Latest / AI DTC WTF / Fable 5.1: What Actually Changed and What to Use It For | EP 13

Transcript

Fable 5.1: What Actually Changed and What to Use It For | EP 13 · AI DTC WTF

• 0:03
  If you work in DDC and you use AI
• 0:05
  and you're wondering
• 0:06
  what the F is going on every week,
• 0:08
  this is your podcast.
• 0:09
  The AI to see
• 0:11
  F podcast
• 0:12
  your home for tactical
• 0:13
  tips, strategies, and ideas
• 0:15
  that you can implement right now in your
• 0:17
  AI workflows to make your brand or agency
• 0:19
  more money.
• 0:20
  Welcome to another episode of AI, DDC,
• 0:24
  and this is a really
• 0:25
  an emergency episode.
• 0:26
  Sound the sirens.
• 0:27
  Get everybody together.
• 0:28
  Fable 5.1.
• 0:31
  What is going on with this?
• 0:33
  Let's talk about it.
• 0:33
  That's what we're here to get into.
• 0:34
  Today,
• 0:35
  I am joined by my illustrious
• 0:37
  co-host Thomas Moen and Will Sartorius.
• 0:40
  Guys, let's talk about it.
• 0:41
  What's what are your thoughts on it?
• 0:44
  What does it do?
• 0:45
  Well,
• 0:46
  is everybody speaking Claude language?
• 0:49
  What do we think about this?
• 0:51
  Thomas,
• 0:51
  what are your thoughts right away?
• 0:53
  Well,
• 0:54
  I have a lot of feelings,
• 0:55
  and I've actually I've been kind of upset
• 0:58
  about anthropic
• 1:00
  and club for the last 2 or 3 months.
• 1:02
  I've felt frustrated
• 1:03
  that it's getting dumber
• 1:04
  and it's just eating up on the tokens
• 1:07
  and all that kind of stuff.
• 1:09
  But I
• 1:10
  honestly,
• 1:10
  I felt a little bit in love again
• 1:12
  with fable 5.1, they will find was
• 1:15
  very greedy in terms of the token spend.
• 1:19
  And, you know, it would tell you
• 1:23
  10,000 of words
• 1:24
  to answer simple questions.
• 1:26
  And my impression of fable 5.1
• 1:30
  so far is that
• 1:32
  it is more concise and more to the point
• 1:35
  in the way
• 1:36
  it is
• 1:36
  communicating,
• 1:37
  and it's also not so token hungry,
• 1:40
  so it's a little bit cheaper to use,
• 1:45
  but I still use table
• 1:50
  one mostly and mainly to
• 1:55
  to plan what I'm going to do.
• 1:57
  And I'm still using Codex
• 1:59
  as the kind of the hands of what's
• 2:01
  actually doing doing the work.
• 2:04
  But fable 5.1 I'm I'm positive
• 2:09
  versus favorable.
• 2:09
  Five yeah,
• 2:10
  I would tend to agree
• 2:12
  net positive obviously anthropic.
• 2:14
  It's just really been
• 2:15
  botching their marketing.
• 2:17
  And it's just like obviously
• 2:19
  being the forerunner in the space
• 2:20
  is far more challenging
• 2:22
  than being the underdog.
• 2:23
  But by
• 2:25
  I think it was maybe a few days ago,
• 2:26
  they were saying, oh,
• 2:28
  we're increasing limits by 50%,
• 2:30
  but they were actually,
• 2:31
  you know, decreasing limits by 17%.
• 2:33
  And so like, you know,
• 2:34
  they have a marketing issue.
• 2:36
  So like for the marketers out there,
• 2:38
  you know,
• 2:38
  maybe talk to them, I'm
• 2:39
  sure there's great opportunities.
• 2:41
  But in any event, like,
• 2:42
  I think the thing that I found
• 2:44
  to be the most useful for fable 5.1 is
• 2:48
  we've built so many skills and,
• 2:51
  you know,
• 2:51
  marked on files or agents or whatever
• 2:53
  you want to call them.
• 2:55
  And oftentimes when I was running
• 2:58
  the skills, whether it's like creating a
• 3:02
  a pitch deck or a series of briefs,
• 3:06
  it would skip steps.
• 3:08
  And I always was trying to audit,
• 3:10
  you know, these skills and be like,
• 3:12
  why does Claude keep skipping steps like,
• 3:14
  this is so frustrating.
