// transcript — 630 segments
0:00 Intro
0:02 I'll let you in to a little secret. There's a tool called Weev AAI that is
0:08 the secret to creating absolutely jaw-dropping beautiful mobile apps,
0:14 software, etc. So, if you're anything like me, you vibe coded using tools like
0:21 Google AI Studio, Claude Code, Lovable, and the results don't look that
0:25 beautiful. Like, they they look fine, but they don't look beautiful. And why
0:29 beautiful matters is you want people to fall in love with your product. So, how
0:35 do you actually do that? Well, I brought on my friend Sarah, who's an incredible
0:40 designer. She sold her last company to Snap, and she walks us through her
0:46 entire AI workflow, including how to use Weebi AI, the secret tool that is so,
0:52 so, so powerful that could help you create beautiful jawdropping designs. In
0:58 this episode, we do a live build. We actually show you how we can live build
1:04 using Google AI Studio, using Claude, using Weevai, using Figma, and going
1:09 back to Google AI Studio. So, this is the entire workflow. People who stick
1:13 around to the end of this episode are going to be weapons. This is a tutorial
1:17 that doesn't exist anywhere on the internet. So, give a like, comment, and
1:22 subscribe for more in this in your feed. Let's get going. time, baby.
1:34 >> I begged her to come on to the pod. Sariah, a dear friend of mine, probably
1:39 one of the most talented designers and you, 99% of people haven't heard of her.
1:44 She sold her last company to Snap. She's incredible. And I asked her to come on
1:49 for one sole reason. There's a lot of us who are vibe coding apps, but they look
1:55 kind of garbage. and Sariah's AI workflow. She uses tools like Weevai,
1:59 she uses Claude, she uses Flux, she uses a lot and she's able to create apps and websites and SAS
2:08 products that really resonate with people. So Sariah, thank you for coming
2:13 on and I'm hoping that you can teach us how to do how to how to create gorgeous
2:19 vibecoded apps. >> Thank you for having me. Very kind intro. Um, yeah, I think this will be
2:24 fun. >> Okay, so if people stick around to the end of the episode, tactically like what
2:30 are people going to learn? >> Okay, you will learn how to design
2:34 something that looks like an agency designed it. >> Okay. And not just not just any agency.
2:42 >> No, no, like a real brand designer. And maybe more importantly, you'll learn how
2:45 to make something that someone actually wants to download. Um, and you can use
2:51 yourself as an example. you'll see an ad for it yourself and be like, "Oh, I
2:54 would actually use that." >> Yeah. And to be honest, like, as much as
2:59 I wanted to do this for the pod, I wanted to do this for myself. Um,
3:02 because I wanted to learn, I think there's so much opportunity right now to
3:07 be vibe coding products that people love using, you know, IG reels and Tik Tok
3:13 ads to to get it to market. But, you know, a lot of times when I'm vibe
3:17 coding stuff, it just looks kind of kind of lame. So, what you're showing me
3:21 today, before we get into it, what you're showing me today, I'll be able to
3:27 to to basically design products this this beautiful. >> Yes.
3:32 >> Okay, cool. Well, show us how. >> Okay, amazing. Um, okay. Yeah, by the
3:35 Why Every Vibe-Coded App Looks the Same
3:38 end of this episode, we should have something like this designed and walk
3:42 through how we do it. So, let's get started. Um, we already introed this,
3:46 but we should build something that doesn't look like a robot made it. Um,
3:52 anyone can build an app now. It's amazing. So much has changed, but now
3:57 everything looks the same. And so, if you're going to make something that
4:00 looks like everything else, it's just really tough to actually expect anyone
4:05 to download it. Um, I am obsessed with branding. I love branding. We should
4:10 think of it as the same thing as the product, same thing as engineering, like
4:14 everything is branding. But to actually prove why um okay, let's make an app. We
4:20 thought we would use a voice journal app as an example. I feel like it's a right
4:26 of passage to want to make a journaling app. >> That's right.
4:31 >> I'm always like, I should journal. And then I'm like, why don't I just make a
4:34 Prototyping with Google AI Studio
4:35 journaling app instead? Um, okay. So, for this demo, we're going to use Google
4:40 AI Studio. It's definitely one of my favorite vibe coding apps. It's really
4:46 good at oneshotting interfaces, but we're more so focusing on design here.
