[0:00] what do all those symbols in stable
[0:01] diffusion really mean how can we write
[0:03] prompt words so that the pictures we get
[0:06] are more what we want when you see those
[0:08] complex stable diffusion prompt words do
[0:10] you feel confused too don't worry this
[0:13] video will help you with all kinds of
[0:15] problems when you fill in prompts first
[0:18] off let's check out the basic principle
[0:20] of prompt words prompt words are
[0:22] separated by commas and you can put
[0:24] prompt words on different lines but you
[0:26] still need a comma at the end of each
[0:28] line as for weight each prompt word has
[0:31] a default weight of one but the words at
[0:33] the front get a higher weight so put
[0:36] important words at the front if you can
[0:38] finally keep the number of prompt words
[0:40] within 75 if there are too many words
[0:43] they won't have much control over the
[0:45] picture next up let's talk about what
[0:48] each symbol means first off there are
[0:50] parentheses square brackets and curly
[0:53] braces these are mainly used to tweak
[0:55] the weights of keywords when you put a
[0:58] hint word in parentheses it's weight
[1:00] becomes 1.1 you can use up to three
[1:02] parentheses at the same time when n
[1:05] parentheses are used the weight of the
[1:07] hint is 1.1 to the power of n curly
[1:10] braces are also for increasing the
[1:11] weight of a hint curly braces are used
[1:14] to increase weight too each layer of
[1:16] curly braces increases the weight by
[1:18] 0.05 the adjustment with curly braces is
[1:21] much smaller than that with parentheses
[1:24] square brackets are used to decrease
[1:26] weight one square bracket makes the
[1:27] weight of a hint word 0 point 9 times
[1:30] less three square brackets make the
[1:32] weight 0.7 to 9 the easiest way to
[1:36] adjust weights is to add a colon after
[1:38] the hint word in parenthesis and then
[1:40] fill in the desired weight value
[1:42] directly after the colon It's
[1:43] recommended to set this value between
[1:46] 0.3 and 1.5 for example if you enter
[1:49] cherry blossoms kittens green leaves and
[1:52] set a different weight value for cherry
[1:54] blossoms you can control the proportion
[1:56] of cherry blossoms in the whole picture
[1:58] the pointed bracket are mainly used to
[2:01] call Laura the format is Laura trigger
[2:04] word weight value after calling Laura
[2:07] images with specific features can be
[2:09] generated it should be noted that in
[2:11] tensor Laura tritter words can be
[2:13] directly used with one click which is
[2:16] very convenient underscores act as a
[2:18] link to make the keywords more closely
[2:20] connected for example a milk cake might
[2:23] be understood by stable diffusion as a
[2:25] glass of milk and a cake but if milk and
[2:28] Cake are connected by Under scores then
[2:30] stable diffusion will understand milk
[2:32] and cake as a whole next let's check out
[2:35] some Advanced syntax now how do we
[2:38] control when a prompt kicks in first off
[2:40] you can use a combo of square brackets
[2:43] and colons there's a number after
[2:45] cologne it means the prompt starts
[2:47] rendering from that point a double
[2:50] cologne that means the prompt in
[2:51] parentheses renders till that point when
[2:55] there's one prompt it's like this but
[2:57] there can be two prompts in parentheses
[3:00] like this the two prompts are separated
[3:02] by a cologne so for the first 70% of the
[3:06] rendering stone is on and for the next
[3:08] 30% flower is on if the two prompts in
[3:12] parentheses are separated by a vertical
[3:14] line like this it's alternate sampling
[3:17] so red hair and blue hair prompts
[3:20] alternately and you'll end up with red
[3:22] and blue hair after you get the hang of
[3:24] the basic syntax let's check out the
[3:26] recommended way to describe the overall
[3:28] prompt words picture quality and
[3:31] painting style words have a big impact
[3:33] on the overall look of the picture so
[3:35] it's a good idea to write them on the
[3:37] top I've got some tips on picture
[3:39] quality prompts mostly they work for
[3:42] everything but for different styles of
[3:44] pictures there will be some specific
[3:46] quality words too like pixel style Pixar
[3:49] style ink painting style and so on add a
[3:52] comma at the end of this and you can
[3:54] start a new line if you want then
[3:56] there's the description of the main part
[3:58] of the picture my character age
[4:00] hairstyle hair color what they're doing
[4:03] and so on the more detailed you are the
[4:06] more accurate the generated picture will
[4:08] be add a comma after writing and start a
[4:11] new line next up is the description of
[4:13] the lighting in the environment like a
[4:15] Snowy Evening or a Sunny Meadow add a
[4:18] description of the lighting used at the
[4:20] end of this the last line is for adding
[4:23] the trigger word of the Laura you want
[4:25] to use after positive prompts fill in
[4:28] the negative prompts
[4:30] generally just filling in some general
[4:32] negative prompt is fine if you're using
[4:35] the flux model it won't let you fill in
[4:37] negative prompts on your own when you're
[4:39] using tensor's classic workbench you can
[4:42] just fill in the prompts in the language
[4:44] you know after that click the
[4:46] translation button right below when
[4:49] you're out of ideas for the prompts you
[4:51] can choose to generate them randomly or
[4:53] by recognizing pictures oh and you can
[4:56] also check out the posts recommended on
[4:58] the homepage to get some
[5:00] inspiration there are a bunch of tensor
[5:02] models for you to pick from all of
[5:04] tensor's models are online no need to
[5:07] download they're one click and don't ask
[5:10] too much of your graphics card best of
[5:12] all they're free you can get free
[5:14] credits every day by doing some stuff
[5:17] click the website link under the video
[5:19] to give it a try any questions or things
[5:22] you want to hear about next time just
[5:25] subscribe me in the comment section see
[5:27] you