Full transcript
0:00Claude Fable 5.1 is here, and I have
0:02been actually genuinely impressed while
0:05testing it out. So, today we're going to
0:07go through some of those tests, and I'll
0:09walk you through what is good about it,
0:11what it's not so good at, and how I
0:13recommend incorporating it into the
0:15writing process. Now, as is the case
0:18with all of these test prompts that I
0:20use with these models, what I'm looking
0:22for is how well does it do out of the
0:24box without any sort of manipulation or
0:28anything like that. Because we all know
0:30at this point that it's easier to get
0:32better results if you, you know, write
0:35the first draft, then have another model
0:37analyze it, then have another model like
0:39go in and surgically correct some
0:40things, and all of that. We're not doing
0:43that. We want to know how well can it do
0:45with just a single prompt, often a
0:47simple prompt, without any kind of
0:50tricks or anything like that. So, let's
0:53get into it. The first one is a logline
0:55prompt where I ask it for 20 epic
0:57fantasy loglines. And by the way, I use
1:00epic fantasy here for all of these
1:03prompts, not just because it's a genre
1:06that I write in and understand better,
1:08but also because it is one of those
1:11things that AI tends to get overly
1:13dramatic on and
1:16a little less natural sometimes. It's
1:18harder to get good results from the
1:20fantasy genre with AI than it is in,
1:23say, contemporary romance. So, we're
1:26going to be going through this and
1:28seeing what we get. So, a couple of
1:30these We're not going to read all 20 of
1:31these, but I did find that a lot of
1:33these actually were decent ideas. Let's
1:36start with this one. The Unmaking of
1:37Kings. When a disgraced court historian
1:40discovers that the realm's thousand-year
1:41peace was written into existence by a
1:43forgotten spell, she must rewrite the
1:45history books before a warlord burns the
1:48last copy. But every edit she makes
1:50erases someone she loves from ever from
1:53having ever existed. I actually think
1:55this one is okay. I we would need to
1:57change a few things because like why
2:01would it be someone that she loves for
2:03instance, but maybe just a racist
2:04someone or has some kind of other
2:07consequence that we could insert in
2:08there, but not too bad. I also thought
2:10this one was intriguing. A prophesied
2:12hero who has failed eight previous
2:14incarnations must break the cycle of
2:17reincarnation by defeating the dark lord
2:19for good until he learns the dark lord
2:20is the only one who has ever been on his
2:22side. Now of course this doesn't explain
2:24why the dark lord has been the only one
2:26on his side. But that was enough of
2:29intriguing mystery to me that I thought,
2:32well, this might be something that I
2:33could explore and get a decent story
2:35idea out of it. And then I also really
2:37liked this one here, the tithe. In a
2:39kingdom where every firstborn is
2:41sacrificed to keep the sun rising, which
2:43is a terrifying prospect, a mother who
2:46hid her child for 12 years must stop the
2:48high priest from finding him and
2:49discovers the sun never needed the
2:51sacrifices at all. That to me is an
2:53excellent epic fantasy premise. Um you
2:57could probably take that in a couple of
2:59different fantasy sub genre directions.
3:02It seems a bit dark so like this could
3:04be like a grim dark or a sword and
3:06sorcery, but yeah, I thought that was
3:08pretty good. We did have a few that were
3:10pretty obvious. Like I always get a
3:12cartographer in here for some reason.
3:15Um and I just think that's a little bit
3:18at this point I've seen it so often that
3:19it just like I just roll my eyes every
3:21time I see it again. Uh but overall,
3:24there were definitely some intriguing
3:26ideas among this batch so it's
3:27definitely a good one for brainstorming.
3:29Moving on,
3:30the outline test prompt. This is where I
3:32give it my full plot module 40 chapter
3:35outline template and a story premise.
3:39Uh there isn't a whole lot of detail in
3:41the story premise because I want to
3:43see what it comes up with without too
3:45much direction. So I just give it that
3:47and it spits out an entire outline and
3:51this is the longest result that I have
3:53here. This thing's like thousands of
3:55words long. It did start out with a nice
3:57little intro here.