• 3:15
  And I'd have to say, hey,
• 3:17
  you know, to Thomas's point,
• 3:18
  it felt like it was getting
• 3:19
  a little dumber.
• 3:20
  Well,
• 3:20
  what actually
• 3:21
  was the most useful thing I found with
• 3:23
  5.1 was just using,
• 3:25
  you know, slash clawed API prompt audit.
• 3:28
  And I had it go through all of my skills
• 3:30
  to find redundancies.
• 3:31
  And then
• 3:33
  as soon as that happened, I reran the
• 3:37
  the prompt
• 3:37
  to generate briefs
• 3:38
  for one of the creatives.
• 3:40
  And it happened
• 3:40
  and I just
• 3:41
  and I ran it through fable,
• 3:43
  just regular fable.
• 3:45
  And it happened much faster
• 3:46
  and it was actually far better.
• 3:49
  So what I'm seeing is that, like,
• 3:52
  you don't necessarily
• 3:53
  need to use this model for everything.
• 3:55
  It's an expensive model,
• 3:56
  but ultimately
• 3:57
  like using it to audit
• 3:59
  some of your past work
• 4:00
  and just like cleaning up your skills,
• 4:01
  cleaning up your ages,
• 4:02
  cleaning up your markdown files,
• 4:03
  I find that to be extremely valuable.
• 4:06
  And it also has just like
• 4:08
  cut out a lot of like fluff.
• 4:10
  Like if I look through like a markdown
• 4:11
  file of mine
• 4:12
  or I look through a skill of mine,
• 4:13
  I just like,
• 4:14
  I can't really tell, to be honest.
• 4:16
  Like, you know, is this helping?
• 4:17
  If I delete this,
• 4:18
  is this going to be problematic?
• 4:20
  So yeah,
• 4:20
  it's it's literally
• 4:21
  just slash cloud API prompt audit
• 4:24
  and it will go through you, you know,
• 4:29
  command shift the skills or the files
• 4:31
  you want it to audit.
• 4:32
  And it made
• 4:33
  it made a pretty big difference.
• 4:35
  Great use case.
• 4:36
  So many people going.
• 4:38
  It's great to audit it
• 4:39
  because ever see
• 4:39
  this stuff changes so quickly
• 4:40
  I think that's an awesome idea.
• 4:42
  The other thing that I've been hearing
• 4:43
  a lot of people doing
• 4:44
  is talking about like,
• 4:45
  okay,
• 4:45
  being basically agnostic around
• 4:48
  which platform is using, but using,
• 4:51
  you know, saying, hey,
• 4:52
  this is going to root to this task.
• 4:53
  This is going to root to
• 4:54
  this task for cloud.
• 4:55
  It's going to root to this task for grok.
• 4:57
  Like what are
• 4:58
  how have you guys done
• 4:59
  this, built this
• 5:00
  like in this 5.1
• 5:01
  taking over more now or like
• 5:03
  how are you building a system
• 5:04
  where, hey,
• 5:05
  you're putting it in and saying, okay,
• 5:06
  I don't care
• 5:06
  what model you use to accomplish this.
• 5:08
  Use a combination of them,
• 5:10
  because that's what
• 5:11
  a lot of the dialogs around.
• 5:12
  Yeah, I think, you know,
• 5:13
  I'm I'm rooting more and more specific
• 5:16
  tasks to specific things.
• 5:17
  And I also actually honestly been quite
• 5:20
  impressed by my 4.6 and kind of the,
• 5:24
  the workhorse that is
• 5:26
  and it's it is so fast.
• 5:28
  It's really, really fast.
• 5:30
  So I find myself
• 5:32
  routing more
• 5:33
  and more tasks to different models.
• 5:36
  And I've spent a lot of my time
• 5:38
  just optimizing my whole setup of skills
• 5:41
  and MDP files and everything
• 5:43
  to be platform agnostic.
• 5:45
  So everything will just run smoothly on
• 5:48
  on all of the different kind of
• 5:51
  models that is out there.
• 5:54
  But just ask a question on that.
• 5:55
  So, so one is, is like
• 5:58
  grok is like, what are you using it for?
• 6:00
  And how are you assessing that?
• 6:01
  It's faster.
• 6:02
  Like, I just how do you do that?
• 6:04
  And then like the second thing is like
• 6:06
  how do you build a system
• 6:07
  of going through
• 6:08
  and like having a choose automatically.