4:49 But, let's just use this kind of basic prompt like build a mobileiz voice
4:54 journaling app. Users can see a home screen, record voice entries, see the
4:59 transcript, browse past entries, and have it build it. Let's see how long
5:03 this takes. It's pretty quick. I would say like two minutes or under.
5:09 >> Yeah, it's honestly so quick. It's amazing. It's kind of become my favorite
5:13 oneshot um tool for vibe coding. >> Yeah, I'm s I like Have you used any
5:20 other tools? Like why you why suggest Google AI Studio? >> I use AI So AI Studio is my favorite for
5:27 like a oneshot prototype. So if you just want to see something, you're not
5:31 working in an existing codebase, you're just like, what would this look like?
5:34 Google AI Studio is really good because it'll do it end to end. When I'm
5:37 actually working day-to-day in an existing codebase, um like in my current
5:42 startup, we typically focus on using cloud code just because it's really good
5:47 at working within a codebase. Um, and then sometimes if I do want the
5:52 capabilities of Gemini, but I still want to work in um an existing codebase, I'll
5:58 use Gemini's model and cursor. And then I found that Gemini is really good at UI
6:03 and shader work and like more complex animations, >> which is key, right? Cuz we're going to
6:08 take this one shot and then we're just going to trying to paint it beautifully,
6:12 right? >> Exactly. Exactly. we're going to take this one shot and be like, "Okay, we
6:17 have the functionality. Now, how do we make it something that a human would
6:21 actually use?" >> And I think the problem is what a lot of
6:26 people do is they go back and forth uh you know, pingpong so to speak with
6:31 the AI models when they need to kind of take a step back and then figure out the
6:36 branding and then apply that branding to the app. Right. >> 100%. Like you can think of it as like
6:41 what do you want to outsource and what do you not want to outsource. Like you
6:46 could say this focuses on like what it should do but you want to keep as much
6:51 like how it should do it to your own brain so that it like actually feels
6:55 distinct and looks different. Otherwise like it makes sense that everything
6:58 would look the same if you're outsourcing like all of the thinking. Um
7:02 but what's amazing is we can now outsource stuff that are solved
7:07 problems. like it's a solved problem to build a lot of this, but your unique
7:12 take is pretty important. >> And I'm I'm actually really interested
7:16 to see how you use Weebi AI to do it. I know we're going to get there, but
7:20 that's a tool that I've been meaning to try for a very long time. I just it's a
7:24 bit o it feels like a bit overwhelming to be honest. Uh >> I think actually even as a designer like
7:31 I used to not like nodebased tools even though like origami and stuff was really
7:35 popular, I never personally liked it. Um, but I think with like AI models are
7:41 the unlock for it for making them like actually super useful. Okay, let's see
7:44 what this generated. Okay, this generated. Good morning. You can record.
7:49 Hello. Hello. Hello. It's going to record something. Stop recording it. Pretty cool. I mean,
7:55 it's still obviously really epic to see that this could get generated in like 20
7:58 seconds or 2 minutes or however long it took. Okay, cool. You can see the home
8:03 and you can see the journal. Let's take screenshots of this and see what it
8:09 looks like in our Figma. So, this is the app that got generated via Google. And I
8:14 mean, honestly, like if the app store was just invented, it wouldn't be that
8:19 bad cuz it was a novel thing to be able to record back then, but now that we're
8:25 so used to software, like if we were to see this on Instagram or Tik Tok, um we
8:30 would not download it ourselves to use. It's true, you know, and I don't think a
8:34 lot of people are saying that, right? So, yeah. So, how do we how do we this
8:38 is really cool and it's got a good foundation. How do we take this
8:41 foundation and build something beautiful? >> Okay, so the helpful part about this is
8:45 we know what it needs to do. So, let's just like quickly write down what it
8:49 needs to it needs to record record audio. It needs to Let's see what
8:55 else. It needs to have a history. >> Yeah. what you record and it needs to
9:01 have like a button to actually start recording or let's we already have that
9:06 with record audio. It needs to have timer while you're recording.