3:59Uh just kind of like giving
4:01solidifying a few things about the story
4:03Bible and stuff like that. I thought
4:05this was kind of fun. Uh gives me the
4:07cast and the main cast and all of that.
4:10There are a couple of AI common names in
4:14here like Corbane or or Corvin I see a
4:17lot. Aldric
4:19uh Vale, but also a few that I wouldn't
4:23expect to see from AI as much. But a
4:25couple of things that I've noticed that
4:26AI does with this particular prompt is
4:29that I now look for as like signs of is
4:32it doing okay or not. Uh one of those is
4:36the prologue. So a lot of the other
4:38models when I test this thing will have
4:40just the barest hint of a prologue.
4:44There isn't actually any story in it.
4:45There often isn't even a character in
4:47it. It's just like a single image uh
4:51just that it describes, right? And in
4:54this case it didn't do that. It thought
4:56a little bit more carefully and actually
4:57has
4:59some characters and then this is an
5:00actual scene here. So I was pretty
5:03impressed with that.
5:04Uh then it moves into the first chapter
5:07and gives us an opening image. Uh this
5:10is all based on the template that I give
5:11it. An obstacle,
5:13a save the cat moment, and a glimpse of
5:14the flaw. And then I outlined what all
5:16of those are. Uh it started with a
5:19premise that I haven't seen before uh
5:21where it had like the this character uh
5:25this character is like a guard having a
5:28a shift uh I mean they're they're a
5:30prince,
5:32but they're having like a shift as a
5:35uh at a gate to like oversee this gate,
5:38you know. It's kind of hard to explain,
5:40but I thought it was actually pretty
5:42good. And the other big thing that I
5:43look for when going through with
5:45outlines is the save the cat
5:47cuz so many of the models I've tested
5:49this prompt on just kind of shoehorn the
5:52save the cat moment in there and it
5:54really has nothing to do with the rest
5:56of the plot and it's always a moment of
5:59kindness.
6:01Uh like Aladdin giving the bread to the
6:04two little kids in the Disney Aladdin
6:06movie. Like that's a classic save the
6:09cat moment and it's usually something
6:11like that even though a save the cat
6:12moment does not have to be an act of
6:14kindness. It just has to be anything
6:16that makes the character look compelling
6:18to the reader. And in this case it was
6:20still an act of kindness
6:22but unlike most of the other things that
6:25I
6:26have uh the
6:27the other results I've tested from other
6:29models, this one did a good job of
6:31actually integrating it into the plot
6:33line a little bit. So it like has to do
6:36with the scene that he's in. It uh
6:39it's definitely something that fits and
6:42that was important to me. I wanted to
6:44make sure it did that. I then just kept
6:46reading for a while um and I was pleased
6:48to see another thing that often tends to
6:51happen is especially when you give it a
6:52template that each scene, each chapter
6:55will feel a little disjointed and a
6:58little uh you know like they're not
6:59quite connected to each other and I was
7:01pleased to see that that was not the
7:02case here. Fable 5.1 was very good at
7:06making the story feel cohesive and I
7:09thought it was really good. And so I
7:10kept reading for quite a bit here until
7:13we got at least to act two and I'm I'm
7:16pleased to see the the whole thing felt
7:18like an actual story. Like I was reading
7:20it, I was actually getting into it a
7:21little bit and and uh enjoying seeing
7:24where the plot was going. That's always
7:26a good sign when that happens. So
7:28overall, really good job. Could it have
7:30been better? Yeah, of course but I think
7:33it could have only been better with a
7:35little bit more human guidance. I
7:37intentionally did not give it a whole
7:38lot of human guidance here and so had it
7:41had a little bit more human guidance for
7:42me, it would have been so much better.
7:44So, overall, really good job on the
7:46outline and it did actually go through
7:48and go all the way to chapter 41, the
7:51epilogue,
7:53and there's plenty of information for
7:56each chapter. Like you can see, like
7:57there's a
7:58pretty solid amount of information here,
8:00even at the end, for each chapter here.
