Full transcript
0:00Okay, in this video we're going to go
0:01through another automation. This time,
0:03this is a redo of a previous automation
0:05I done, but this time I pulled out all
0:07the stops and tried to make this as
0:09perfect and as advanced as I possibly
0:11could. My goal with this and the
0:14automations that follow it is I want to
0:17get to a point and this is going to
0:18freak out some people. I want to get to
0:20a point run these automations without
0:23any human input and it would actually
0:26give you something that is sellable that
0:28would actually get a like four star or
0:32higher average on Amazon and people
0:35wouldn't know that it's AI. Now, that's
0:36a tall order and I will always say that
0:40doesn't matter how good AI gets on its
0:43own. AI plus a human can always get it
0:45better. So, this is not saying that I
0:47want to flood Amazon with a bunch of
0:48books that are purely AI written with no
0:51human input. What I want to do is be
0:53able to get the book get it as good as
0:55possible so that the human can make it
0:58even better. Because if it's if the AI
1:01can only get you 50% of the way there
1:03and then your efforts can get an
1:04additional 50% you get it to 100%.
1:07That's fine. If the AI can get you 90%
1:09of the way there, then with your efforts
1:11you could get it to 150%. If that makes
1:14sense. That's the idea behind this. I
1:16really want to get it to as good as
1:18humanly possible. Kind of a odd phrase
1:20there, but yeah, you get the idea. So,
1:22the way this is working is this workflow
1:25that we're going through today is my
1:27brain dump to story dossier automation.
1:30I've done one of these before, but this
1:31one I really went back to the drawing
1:33board and built it from the ground up
1:36trying to make this as good as possible.
1:39And first of all, I've designed it to
1:41work with my uh templates. So, you see
1:44these this first part of the automation.
1:47First of all, you say execute workflow
1:50and it will pop up something like this
1:52where you can put in your brain dump
1:54info. Just whatever you already know
1:56about this book that you want to do, you
1:57put it in here. That way, it has that
1:59context. And then from there, it runs
2:01through these templates and grabs them
2:03from various documents. Now, you might
2:05be wondering, where on earth do I get
2:07these templates? Well, I have some of
2:09them built out myself inside of my story
2:11hacker gold group. I've written these
2:12genre guides that all have these
2:15templates associated with them. You can
2:17get these at no extra cost when you join
2:19the membership. They are the book itself
2:22is available for $5 on Amazon for each
2:25of these. I've got 30 of them written so
2:28far, even though you don't see all 30 of
2:30them here. They'll be here very soon. 30
2:31of them written. And my plan is to get
2:34at least 100. So, you could grab these.
2:36If we look at like the tropes dock here,
2:38for example, here we have the list of
2:40major tropes like the masquerade. This
2:41is the bread and butter of the genre.
2:43Simply put, the masquerade is the veil
2:44that keeps the magical world hidden from
2:46the mundane one. And it goes through and
2:47basically goes through every single
2:49major trope of this series. And so we
2:52gather all of these docs that we need
2:54and we put them in this automation. We
2:58just take the document ID, stick it
3:00right here, and then run through the
3:02automation. And then we also grab a
3:04blank document in which to put the
3:06finished story dossier later. So, it'll
3:09be blank when you start, but eventually
3:11we'll get to it and it will fill it up.
3:13I'll show you an example of a fleshed
3:16out the story dossier at the end of this
3:18video. All right. So, then we run
3:20through a series of AI automations. And
3:23you can tell we've got a bunch of them
3:25here. Some of them are more powerful
3:27than others. And you'll see one of the
3:29things I'm doing now to try and clean
3:31these things up is here where you
3:35identify the model that you're using.