• 6:10
  Like what
• 6:11
  what what's technically in between there?
• 6:13
  I mean,
• 6:14
  I sort of know,
• 6:15
  I think, but I'm just curious
• 6:16
  how you have it set up.
• 6:17
  Yes, absolutely. So,
• 6:20
  so first of all, I use grok
• 6:22
  a lot for small, simple coding
• 6:25
  or research tasks
• 6:26
  because it, again, is quite quicker
• 6:28
  than the other models, and I like it
• 6:31
  better than using
• 6:33
  the term model for from, from OpenAI,
• 6:36
  for example,
• 6:37
  or the opus agents
• 6:38
  when it's small quick tasks in terms
• 6:42
  of how you get the model
• 6:43
  to kind of switch and understand
• 6:45
  what is going where, it is as easy
• 6:49
  as just telling the model
• 6:50
  that, I would like you to use this
• 6:52
  and this.
• 6:53
  And now kind of in my muscle memory,
• 6:56
  I will say using opus
• 6:57
  five subject for this,
• 6:59
  use a subject for this,
• 7:00
  and it will just kind of go out
• 7:01
  and do it.
• 7:03
  And I have a few friends
• 7:05
  who are using their their agent
• 7:09
  or cloud MD
• 7:10
  to also set in rules and kind of automate
• 7:13
  kind of these kind of things.
• 7:14
  But I've been stripping down
• 7:17
  all my MDF files a lot,
• 7:19
  and I think as the models are getting
• 7:22
  smarter and smarter, the less instruction
• 7:25
  you as a human have on the models,
• 7:28
  the better it delivers.
• 7:29
  Just like back in the days,
• 7:31
  you know,
• 7:31
  when you were doing meta
• 7:32
  ads, you had to really control everything
• 7:35
  of who you were targeting, etc., etc.
• 7:37
  remember six, seven,
• 7:38
  eight years ago
• 7:39
  it was all about
• 7:40
  just micromanaging things all the time.
• 7:42
  But now, because, you know,
• 7:45
  Meta's so smarter,
• 7:47
  you don't have to interfere that much.
• 7:48
  And in my opinion,
• 7:50
  that is also the stage
• 7:51
  where we're getting with these models.
• 7:52
  Now that the more you kind of
• 7:54
  overcomplicate your your MDM files
• 7:57
  and the more
• 7:57
  you send your instructions, the
• 8:00
  the more confused
• 8:01
  and the less good results you get.
• 8:04
  I don't know.
• 8:04
  Well, what's your experience?
• 8:06
  Yeah, I tend to completely agree.
• 8:07
  I think I
• 8:08
  bloated a lot of my markdown files
• 8:10
  early on, and
• 8:13
  I just assume like that they just like
• 8:16
  this was the best way to go about it.
• 8:17
  And again, in auditing our brief flow,
• 8:21
  which you know very quickly
• 8:23
  what it does is like,
• 8:24
  if I'm starting from scratch,
• 8:26
  it will use apathy,
• 8:27
  it will scrape subreddits, it'll try to
• 8:29
  find net new personas,
• 8:31
  you know,
• 8:32
  having already pulled in the ad account
• 8:34
  data and tagged all the ads,
• 8:35
  we have a good sense of like
• 8:36
  what new personas we can be testing.
• 8:40
  But, you know,
• 8:41
  what was happening before is
• 8:42
  it was writing the copy before
• 8:44
  it was actually finding,
• 8:47
  you know, this up, Sonus and read it.
• 8:49
  And I didn't really realize that.
• 8:50
  And I was like,
• 8:51
  this is why our copy feels so,
• 8:53
  you know, it lacks persona orientation.
• 8:56
  So ultimately having that audit
• 8:58
  and what I will say is that, like
• 9:01
  the caveat here is fable
• 9:03
  5.1, all of the copy
• 9:04
  it writes is going to have a watermark.
• 9:07
  So just like be aware of that.
• 9:09
  Like if you're writing a newsletter or,
• 9:11
  you know,
• 9:11
  putting out, you know, blog posts
• 9:14
  with, with fable 5.1,
• 9:16
  it'll be very easy to detect
• 9:17
  that you've done that.
• 9:19
  But what I will say is 5.1
• 9:21
  the copy style.
• 9:22
  It feels much less like conventional AI.