9:11 >> So basically you just you just outline what the screens need to do at this
9:14 stage. >> And most people stop here, right? Like vibe coding focuses a lot on like what
9:20 does the thing do? It records, it shows your history, and there's a timer. Um
9:25 but that's like not what a product is. is like you don't really think of your
9:28 favorite products in the world and think, "Oh, it does this." You think,
9:32 "Oh, it makes my life easier and I just feel so good using it and like blah blah
9:34 Defining How Your App Should Make Someone Feel
9:35 blah blah blah." So that's what we're going to focus on now. An equally
9:39 important thing to what it does is like how should it do it? So I literally
9:44 always start here. Like I define how something should make you feel. And even
9:47 though that sounds like a little designer hippie-dippy, it's like really
9:53 not. Um, and so, okay, I kind of wrote out this prompt beforehand. Um, and this
9:58 is exactly something that I would paste in Claude, and we're going to paste it
10:01 in Claude and see what it thinks. But let's just talk about how the audio
10:07 journal app should make someone feel. I think it should make someone feel, or
10:14 let me rephrase, um, who are we making this for? We're making an audio journal
10:19 app for someone who wants to get their thoughts out. It's too long to write
10:24 them out on pen and paper, but there might there may be tired of just like
10:28 using tech all the time and like being on their phone all the time. And so that
10:31 tells us that we don't want people to feel like they're just in any other app.
10:35 So that means that our original design already failed because this like 100%
10:39 feels like every other app on the planet. And so it kind of goes against
10:43 like why someone would download an AI journal in the first place, which is
10:47 like to feel calmer and just get your thoughts out. So now we know, okay, we
10:53 want someone to feel maybe like they're in something analog. We want to be able
10:57 to think out loud, but not feel like they're on their phones. Um, and they
11:01 shouldn't feel like we're trying to steal their attention and like force a
11:05 bunch of ads and random stuff down their throat, like a lot of apps. Let's go
11:07 Using Claude to build Brand Identity/Guidelines
11:11 ahead and paste this into Claude so that it can help us along the way. I'm
11:16 designing a voice journaling app. Let's have it help us define how it feels,
11:24 blah blah blah, and see what it says. >> And what's your advice for people who
11:28 like I I feel like your insights there were really bang on. Like any advice for
11:32 people to come up with insights like that? >> Good question. Um, I think an example is
11:40 I I I actually use movies and TV shows as a reference a lot. Like movies and TV
11:46 shows have a beginning, middle, end. And if we think of a product like that, it
11:49 becomes a lot easier. Like if we're building like a health tracking app, we
11:54 don't really care about the middle and end yet. We care about the beginning and
11:58 that's like how someone feels. So we might say like, okay, we want someone to
12:01 feel not anxious by like having some annoying coach telling them what to do every 5
12:08 seconds. Like that might not get the girl who I want to download it to
12:11 download it. >> And so if you just think really like as
12:14 a human, I think that's why people say like build stuff for yourself cuz it's
12:18 like you naturally know how you feel and how you want to feel. Um, I don't think
12:23 you have to build for yourself. Like I But you also have to think about like,
12:26 okay, then how would that person want to feel? And then you work backward
12:31 backwards like, okay, I want this girl to feel calm. Therefore, it probably can't have
12:38 like 75 timers saying like track these 5,000 things and like work backwards of
12:43 those like key words. But also like Claude helps with it a lot. So, even if
12:47 we said something like, "I'm designing a voice journal app. I want people to feel calm or they're
12:54 tired of their phones." Like, you could kind of see, okay, what does it say? How
12:57 should it feel? Like, do I actually like what it's saying or do I not? And like
13:01 you keep I I think Claude is like the perfect brainstorming tool.
13:04 >> Cool. That's helpful. >> So, here, let's see what it says. It
13:08 says this is for the overinker that needs to exhale. Mind's constantly
13:13 running. Sure. Whatever. Over stimulated. Yep. Like this all sounds
13:18 like pretty spot-on. They crave permission to be unpolished. They
13:24 suspicious of apps that um want too much of them. Analog warmth. That's cool
13:29 because it's highly visual. So I think like the things to take from this are
13:32 like what are things that are like very visual. And then also what's equally
13:36 helpful is like what is it not? It's not productivity tool. That's important
13:39 because that's really the opposite of like a private space to rant. It's not
13:46 social. um not needy. Okay, this is really helpful. And I don't tend to use
13:50 the questions it asks for this branding process cuz it's more about like what do
13:55 we want from this? And then let's go back to the Figma. So now Claude knows enough to kind of
14:01 help us create brand guidelines. I feel like a lot of people think of brand
14:05 guidelines as this like corporate like afterpaint structure, but it's really
14:09 not. Brand guidelines are really just like um a prompt in it of itself and
14:14 it's a prompt we're going to bring into Weevi. So let's call our app cassette.