3:37I've labeled it whether you should use
3:39an inexpensive or a medium or a highly
3:42powerful model for this step. And for
3:44instance, this one I tell you to use an
3:46inexpensive model because all this step
3:48is doing is looking at the tropes and
3:50identifying what the genre of those
3:52tropes are. So I I give it all of those
3:54templates and then say identify the
3:56specific subg genre for these templates
3:57and output just the name of the subg
3:58genre, nothing else. It's a very simple
4:01thing that even the smallest model can
4:03do. So, in this case, I am using Gemini
4:063 Flash for this step to keep it really
4:08inexpensive. The reason I do that is
4:10because I sometimes want to plug in the
4:13name of the genre elsewhere down the
4:15line, but that genre can vary depending
4:17on what we're doing here. So, I just
4:19want it to be able to look at everything
4:21that it's got, say, okay, this is a
4:23genre, and then later when I need to
4:25bring in the name of the genre into the
4:27prompt, I can just have it identify the
4:30output of this step. Does that make
4:31sense? Hopefully, it makes sense. All
4:32right. Then we go into the brainstorming
4:34pitch. And these three prompts here are
4:38kind of a set here. What this one does
4:41is it brainstorms five pitches based on
4:44the genre and based on your brain dump
4:46info. So whatever you put into the brain
4:48dump info, even if you've got a pretty
4:50well-formed synopsis there, it will
4:52actually give us five additional pitches
4:56based on this hook rubric that you can
4:59see here. I've got a pretty extensive
5:00rubric for what makes a good hook. It
5:04keeps that in mind and then goes through
5:05and says, "Above, you will find a
5:07variety of tropes and a plot template
5:08for this genre colon." And then it pulls
5:10in the name of the genre that was
5:12generated in the previous step. You may
5:13also find an author's brain dump for a
5:15book or series. Your highly complex task
5:17is to brainstorm five story pitches for
5:19a new book in this genre. Make sure that
5:21it has an original setting and premise,
5:23but still sticks to the tropes and
5:24overall plot outlined, as well as adhere
5:27to the author's brain dump. Use the hook
5:29rubric that I have above to formulate
5:31your pitches. Format each story pitch in
5:33markdown as follows. And then I have it
5:36constructed this way. All right. And
5:37then it runs a specific or this one
5:41right here. It runs a specific prompt
5:44designed to identify which one is the
5:47best, which it can usually do pretty
5:50well. By the way, for this brainstorming
5:52pitch, I have I recommend a powerful
5:54model like Claude Opus 4.6 six or Gemini
5:563 and probably at least something on the
5:59level of Gemini 3, maybe Claude sonnet
6:02or any of those for this step where
6:05you're picking the best. I have found
6:06that there is a remarkable amount of
6:08consistency when you ask the AI what it
6:10thinks the best is across multiple
6:12models. I found that that's like 80% of
6:15the time they're picking the same one.
6:16So, I think that's pretty cool. So,
6:18it'll pick the best one here. And then
6:20when it goes through this prompt,
6:23there's a lot of stuff that I have it
6:25do. Like, I have it explain exactly why
6:27it thinks that that pitch is the best
6:29one. Goes through this whole process.
6:31The output's kind of long. So, I don't
6:33necessarily want all of that craziness
6:36around it. So, I then run with an
6:38inexpensive model. I have it select just
6:41that one pitch that was picked in the
6:43last step and just output the text of
6:46that pitch verbatim. Just so we have the
6:48one text that we can now plug and play
6:51into the next few steps. All right.
6:54Okay. This is where and all of that that
6:56we've done is very similar to what I had
6:58in my previous brain dump to dossier
7:00automation. Although I have refined the
7:02prompts a little bit better. They work a
7:04little better. But this next part, these
7:06next like couple of steps are unique.
7:09Quick pause. If you're an author who
7:10wants to start writing more books, I put
7:12together my complete AI prompt pack that
7:14helped me and many others in my
7:16community write a book in a month. It's
7:18completely free. Hit the link in the
7:19description. Join the free weight list
7:20for Story Hacker Gold. Then you just
7:23answer a couple quick questions and I'll
7:25send you the full prompt list plus a few
7:27extra free goodies. I'll also email you
7:29when I open the paid Story Hacker Gold
7:31group so you can join if it feels right.
7:33There's no commitment to join though,
7:35and you'll still be able to get all of
7:36the freebies you signed up for either
7:38way. All right, back to the video. So,
7:39first of all, let's start with this one.