• 9:27
  It doesn't like keep reaching for these,
• 9:28
  like sort of like bold
• 9:29
  headlines and like,
• 9:30
  if it's not this, it's
• 9:31
  that it is far more organic
• 9:35
  in a way that I hadn't seen before,
• 9:38
  because ultimately
• 9:39
  I would have to write copy for ads.
• 9:41
  And then, you know,
• 9:43
  and my company still does this,
• 9:45
  you know,
• 9:45
  we rewrite a lot of the cloud copy
• 9:48
  or we're just like sort of
• 9:49
  starting from scratch,
• 9:50
  but it's it
• 9:52
  you're getting far less lists, like,
• 9:55
  you know,
• 9:55
  you're not getting those bold headers
• 9:57
  and then like,
• 9:58
  you know, the the paragraph. Right.
• 10:00
  And you're able to tell right away.
• 10:02
  That's the one thing
• 10:04
  that I've noticed as well.
• 10:06
  Yeah.
• 10:06
  That was the immediate tell for me.
• 10:09
  And I think that the like authentic
• 10:11
  look to it is going to change
• 10:13
  because I think that's,
• 10:14
  that's like the big
• 10:14
  one of the big complaints of super users
• 10:16
  is, hey,
• 10:17
  like this all looks very similar.
• 10:19
  You can tell it immediately
• 10:20
  and you can see
• 10:21
  people complaining about it.
• 10:22
  So the more that that gets removed
• 10:23
  and the more that it can also scan
• 10:25
  like understand
• 10:26
  your your native tone and everything.
• 10:28
  If you attach it, the faster
• 10:30
  or adapts to that, the better off
• 10:32
  it's going to be in terms of stickiness,
• 10:34
  because then people will say like,
• 10:35
  okay, well,
• 10:35
  I can just use this over here.
• 10:37
  People are going to be
• 10:37
  aren't going to be able to tell
• 10:38
  that AI is using this.
• 10:40
  And I think it's interesting.
• 10:43
  So as it relates to
• 10:45
  these models are going to keep changing.
• 10:47
  These models are going to keep updating
• 10:48
  and getting faster and different, etc..
• 10:50
  And you know, Reza from Motion
• 10:54
  Run has been talking a lot about
• 10:57
  the rise of an icy.
• 11:00
  I'm sure you've seen this
• 11:03
  an individual contributor
• 11:04
  in terms of somebody who's who's,
• 11:07
  you know, really
• 11:08
  AI forward in an organization,
• 11:10
  but also has the fundamentals
• 11:12
  of being a great marketer.
• 11:15
  How do you think,
• 11:16
  just as people go through this,
• 11:17
  like how do how can they position
• 11:20
  someone or themselves in an organization
• 11:24
  to be ready to continue to test
• 11:28
  and understand these things
• 11:30
  and these models and these changes?
• 11:32
  And what does that look like
• 11:33
  in marketing organizations
• 11:34
  and even six months
• 11:35
  by the end of the year?
• 11:37
  Yeah, that's a great question.
• 11:39
  And Thomas and I may differ on this one
• 11:42
  if you ask me this question,
• 11:43
  probably 6 to 8 months ago,
• 11:45
  I'd have a much different answer.
• 11:47
  You know where I'm at today again,
• 11:49
  like sort of
• 11:49
  coming out of the AI psychosis state is,
• 11:53
  you know, internally,
• 11:54
  we don't let any of
• 11:55
  our creative strategists touch
• 11:56
  any of our internal tools
• 11:58
  until we can trust
• 12:00
  they can write creative
• 12:01
  strategy, analog style.
• 12:03
  The tooling out there is best
• 12:07
  for amplification
• 12:08
  of existing hypotheses or theses.
• 12:11
  Right.
• 12:11
  And so like
• 12:13
  what I tell my team
• 12:14
  and what these models are very good at is
• 12:18
  serving as a compass,
• 12:20
  very good at pointing you in a direction,
• 12:22
  but you don't know what the end goal is.
• 12:24
  But if you start walking
• 12:25
  in that direction,
• 12:25
  you'll start seeing things
• 12:26
  that actually might make sense.
• 12:29
  And so to that end,
• 12:31
  if you are sort of like
• 12:33
  looking to get into the field
• 12:35
  as much as it sucks like
• 12:39
  strategy the old fashioned way and like,
• 12:40
  what is the old fashioned way?
• 12:42
  Well, it's just like figuring out,
• 12:45
  you know, your ICP,
• 12:46
  you know your persona,
• 12:47
  figure out their problems,
• 12:48
  go onto a subreddit.