14:21 So we can say let's call our app Based on that, write me brand
14:30 And we'll just kind of use this throughout. I'm happy you said, by the way, that you
14:37 think brand guidelines sounds corporaty I sort of know that people are thinking
14:47 that in their heads. >> Like a different phrase. Oh gosh, it's
14:51 going really extra. It's making like literal brand guidelines. No, no, no.
14:57 Just write them out, please. What people don't realize is the brand
15:00 guidelines is what's going to get people to download your app.
15:04 >> It's the opposite of corporate jargon and corporate stuff, right?
15:08 >> 100%. Like if a brand is successful, you won't notice it in the final app. It'll
15:12 just feel good. >> And that's really the goal of like all
15:16 these complexities and processes. Like you don't do processes. Like sometimes
15:19 you could literally just sketch something and that's it and you've got
15:23 it. It's just about knowing like what you want. But I think these tools are
15:26 nice because they help you get to know what you want. Like we can look at this
15:30 and see, do we like it or not? Like cassette. Deliberate misspelling of
15:33 cassette. That was not deliberate. That was my bad. That's really not important.
15:38 Cassette is a voice journal for people who think out loud but are tired of
15:42 being on their phones. Totally spot on. Let's put that in the Figma. We like
15:46 that. That seems like a good thing to put into Weevie. And that's kind of just
15:50 what you do. Like you just take stuff from here that you like. Um, and like
15:58 let's see where it talks about cassette is the satisfying click of a
16:03 record button. That feels really nice because we can kind of feel that like,
16:07 okay, maybe we just build our whole visual style around like analog cassette
16:16 um, designs. And so I usually will just like talk back and forth with Claude.
16:21 Um, it could take one chat like this or a few more until you kind of end up with
16:25 guidelines that you like. So, I think parts of this are relevant and then a
16:29 bunch of it we don't really need. But for now, let's just take like these two
16:34 things. Voice journal, people who think out loud. It's focused on analog and
16:38 Building a Mood Board with Cosmos
16:40 like a record button. And then the next tool I use is Cosmos. Cosmos is my it's
16:46 honestly one of my favorite apps just in general. Um, it's an alternative to
16:53 Pinterest and it's beautiful and it's usually where I make mood boards. So, I
16:58 prepared a mood board for this voice app, but I'm going to show you just like
17:04 very easily how um you could add to it or do the same. It's super simple. Like you could
17:10 literally just type vintage cassette. Oh, there's this like cassette and you
17:16 can kind of go down all these clusters. Um, and Cosmos is really nice because
17:20 you kind of just end up in this like endless loop. And so if you really like
17:24 the colors in this, I'm going to add this to our voice app mood board. And
17:30 then you can just keep going until you stop. I usually end up on this site just
17:37 without even trying. Um, so if we go back to the mood board, you can kind of
17:43 feel the vibe that like all of these are matching the stuff we just talked about.
17:47 like they're all really analog. We can lean into the analog vibe to make
17:52 something feel how we said in the beginning, which is like not very techy,
17:58 Intro to Weavy AI: Node-Based Visual AI Tool
17:59 not overwhelming, and like personal. And so the next step is the main part
18:03 we're going to talk about, which is Weebi. So Weevie, super cool app. Um,
18:08 like Greg mentioned earlier, it's nodebased and yeah, kind of like you said, it
18:15 looks complex. Like if you zoom out, it has like this beautiful complex thing,
18:18 but it's actually really not. I found that it's just the best way to try any
18:24 model um visually and like really really quickly for honestly any project. Um, so
18:32 let's get back to branding. I went ahead and I saved all the images from this
18:36 mood board to my desktop. So, you just like go into it, click save. Um, then it
18:43 saves and then you just paste them all into Weevi so that we can use them for branding
18:52 and then they load really quickly. Um, let's put them all together.