8:02So, it did a great job.
8:03Next, this was the the big one, is
8:07testing out the prose and I'm pleased to
8:10say it did really well at this as well.
8:12Now, ignore the fact that there are m
8:14dashes, you know, you're you're always
8:16going to see that with AI, but let's
8:18actually read through a few paragraphs
8:21here. Halder did not know this, he knew
8:23only the mountain ahead and the gate
8:24somewhere on the lower face and the
8:26wind. And behind him, he did not want to
8:28look and looked. He knew the hungry.
8:30They were on the horizon now, where an
8:32hour ago there had been only the long
8:34red glare of the sun. Two dozen at
8:36least, perhaps more. It was hard to
8:38count them because they were the same
8:39gray wasteland, the gray of ash and old
8:42bone, and only their movements betrayed
8:44them. That movement, there was nothing
8:46in the world that moved like the hungry.
8:47They came on in fits, a lurch and a
8:49stillness, a lurch and a stillness as
8:51though each of them were being pulled
8:52forward by a string that someone kept
8:54forgetting to hold. Which, you know,
8:56AI's usually terrible at metaphor,
8:58that's actually a decent one right
9:00there. Not really a metaphor, but
9:02you know, an image there. A head would
9:04jerk toward him, a shoulder would drop
9:06and drag. They did not tire and they did
9:08not hurry and they did not need to
9:09because everything that ran from them
9:11eventually stopped. He knew the laws of
9:13Arland as every child of the mountain
9:14knew it. When the hungry are within
9:15sight of the walls, the gates do not
9:17open, not for a king, not for a mother
9:18with a child, not for the last living
9:20cartographer of the third expedition
9:22whose name was already crossed out in a
9:23ledger he would never see. Halder turned
9:25his face toward the mountain and ran
9:27anyway.
9:28I don't know if you could tell just from
9:29me reading that out, but it flowed
9:32really nicely. I felt very natural
9:35reading that out loud. I
9:37There are a couple of things here and
9:39there that are sort of AI typical, but I
9:42was finding very, very few AI-isms just
9:46in that read-through
9:47compared to what I'm used to seeing.
9:49Now, I will say just visually, if you
9:52look at this, there's large paragraphs,
9:54big chunks of text. It's not the the
9:57most visually appealing, and I think it
9:59could have broken things up a little bit
10:00better.
10:01But, I actually think the prose here is
10:03pretty darn good overall. And this was a
10:06very simple prose prompt. I gave it just
10:10a description of the scene and told it
10:13to go. No style information, nothing
10:15like that.
10:16And overall, I think it did a great job.
10:18Now, this next one is the same scene,
10:20but I do give it some style information
10:22and additional style tips and stuff like
10:24that. Let's read that same section here
10:26with with this prompt. He crested a low
10:29rise and let himself look back. First of
10:30all, you can see visually it it's
10:32breaking up the paragraphs a little bit
10:33better. So, that's good. Figures dotted
10:36the horizon, scattered, spread wide
10:38across the flat the flats like seed
10:40thrown by a careless hand. They moved
10:42wrong, a step, a stop, a lurch forward
10:44as though each one had to remember how
10:45legs worked between strides. The light
10:48bled out of the sky behind them and
10:49turned their skin the same dead gray as
10:51the rock, so that for one breath he
10:52could pretend they were only stones.
10:54Then a stone moved, then another. He
10:56counted 10, 15. He lost the tally past
10:5920 and stopped counting because counting
11:00changed nothing. Two dozen at at the
11:03least, and those were the ones he could
11:04see. Too close, far too close. Every
11:07child in Varlen learned the rule before
11:08they learned their letters. When the
11:10hungry came within sight of the walls,
11:11the gates stayed shut. Didn't matter who
11:13stood outside, didn't matter if they
11:14screamed or begged or wore the seal of
11:16the Surveyors on their coat. The watch
11:18would look down from the parapet, and
11:20they would keep the gates shut, and the
11:22city would go on living behind them. His
11:24hand closed around the compass until the
11:25brass bit into his torn palm. He turned
11:28east and ran. I actually think it did
11:30some things better and some things
11:31worse. I noticed there were a number of
11:34small metaphors here and there that felt
11:36very AI-ish to me. Uh there were a
11:39couple of moments where it was like
11:41almost too punchy with the the
11:44sentences. He counted 10, 15. Uh or
11:47where was it? Then a stone moved, then
11:49another. Uh they moved wrong is like the
11:52word using the word wrong in this
11:54context is something I see AI do now and
11:56then. Uh it fits the scene, but I I I
11:58feel like it's kind of common.