• 12:50
  Maybe your persona is
• 12:51
  like someone with ADHD.
• 12:52
  Go on to those subreddits,
• 12:53
  figure out
• 12:54
  what those people are struggling with,
• 12:55
  and then find
• 12:56
  specific angles for your ADHD persona.
• 12:59
  And that's your angle or
• 13:01
  angle or problem.
• 13:02
  And then write copy
• 13:04
  based on that angular problem.
• 13:05
  Right?
• 13:06
  Oftentimes
• 13:07
  you can just take the copy verbatim
• 13:09
  from the subreddit, and that works great.
• 13:11
  Once you sort of
• 13:12
  get the hang
• 13:13
  of knowing
• 13:13
  how to identify the ICP
• 13:15
  and the angle or problem
• 13:16
  and writing the copy,
• 13:17
  then you have a foundation
• 13:20
  of what could create a strategy is,
• 13:21
  and then you can start
• 13:22
  utilizing these models
• 13:23
  to amplify that process
• 13:25
  so you can have it go search for.
• 13:26
  Net new personas for you
• 13:28
  that you haven't tested.
• 13:29
  You can go have it, search for angles
• 13:30
  that you haven't tested,
• 13:31
  and then you can have it
• 13:32
  help write copy for you.
• 13:34
  And so, you know, to that end like
• 13:38
  you know asi AGI side,
• 13:42
  right now we are in
• 13:43
  a space where you can amplify
• 13:45
  really, really good people.
• 13:47
  And if you have the foundation of
• 13:51
  strategy under your belt,
• 13:52
  then you can really,
• 13:53
  really ten x yourself
• 13:56
  in a way that I haven't seen others do.
• 13:58
  And so like again, in our organization
• 14:01
  we teach analog style.
• 14:02
  And once you
• 14:02
  sort of pass the OKRs analog style,
• 14:05
  we introduce the tools to you.
• 14:06
  And then people are like, oh my God.
• 14:08
  Like I can be so much more powerful now.
• 14:10
  I can cover 2XA number of brands
• 14:12
  and get the same results.
• 14:15
  And I think that's
• 14:16
  I mean,
• 14:16
  I think that's the argument
• 14:17
  for one of the main arguments for, for,
• 14:19
  for utilizing and superpower
• 14:21
  is super powering.
• 14:22
  I just
• 14:22
  I think the organizations keep shifting.
• 14:24
  I think what you're,
• 14:25
  what you're saying is really smart,
• 14:27
  which is starting with the basics
• 14:29
  and looking at this, Thomas,
• 14:30
  what do you think
• 14:31
  in terms of like
• 14:32
  where the marketers
• 14:33
  going to look like in this
• 14:35
  and understanding
• 14:36
  the changes
• 14:37
  in these models and stuff, because,
• 14:38
  you know,
• 14:38
  you're spending
• 14:39
  a certain percentage of your time
• 14:41
  looking at this
• 14:42
  all the, you know,
• 14:43
  and messing around with it
• 14:44
  and testing and stuff.
• 14:45
  So, you know, is it you?
• 14:46
  Is it the founder?
• 14:47
  Is it somebody else?
• 14:48
  Is it, you know,
• 14:50
  what does that look like?
• 14:50
  Because I think it's
• 14:52
  I think it's changing a lot a lot.
• 14:54
  And I'm sorry to be boring.
• 14:56
  I think
• 14:56
  I tend to agree with, with Will
• 14:59
  most of the things he was saying.
• 15:00
  So so first of all,
• 15:01
  I think you really have to
• 15:02
  be good at your craft
• 15:03
  and you have to do the reps.
• 15:05
  And the way
• 15:06
  I'm looking at it now is in our agency.
• 15:09
  We stopped hiring
• 15:10
  junior people
• 15:11
  and we're just looking for all dogs
• 15:14
  with ten,
• 15:14
  15 years of experience
• 15:16
  that has that gut feeling,
• 15:17
  that intuition of what's
• 15:19
  what's good and how processes worked
• 15:21
  and how it used to work.
• 15:23
  Because in my experience,
• 15:25
  those kind of people is the one
• 15:26
  who can really 10
• 15:27
  or 100 extra their output
• 15:30
  because they know what
• 15:30
  they're talking about.
• 15:31
  Right.
• 15:32
  So, so I think that having the craft
• 15:33
  and have done
• 15:34
  the reps is super important.