7:41This one isn't so unique. We basically
7:43take that story pitch and create a story
7:44dossier. What that looks like is
7:47something like this. If we go in here, I
7:50give it the premise, which I got in the
7:52last step. I give it all of the tropes,
7:53the outline, the character, the world
7:55building, and the author brain dump, and
7:57ask it to create a story dossier with a
8:00list of all of the characters, the world
8:02building elements, the synopsis, and
8:04certain points on the outline that I
8:07want it to do. I wanted to create a
8:09whole thing that will basically list all
8:12of the different elements that we need
8:13for this story. All right. So, you know,
8:15it looks pretty good. But what I was
8:17finding when it did this before is that
8:20a lot of stuff in there either didn't
8:22quite make sense or there were a lot of
8:24like AISMs in there, especially with
8:26names. Like if it was coming up with
8:28names of side smaller characters or
8:30anything, you get all kinds of names
8:32that pop up in AI all the time. then
8:35went through a couple of critiques or
8:38checks. In fact, I think I'm going to
8:39rename these as checks. Emotional check.
8:42So, named because we want the AI to
8:46check on these things. So, here we have
8:48two checks, an emotional check and a
8:50character name check. The emotional
8:51check is looking through this and saying
8:53like, does it actually resonate? Like,
8:56is this an interesting idea? Basically,
8:57you will analyze the above story dossier
8:59for theme cohesion and emotional payoff
9:01quality as we want this to be an
9:03emotional gut punch for the reader no
9:05matter what genre we're in. Identify the
9:07primary themes and secondary themes in
9:09this dossier, not in the template. In
9:11your own words, no vague labels like
9:13love or hope without explaining how they
9:14show up for each primary theme. List any
9:16major moment where the theme should show
9:18up but currently doesn't or feels weak,
9:21etc., etc. And it goes through and
9:22basically runs this emotional check on
9:25the dossier. It then goes through
9:28another check. So, this one doesn't look
9:30at what the emotional check had. It's
9:31just looking at the dossier, and it is
9:33specifically looking for character names
9:35that are AI obvious, AI sloppy. We'll
9:38put it that way. All right. So, I give
9:40it the author brain dump. I give it the
9:42story dossier that was completed two
9:44steps earlier. And then I give it a list
9:46of names to avoid. This is just a list
9:49that I've built up over the years. Some
9:51obvious ones in here like Marcus and
9:54let's see, Nora. We got some some more
9:57fantasy ones like Thor, Kale, Kalin,
10:00Kayla. We got LRA in here. We've got
10:02Lara in here. Silus and Vain and Vesper
10:04I've seen a lot recently, so I added
10:06those in. Chen, Dr. Chen is in here a
10:09lot. Blackwood. Anyway, so you get a lot
10:12of these in AI, so I specifically call
10:15them out. But then I also say you will
10:16examine the characters in the above
10:18story dossier for uniqueness and name
10:20quality. Extract a character list for
10:22each named character list, the full name
10:24and ro, etc. Name AI telly check. Flag
10:28any characters whose first or last name
10:29is exactly any of the above names to
10:31avoid unless included in the author's
10:34brainstorm because maybe you want a
10:36specific one of those names that was
10:38mentioned. You know, I've got common
10:40names in there like John and Sarah. You
10:42might want to have a John or Sarah in
10:44your book. So, I make sure that is not
10:46looking at the ones that you asked it to
10:47put in. For each flag name, propose
10:49three alternative names that fit the
10:51character's culture, setting, and social
10:52status are memorable but not
10:54melodramatic, are unlikely to be
10:56overused in AI generated fiction and are
10:59not included in the list of names to
11:00avoid above. Briefly justify one to two
11:02sentences why each alternative is a
11:04better fit, etc. And it goes through and
11:06just gives a bit more instruction on how
11:09to find good names. Now, it's not going
11:11to be perfect. strongly recommend you go
11:14carefully through the completed story
11:16dossier at the end and make sure you all
11:18your named characters are names that you
11:20like and um aren't obviously AI. But
11:24once it's done that, it's now created
11:26out of these two checks, it's created
11:28two lists of instructions on things to
11:30improve. So at this point, I kind of
11:33want to implement all of that into this
11:35dossier. So I have it do a rewrite.