18:58 And now the question, this is what's not really, this is what you couldn't do in
19:02 the past. Like in the past, you'd have a mood board and then it's like, okay,
19:05 then what? It's all this work and you have to like contact a designer and have
19:09 them like look at the mood board, create custom assets from it and like all this
19:13 stuff. But now, literally anyone can do it. Um, let's go back to the Figma.
19:19 Okay, so we did the mood board. Cool. And now we're going to start a weebi.
19:23 And so the way you use Weevie and the way I use it is I kind of switch between
19:28 Claude and Weevie and I have Claude create the prompts that I want to put
19:32 Generating Color Palettes with Flux 2 Pro in Weavy
19:33 into Weevie. And so okay, we're going to start with colors. Colors is like a mood
19:40 board kind of the most primitive easy thing to start with and then it's going
19:44 to be like building blocks like colors and then backgrounds and then buttons
19:47 and images. And you could think of it as like a literal painting and we're kind
19:50 of starting with the background. So, we have this image in Weevi, one of the
19:53 mood board references. You're going to take it as an input and let's use an
19:58 image model. We're going to use Flex 2 Pro. It's a very good just like general
20:02 model to use for a bunch of stuff. And so, then in the prompt, we can write
20:06 something like, I need a color palette. Extract colors And let's see what it does.
20:23 And usually when I run these models, I run a bunch of different ones. Like,
20:28 let's do the same thing on this one cuz this is kind of nice. And let's just say
20:34 give give me a color palette based on this image, but make it textured and vintage.
20:46 >> And you're doing that cuz you're you want to see like what's going to get you
20:50 the best result. Yes. So, you want to see what's going to get you
20:55 the best result and what's going to get you I mean part of this is developing
20:58 your own taste in general and in the product you're making. So, like if we go
21:03 back to okay, I want this person to feel retro, calm, cool. Um what else? Like
21:09 not too techy. Like what matches that? So that's another reason I start with
21:12 like how do you want the end person to feel is like any step of the way you
21:15 just kind of go back to that and are like is this true or not true? So, like
21:20 I kind of like thought I would love this more, but I kind of love this more
21:24 personally. Like it just feels really cohesive and there's like a really nice
21:28 textural element to it. >> And so maybe that tells us like, oh, we
21:34 should lean into like the textures of cassette tapes and um stuff like that. Like actually
21:41 earlier when I kind of was thinking through this demo, um I did some other
21:48 variations and kind of ended up with like a similar metaphor of like, okay,
21:53 how can we lean into this like analog cassette tape? Um so maybe we kind of
21:58 run with that. So now we have like a couple color palettes. Um
22:02 but I don't know about you, but I personally also am like obsessed with
22:06 this image, honestly. Like this could literally be the app. But and honestly sometimes I do that
22:14 like I'll find one image I like and then make a whole company based off that
22:17 image. >> Like my last startup I worked on literally I
22:23 just found like one image and then the whole branding came from that one image.
22:25 So I wouldn't be afraid of doing that. Like you could literally just love this
22:28 image so much and be like this is the whole app the whole product like every
22:32 >> I mean didn't Johnny I do that with Dita Rams and Apple like that's basically
22:39 what he did I think you know Johnny Ives who used to run design at Apple was a
22:44 huge Dita Rams fan. I think Dita Rams was you know creating stuff in the 50s
22:49 and 60s and just saw some of his stuff and then that inspired a whole set of
22:55 Apple products from that. So, we're not reinventing the wheel here. We're just
22:59 doing it using different tools. >> 100%. It's all the same exact thing. And
23:03 even for the new product I'm working on right now, like I literally just found
23:07 one image that I really liked and now the entire product is based on this like
23:10 one shadow image. >> So, this one I feel like it's kind of
23:14 nice to find that image. So, like we have these two color palettes, but I
23:17 also just personally like this image. And I think why do we like it? like we
23:20 have to think about like, okay, well, we're using the colors that feel analog,
23:24 but we still have like maybe a background that feels calm and like
23:30 versatile. So, now we kind of know like, okay, we have a bunch of colors to play
23:33 with. We're experimenting with colors and you can kind of go back and forth
23:37 between like what do you need help with from Claude and what do you just want to
23:40 put into Weebi. Like sometimes if something isn't coming out in Weebi how
23:44 you like it, just tell Claude, say to Claude like, "I want a better prompt for
23:48 getting a color palette from an image." Or you can just go directly into Weebi.