• 15:38
  Standing point.
• 15:38
  And I think
• 15:39
  you would grow and position yourself
• 15:41
  much faster
• 15:42
  if you understand
• 15:44
  the craft that you're trying to amplify.
• 15:47
  But I also I've spent a lot of time
• 15:50
  in the last few weeks,
• 15:52
  both in Norway, Denmark
• 15:54
  and in Finland,
• 15:55
  in Scandinavia,
• 15:56
  and also a little bit
• 15:57
  talking to agencies in Spain.
• 16:00
  And I realized that we are all
• 16:04
  we as in the part
• 16:07
  we are looking at the top tier of things,
• 16:10
  of the YouTube videos, the expo,
• 16:12
  the expos, etc.
• 16:13
  and we feel like we're behind.
• 16:15
  But but the world is really, really
• 16:19
  there, much slowly adapting to this
• 16:22
  AI world than I think that we feel like
• 16:26
  it's doing. I feel like I'm behind.
• 16:27
  But when I talk to both big ideas in
• 16:30
  smaller agency all over Europe,
• 16:32
  they're they're like mind blown
• 16:34
  about the things we're doing
• 16:35
  and how we're thinking about things.
• 16:37
  But what I'm saying is
• 16:38
  that if you love your craft
• 16:40
  and you understand
• 16:40
  the crafts have done your reps,
• 16:42
  I don't think it's so hard
• 16:43
  to position yourself
• 16:44
  because you just have to be curious
• 16:46
  and try to figure out and ask
• 16:48
  and have a conversation
• 16:49
  with these different tools
• 16:50
  on how you could amplify
• 16:53
  the more effective share work
• 16:55
  or accelerate. Cetera.
• 16:56
  So I think that loving your craft
• 16:58
  and having a little bit of
• 16:59
  that entrepreneurial curious
• 17:02
  is a good match to, to
• 17:06
  to get a much better output.
• 17:08
  It's interesting.
• 17:09
  So in fable five one
• 17:10
  you guys feel like
• 17:11
  the language is more native.
• 17:13
  You feel like it's it's
• 17:14
  faster to a degree.
• 17:15
  You feel like it's a it's able to do
• 17:20
  complex tasks.
• 17:21
  I mean, I've certainly seen people
• 17:22
  talking about image generation
• 17:24
  and video generation being better,
• 17:27
  being like world's better on five one.
• 17:31
  Yes.
• 17:31
  What are the other main
• 17:33
  big things that you guys see?
• 17:34
  See that's really changed in 5.1?
• 17:37
  Just just to point out the video part,
• 17:39
  because
• 17:41
  we have a lot of instructional videos
• 17:47
  on, on in Norwegian
• 17:49
  and the last few weeks
• 17:50
  or the last week
• 17:51
  we spent time with fable 5.1 actually
• 17:55
  translate everything.
• 17:56
  And Claude
• 17:57
  have been creating
• 17:59
  ads, creating voiceover, creating
• 18:01
  text, etc..
• 18:02
  Yeah, so,
• 18:02
  so video creation is so much better now
• 18:06
  and the quality is super high.
• 18:08
  And we'll talked about earlier on that
• 18:13
  that it's writing is better.
• 18:15
  But I also think it's
• 18:16
  reasoning is better.
• 18:17
  It's project management of orchestrating
• 18:20
  your project is
• 18:21
  is much better because again
• 18:24
  the language and the thinking is better
• 18:26
  and that makes everything else better.
• 18:28
  So for me,
• 18:29
  having it
• 18:30
  look at your projects
• 18:31
  and finding patterns
• 18:33
  or finding out what to focus on
• 18:35
  is also something
• 18:36
  I got a lot much more out of in five one.
• 18:39
  Yeah.
• 18:39
  I'm curious
• 18:41
  your take on sort of the video component
• 18:44
  as well.
• 18:44
  Like I assume you're not having anthropic
• 18:48
  models build the videos, right?
• 18:51
  You're using a model and then you're
• 18:54
  we used to use Higgs Field
• 18:56
  and all that stuff for that,
• 18:57
  but not with 5.1.
• 18:59
  I'm basically jumping in the video
• 19:01
  saying, here's an original video,
• 19:03
  make an English version,
• 19:04
  make slides for it,
• 19:05
  sure got voiceover, etc.