11:38Notice I say you can use an inexpensive
11:40model for this because this is actually
11:41a fairly simple prompt. I just give it
11:44the story dossier, the emotional
11:45critique and the name critique from the
11:47previous two steps and say using the
11:49text of the story dossier and the two
11:50critiques. I want you to implement the
11:52suggestions from the improvement plan
11:54into the dossier only implement the
11:56suggested changes and do not change
11:58anything else about the original
11:59dossier. Reproduce the entire dossier
12:01with the suggested changes made. So
12:03basically I'm asking it to implement the
12:05suggested changes but change nothing
12:07else and to reproduce the text. All
12:09right. And then I repeat that cycle.
12:12We're doing another check and then
12:13another rewrite. But in this case, the
12:16reason I had it do a rewrite before this
12:19logic check is because I want this logic
12:21check to be working with the latest
12:24version of the dossier. I you know we
12:27will uh presumably have already seen
12:30some improvements from the emotional
12:31check and the dossier rewrite or the
12:34emotional check and the character name
12:35check. So I want that updated version
12:38ready when we performed the logic check.
12:41Now I recommend using a medium to
12:42advanced model for this. You don't need
12:44to have an advanced model for these
12:46previous two checks, but for this one
12:48because we're talking about logic and
12:50implausibility, we definitely want it to
12:53be able to think hard about this one.
12:55And I did a whole video just about the
12:57logic check cuz it's a pretty
12:59interesting one. But basically, we run
13:01this thing. It checks for logical
13:04inconsistencies, plausibility issues,
13:06etc. Creates another improvement plan.
13:08And then this dossier rewrite number two
13:12implements that feedback once again. So
13:14it's pulling from this dossier rewrite
13:18and then adding the improvements
13:20recommended here in the dossier in the
13:22logic check into its second rewrite.
13:24Then that second rewrite gets dumped
13:26into a dossier document here and that
13:28concludes this automation. So what does
13:32this actually look like in practice? Uh
13:34at the end I'll show you an example of
13:36one. So in a previous video I had
13:39outlined a vigilante series that I
13:43wanted to do but in a fantasy setting.
13:45So I took all of those templates from a
13:47vigilante crime thriller and adjusted
13:50them to work in a fantasy setting. And
13:52then I brainstormed some ideas for what
13:54that would look like. Off camera, I then
13:56took those ideas and ran it through this
13:59exact story dossier. And this is what I
14:02ended up getting. So, it's called story
14:04dossier, the red node. Of course, that's
14:07just the name it gave it. I'm probably
14:09going to name it something else because
14:10that sounds too sci-fi. There is
14:12something in here called a node because
14:14this fantasy world does have some
14:16elements that are sort of influenced by
14:18a more technologyish thing, but that not
14:22really something I dwell on too much.
14:24And so this is the synopsis that I was
14:26given from everything that I had put
14:28together in my brainstorming of the
14:30series. A disgraced special ops soldier
14:33hunting for the only witness who can
14:35prove a government coverup discovers his
14:36former subordinate is currently being
14:38rendered into magical fuel. and the
14:40conspiracy that killed his unit now
14:42controls the entire city. Now, I will
14:44modify this a little bit because this
14:45sounds too much like a modern-day thing.
14:48So, instead of special ops, we could
14:50call him like government
14:52sanctioned
14:54assassin hunting for the only witness
14:57who can prove a government cover up.
14:59Discovers former subordinate is
15:01currently being rendered in magical
15:03field. The conspiracy that killed his
15:05unit sounds too military like modern
15:08military that killed his battalion maybe
15:12now controls the entire city. And then
15:14we have the full pitch. So that was just
15:16the log line. Here's the full pitch. Cor
15:18Morick just wanted to find Jace. For 3
15:21years he's been hunting through the
15:22vertical slums of Arch City, a
15:23metropolis carved into the walls of a
15:25mountain chasm. searching for the one
15:27surviving member of his black ops unit
15:29who can prove their government ordered
15:30them to massacre civilians to cover up
15:32illegal human experiments. Instead, Core
15:35finds a nightmare during a desperate
15:37raid on a shipment of blood nodes
15:39stabilized magical fuel drained from
15:42living husks. Core used a nose a node to
15:44survive and hears a stranger's dying
15:46scream, but the iron glass vial bears a
15:48batch number that leads him to an
15:49industrial slaughter house. This kind of
15:51like industrial age fantasy. There he
15:54finds Jace strapped into the husk
15:56machine. The rescue comes too late. Jace
15:58dies and in the extraction and resulting
16:00red node pulses in Coror's hand, it used
16:02to be his friend. Now Core isn't hunting
16:04for a witness. He's hunting for the
16:06light elf noble who ordered the
16:08experiments. The gray skinned enforcer
16:10running the kidnapping operations in the
16:12spans. And every corrupt watch captain
16:14who signed the orders to look the other
16:15way while the poor disappeared into
16:17processing labs. He has one brand,
16:19Grace, which gives him impossible
16:21balance and precision but zero
16:22firepower. He has a bandelier of
16:24dwindling natural node charges. Verman
16:26magic caught in the spans. And he has a
16:29dwarven investigator named Borin who
16:31started out hunting him but just had his
16:34own murder files sealed by the heights.