12:00And so overall, I think it's still good
12:03and it's definitely different. It
12:04definitely changed some things based on
12:06the style information I was giving it,
12:08but I'm not actually certain that it
12:10flows as well as the first one, which is
12:13interesting.
12:14All right, let's move on to the next
12:15prose prompt and this one is a scene
12:18that has dialogue in it between uh the
12:21man Lathen and his father, the king.
12:23Let's read a little bit of this section
12:26right now. King Store sat where he
12:28always sat on the black seat carved from
12:29the mountain's own heart, a ledger open
12:31across his knees. Firelight from the
12:33braziers
12:35found the gray in his beard and the iron
12:36in his shoulders. Argal stood at his
12:38right hand as he always stood, thin as a
12:41stripped branch, hands folded into his
12:43sleeves. See, there are little tiny
12:46metaphors like that that aren't
12:47egregious,
12:49but kind of take away from it a little
12:52bit.
12:52"A man died at the gate tonight," Lathen
12:54said. His voice came out rougher than he
12:56intended. His arm still ached where the
12:57guards had held him. He came out of the
12:59waste alone. He had no weapon. He was
13:01begging us. He was close enough that I
13:02could see his face, father. I could see
13:04his" He stopped, swallowed, "and we
13:07watched. The whole city stood on the
13:09wall and watched the hungry take him."
13:11Store turned the page. And I got to say
13:13I love that moment. That um that's a
13:17good show-not-tell moment where it's
13:19showing us that Store does not care.
13:22He's still just doing his thing. He's
13:24turning the page in his book that he's
13:26reading. "That's the second one this
13:28month, Lathen said. The second. The
13:29woman with a child and now him. Do you
13:31remember the woman because the people
13:32do? They talked about it in the lower
13:34markets. They talk about the sound she
13:36made. The law is the law, Storr's voice
13:37was flat as slate. The gates do not open
13:40after dusk, not for one man, not for a
13:41hundred. You make the law. Lathen took
13:44three steps up to the dais before he
13:45realized he'd moved. It isn't carved
13:47into the mountain. It wasn't handed down
13:49by the gods. You spoke it and you can
13:51unspeak it and a man would be alive
13:52tonight if Prince Lathen Ar- Argol's
13:55voice slid in soft as oil.
13:58Another metaphor there, but this one I
14:00would leave. It's definitely
14:01appropriate.
14:02Uh he inclined his head the way one
14:04might to a child who had wandered into
14:06the wrong room room. You are grieved.
14:08That is to your credit. Grief is a noble
14:10thing in the young. A small patient
14:12smile. But the laws of Vrolden were not
14:14written care- carelessly. Every one of
14:17them exists for a reason and the reason
14:18is always the same, the safety of the
14:20city of the many. Your father bears the
14:22weight so that others need not. Uh
14:24anyway, we could keep going but overall,
14:25again, I think this is pretty solid. I'm
14:27not like blown away by it like I would
14:31be from a skilled human writer, but this
14:33is some of the best prose I've seen from
14:35AI so far. Definitely a solid baseline
14:38and we could only take it up from here.
14:40The next one is editing.