• 19:08
  so so
• 19:09
  so this is more instructional videos
• 19:11
  for how to use a platform,
• 19:13
  how to think
• 19:13
  about creative work
• 19:14
  for our clients more than creating ads
• 19:17
  and those kind of things.
• 19:19
  And that used to be really sucky,
• 19:21
  you know, a few months ago,
• 19:23
  you're not able to do that at all.
• 19:24
  But the quality is surprisingly good now.
• 19:29
  So we really ramped up our production of
• 19:32
  of our Norwegian videos
• 19:34
  because we're now starting in Finland
• 19:36
  and in Spain, etc..
• 19:37
  So, so localizing our content
• 19:39
  with basically
• 19:40
  just dropping the
• 19:42
  MP for files in and say, hey,
• 19:44
  make this into an English version,
• 19:46
  make this into a finished version.
• 19:47
  It is.
• 19:48
  It is really good these days. Cool.
• 19:50
  I that's really smart.
• 19:52
  I have like consulting clients
• 19:53
  that are
• 19:54
  have always been asking me
• 19:55
  that like translating videos.
• 19:57
  That's interesting.
• 19:58
  And so that's wow okay.
• 20:01
  That's that's incredible. Okay.
• 20:02
  And have you found the same
• 20:03
  for images as well.
• 20:05
  No, I honestly
• 20:06
  I haven't really tried images because.
• 20:09
  Yeah, because of just how bad it was.
• 20:14
  But I expect it to be better.
• 20:17
  But but when you're talking
• 20:19
  about creating images
• 20:22
  and videos for, for ads
• 20:23
  and those kind of things,
• 20:25
  then I think Higgs Field
• 20:27
  and the other usual Suspects
• 20:28
  are a much better way to do it.
• 20:31
  Yeah, no, I completely agree.
• 20:34
  And obviously I've talked about sort of
• 20:36
  my qualms with Higgs
• 20:37
  field in the past here.
• 20:38
  But one thing that I thought
• 20:40
  fable did a very good job at is like,
• 20:42
  and we sort of
• 20:43
  built the system for this,
• 20:44
  and you have to sort of
• 20:45
  combine it with Gemini,
• 20:46
  and maybe I'll talk about this
• 20:47
  more in the New York meeting.
• 20:50
  But like
• 20:51
  if you feed a folder of b roll content,
• 20:55
  what you can have it
• 20:56
  do is have it's slice and dice
• 20:59
  the B roll content, and then have it
• 21:01
  tag each piece of B-roll content
• 21:03
  and then
• 21:03
  have Gemini Flash
• 21:04
  go and like visualize each of those
• 21:06
  and add additional tags.
• 21:07
  And then you can have create mashups
• 21:10
  very easily.
• 21:11
  And I would say
• 21:11
  like with a fable model
• 21:13
  is probably 80% of the way there.
• 21:15
  Like you would sort of
• 21:16
  have to go back and forth in cloud.
• 21:17
  Could be like, hey,
• 21:18
  can you just like move
• 21:18
  this here, move this there?
• 21:19
  I also think like cloud code,
• 21:21
  this may be a hot
• 21:22
  take is like a really good
• 21:23
  editing software.
• 21:24
  Just like you could just be like,
• 21:25
  hey, you know,
• 21:26
  can you just like, move
• 21:27
  that this up thing?
• 21:28
  They also have a really good markup tool
• 21:30
  that I think is pretty slept on.
• 21:32
  Yeah, yeah.
• 21:34
  And then,
• 21:34
  you know,
• 21:35
  to that end,
• 21:35
  I candidly haven't
• 21:36
  tried this, this workflow
• 21:38
  with 5.1,
• 21:39
  but based on what you just said,
• 21:40
  I think there's probably serious
• 21:41
  legs to it. Yeah.
• 21:43
  You should.
• 21:43
  You'll be impressed.
• 21:45
  Oh that's amazing.
• 21:46
  That's a really good tip.
• 21:48
  Cool.
• 21:49
  Well, I
• 21:51
  feel like we've
• 21:52
  covered a lot of good stuff.
• 21:54
  Many more things on fable five
• 21:55
  one to be coming,
• 21:56
  but I think you for listening
• 21:57
  to this quick episode, everyone,
• 21:59
  and have an awesome rest of your day.
• 22:01
  We're always here for you
• 22:02
  and we appreciate any.
• 22:03
  Review
• 22:04
  your feedback, so
• 22:05
  take care and we'll talk to you soon.
• 22:08
  Bye! Thanks all.