16:36The Ilidar Restoration Society promises
16:38to save humanity from a coming war by
16:40perfecting demons. Soldiers force-fed
16:43multiple brands through the blood nodes.
16:45That's something I came from my world
16:46building. Core has already been framed
16:48for one massacre. If he can't expose
16:50Lord Ilon's operation before his node
16:52supply runs out, he'll be blamed for
16:55another, and the only family he had ever
16:57had will have died to fuel the monsters
16:59that inherit the city. Core Promise to
17:01the Reader is a gritty fantasy noir
17:03vigilante thriller set in a vertical
17:05city of bridges and shadows, where magic
17:07runs on finite ammunition, and justice
17:09requires becoming the monster the system
17:11feared. Expect brutal interrogations in
17:13damp tunnels, a count mouse chase across
17:16swaying suspension bridges, and a hero
17:18who must choose between dying as a
17:20wanted murderer or proving that his
17:21government feeds the poor to create
17:22weapons. All right, so that's the core
17:25premise, which I think is pretty solid,
17:27especially knowing that a lot of this
17:29world building came from me, so I kind
17:31of have it all in my head. And then it
17:32lists the main cast and gives us just a
17:34sentence or two about each. So Cororic
17:36is our main protagonist, a disgraced
17:38former spec ops commander branded with
17:40grace. We're going to change this again
17:42to government sanctioned assassin
17:47commander. Branded with grace, kinetic
17:49manipulation, and balance. He operates
17:51from the deep shadows of the spans to
17:52avoid detection. He is driven by the
17:54weight of the dead, physically carrying
17:55the souls of his unit as ammunition.
17:58That's not quite right. So, it is going
18:00to still get things wrong. So, you want
18:02to go through this. I'll just remove
18:04that sentence entirely, I think. All
18:05right. All right. And then we have Lord
18:06Ilon, Investigator Borin, Thovikar,
18:10Sable, Old Doc Mole, and minor
18:12characters like Jace, Captain Goran,
18:15Auditor Krell. These are all names.
18:17You'll notice none of these names are
18:19particularly AI sloppy. Some of them I'm
18:22sure could be that I just haven't
18:24identified recently. But I look at all
18:26of these and I'm not seeing anything
18:28that particularly stands out as being AI
18:30sloppy to me. And then we have
18:32locations. So, it goes through the
18:34locations, gives us a sentence about
18:35each magic and technology. It has an
18:38explanation of what that looks like, and
18:40then part four is the outline where it
18:42actually gives us a couple of key
18:44moments in the story. I don't have it do
18:46the entire outline at once, but we do
18:48want to have the opening image, and then
18:50it'll we'll have the closing image here
18:52at the bottom as well because those are
18:53important to know like how is the
18:55character changed, what's the inciting
18:56incident, uh what's the act one twist,
18:59what's the midpoint, and what's the
19:01climax. These are the ones that I like
19:03to identify before I start outlining the
19:06actual outline. I have it jot all of
19:08those down as well. So that is the story
19:11dossier in a nutshell and what it looks
19:13like. Of course, there are going to be a
19:15few things that need tweaking, but
19:17overall I think this is absolutely a
19:19fantastic story dossier that is way
19:22better than what I had with my previous
19:24automation. Although it is more advanced
19:25and will use more of your AI credits.
19:27But hopefully this video has been
19:29interesting to you. If you want to get
19:30your hands on this specific automation,
19:31it's in my Storyhacker Gold group. Go
19:33ahead and join the wait list for that
19:35down below.