14:43Uh so I give it a section that was
14:44written by an older AI model and it's
14:48the same scene but written by an
14:49inferior AI model. In fact, it's Claude
14:522
14:53that initially wrote it and I asked it
14:55to improve it. And I'm not going to go
14:56through the whole thing, but basically I
14:59can I can confirm this was a significant
15:01improvement over the original draft. So
15:04if you want to take some like maybe some
15:06writing that you wrote a long time ago
15:08that you know isn't great and you want
15:10to give it to AI to improve, I actually
15:13think Fable 5.1 is a good one to try and
15:16do that with. The next one is add
15:18headlines. This is where I ask it for 20
15:20headlines and again, I'm not giving it
15:22too much instruction. You can get really
15:24good copywriting out of AI these days by
15:26giving it a whole bunch of copywriting
15:28templates, headline templates, that sort
15:30of thing. But again, I'm not interested
15:32in what it can do with all of that info.
15:34I'm really interested in what it can do
15:36by itself, just by asking it to write me
15:39headlines and really nothing else.
15:42So, I gave it a description of the story
15:45and asked it to write 20 headlines for
15:48Facebook and it gave me a couple. They
15:50threw this king's son to the hungry. He
15:52didn't die. Everything Latham was taught
15:54inside the mountain was a lie. His
15:55father had exiled him. The reason is
15:57worse than treason.
15:59The prince was cast out to die. What he
16:00found instead will bring the mountain
16:02down. One betrayal sent him into the
16:04wastes. One secret will bring him back.
16:07That was not too bad. And then it gives
16:09me a whole bunch more. The secret have
16:11no thoughts of their own. There are
16:12still humans outside the mountain. That
16:14one's pretty good actually. The king
16:16said there weren't. The rebels knew his
16:18name before he told them. That was the
16:19first warning. These are pretty long,
16:22but actually they're not that bad
16:24compared to what I'm used to seeing from
16:26this prompt. A lot of AI models will
16:29just have terrible ad headlines here.
16:33Some of these I might actually consider
16:34using or adapting, trimming, whatever.
16:38They're not They're not too bad. Not at
16:39all. And I even gave me some quick notes
16:42for testing, which is nice.
16:44The next one is an email newsletter
16:45prompt where I ask it to I give it a
16:48description of the book and say it's
16:50going to come out on this date. Please
16:52write a announcement
16:55of the pre-order for that book. And it
16:57gave me this. Hi everyone, I have news
16:59I've been waiting a long time to share.
17:00My new novel The Crown of Hunger is
17:01finished and it releases on December
17:0321st. This one is an epic fantasy with
17:05the bones of the Count of Monte Cristo,
17:06a story of betrayal, survival, and the
17:08slow burning reckoning.
17:10Latham is the son of a king, yada yada
17:11yada. It gives a description. This
17:13paragraph might be a little bit too
17:14beefy for an email, but it does
17:17accurately sum everything up. Then
17:20Latham meets a band of human rebels and
17:21learns that everything he was taught
17:22inside the mountain was a lie. You can
17:24pre-order it now. Boom. Thank you for
17:26reading, for following along, and for
17:27your patience while I wrote this one. I
17:29can't wait for you to meet Latham. Until
17:30December, signed your name.
17:32Uh
17:33nice and short and sweet, which is how I
17:35like a lot of my newsletters. And
17:38doesn't like I I don't feel like it's
17:40that wordy. Uh a little bit here and
17:42there perhaps. I could definitely trim
17:44it down, but this is decent. This is
17:46decent. Uh I could definitely use that
17:48as a starter.
17:50All right, the last one that I do is an
17:51SEO article super prompt to see how well
17:53it does with uh non-fiction. I ask it
17:56for 4,000 words on the topic of how to
17:58write a fantasy novel, and it gave me
18:004,200 here. So, right on the money there
18:03with the amount of words. And it starts
18:06out, "Fantasy is one of the most
18:07rewarding genres to write and one of the
18:08easiest to get lost in. You have to
18:10invent a world, fill it with people,
18:11tell a story worth reading often across
18:13uh 100,000 words or more. That's a lot
18:15of moving parts for a first-time
18:16novelist. This guide covers how to write
18:18a fantasy novel from the first idea to
18:20the final edit. You'll learn what
18:21defines the fantasy genre, which tropes
18:23readers expect, and the step-by-step
18:25process for building your world,
18:26characters, and plot. It also includes
18:28lessons from my own first fantasy novel,
18:30which most of which I learned by doing
18:32things wrong. If you've read a dozen
18:33fantasy books and played years of
18:35tabletop games, but never written more
18:36than short stories, this is for you.
