
Some final points, make use of the negative prompt to suppress things you don't want, I always find, when doing quicksand/bog pictures that "water" is a good thing to suppress. Also, Perchance, like most other AIs, has restrictions on violent, hateful and harmful images, so you tend not to be able to get the subjects in over their shoulders, though it will occasionally generate this kind of image of it's own accord.
Hope this helps, post some pics so we can see how you are doing and feel free to ask questions.
Cerberus
Johnny Dowd: Be content with your life. It may not get any better!
Johnny Dowd: Be content with your life. It may not get any better!
Who is online
Users browsing this forum: No registered users and 2 guests