18:38This is a bit of a wordy intro. Uh
18:41my background is in SEO writing, and I
18:43know that like we usually want the intro
18:47to be like the length of this first
18:49paragraph. And you want to get straight
18:51to the info as quickly as possible. But
18:54again, it doesn't feel too bad. Like I
18:57could easily trim this up myself. Then
18:59it goes into, "What is a fantasy novel?"
19:01And we've got a nice little intro to
19:03that section. And uh as we go down, you
19:05can see everything is very scannable,
19:07making good use of
19:09uh Um, bulleted lists, bold text,
19:13italicized text. Uh, I really actually
19:16really liked some of this list of
19:17fantasy subgenres that it gave me. Um,
19:19high fantasy, epic fantasy, low fantasy
19:22with examples on each one. And knowing
19:24the fantasy genre well enough, I can say
19:25that all of these examples are correct
19:27and these are good uh, examples of of
19:30each one. Common tropes, these are
19:33definitely common tropes. And once
19:34again, this is all very scannable. You
19:36can just look through this and you're
19:37like, yeah, this is this actually this
19:39looks visually like a good SEO article
19:42and it's definitely covering all of the
19:44different topics and things that I would
19:47want it to go for. So, all in all oh,
19:49also I like to look at the steps that it
19:52has here. And uh, it definitely did a
19:55good job uh, with these steps to to
19:57write a fantasy novel. The only thing I
19:58might change is actually give the steps
20:01numbers. So, make this one step one, the
20:03next one step two, etc. But, all in all,
20:06it did a pretty fantastic job at this,
20:08too. Maybe slightly wordy.
20:11And with SEO articles, you generally
20:13want to say what you mean quickly. You
20:16don't want to draw it out. But, overall,
20:19great job. Like I'm pretty impressed
20:21with it. I like it. Uh, this is pretty
20:22good. So, if you can't tell, I am pretty
20:25impressed with Claude Fable 5.1. Uh, I
20:28have really seen very few big red flags.
20:33The only
20:34red flag that I would say that there is
20:37is its cost. It is not a cheap model.
20:39And if you're using it heavily in your
20:42writing, you're going to probably be
20:44looking at 50 to $100 per book would be
20:47my estimate for how long that takes. Uh,
20:50it depends on how you're using it, where
20:52you're using it, and all of that. But,
20:54uh, I would expect at least that much.
20:56But, certainly a decent model to be
20:58going with and uh, it's it's especially
21:01good, I think, at just understanding
21:04your story as a whole
21:06and not making weird kind of out of left
21:08field assumptions or changes to your
21:12outline or what have you. Like it it
21:14seems to have a much better logical
21:16grasp on the the topics that we cover in
21:21creative writing. And that kind of makes
21:23sense, right? Because models like this
21:25are really good at understanding like
21:27entire code bases to to create a you
21:31know, software and vibe code and stuff
21:33like that. Which I have been getting
21:34into recently a lot as well. And so
21:37being able to
21:39kind of see and get a holistic view of
21:41large amounts of data seems to be
21:44something Fable is incredibly good at.
21:46So those are my thoughts. Definitely you
21:47should go out and test Fable 5.1 for
21:50yourself and see what you think. And if
21:51you got this far you should definitely
21:53check out the Storyteller AI system
21:54where we have our own product now. Used
21:56to be that if you subscribe to
21:58Storyteller
22:00you got the instruction but you also had
22:02to subscribe to Claude or Novel Crafter
22:04or you know, other things. Now we have
22:05our own system so you can do everything
22:07inside of the Storyteller community and
22:10yeah, it's just a great time. We have
22:11hundreds of people at this point and
22:13hundreds that have successfully written
22:15a book in a month. We do them quarterly
22:17channel challenge around that. And it's
22:19just a good time. So if that's
22:21interesting to you, make sure to check
22:23that out down below and I'll see you
22:24there.