23:55 Um, and so one idea that actually came up when I was looking at this image is
23:59 like we really like texture, which we saw here. We also really like this just
24:07 like pretty analog hybrid machine thing. So, what if the anchor of our visual
24:13 system, if you will, is like as you use the product more, this becomes more used
24:19 visually. Like, what do you think? >> I love that. I love that. I mean, like
24:22 digital products, think about it. They're kind of like cartoons. It's like
24:26 The Simpsons never got older, you know, but time went on. So, I like this con. That's a huge
24:32 That's a winning concept. I'm I'm in. >> Okay. I love that. True. And I feel like
24:36 sometimes if you overuse analog things digitally, but there's no purpose to
24:40 them, it feels kind of gimmicky. And so you got to think about too like what's
24:45 the purpose of your brand element? Like it's not for the sake of it. Like we're
24:48 not just putting this to be pretty. But if you're making it more like used like
24:54 you would in the real world, then it's also just like a reason for someone to
24:58 actually use it every day. So then we might say to Claude, "I have a product
25:04 idea. that the app should age with use. Um, new users get a nice interface, blah
25:12 blah blah, but we want to extract that loved and used feeling. And then let's
25:16 say like, write me a prompt. Oh, I have that here. Write me a prompt
25:23 that shows a progression on a vintage audio equipment image.
25:28 Let's see what it says. Let's see what that does. It might be a
25:33 little too detailed, but who knows? It's good to experiment.
25:40 And maybe we use it with this cassette tape because I'm going to
25:45 do it on a cassette tape. >> And so Flux 2, you basically are
25:50 exclusively using Flux 2 Pro >> right now. Yes. For image to for any
25:55 image generation, it's like you don't have to be too deep about which model
26:00 you use. like you just want to see it visually. And I found that Flux is quick
26:05 and just good at like text prompts and it feels like very generically good. Um,
26:11 and then I'll get into like a couple other models down the line when it
26:14 becomes more important to like grab actual assets. Okay, this kind of shows
26:18 us what it would look like to like have a more used cassette tape. And I do love
26:23 the vibe actually. I'm obsessed. So now maybe we take this image we love and say
26:28 um okay let's put that into flux again and the prompt is like show a more used
26:38 version of this audio system. Um and again like there's always a more
26:43 detailed in-depth prompt like you can always paste your prompt into claude and
26:46 be like make this a better prompt and just kind of like experiment it with it
26:51 like you would vibe coding. Let's see what it does. And then while it's
26:54 Creating Record Buttons and Custom Assets in Weavy
26:54 generating, let's go back to Figma. Now we kind of have a visual anchor of the
27:00 whole product. Like it should all kind of get more used as we
27:06 use it, as the user uses it. Um, and like if we're going to go back to what
27:09 the functionality is, we need to be able to record audio, see what you record, and have a timer.
27:16 And now let's say that we have a color palette and we have a visual anchor. So,
27:22 we sh we're probably ready to like actually work on the actual like assets
27:29 and buttons now that we kind of have honestly like all the pieces put
27:32 together. Like I actually love this so much. Oh my god, this is so pretty. Um,
27:39 okay, cool. So, we could start with the buttons. Um, this is another prompt I kind of wrote
27:45 out beforehand is like we need a record button. Um, and something I do a lot is
27:51 I say to Claude like, "Okay, write different prompts of different types of
27:55 record buttons." Like, we might want to try like a minimal version that's
28:00 similar to this. We might want to try like a more intense version
28:06 um that might be similar to some of these versions I did earlier. And you
28:10 kind of just want to see what different buttons look like. So, I'm going to show
28:15 you how these were generated over here because it's the same exact process we
28:18 just used. So, you might take input as a color palette. You might want to take input as
28:26 an as a as a um cassette tape. Any image again can always be your input. And then
28:32 let's see what the prompts are. So, this is for example like generate a record
28:37 button inspired by this cassette recorder. It should be like an analog
28:41 button from the 80s. use this color palette. Um, and I've actually just like
28:46 a quick I found that like a lot of talk online about AI is like how do you get
28:50 consistent characters with like AI images, which is important and like
28:54 challenging. But the nice part about product design is you don't need like a
28:58 face to look the same. You just need colors, shadows, lighting to look the
29:03 same, which is a lot easier. So, it's like a lot easier in app design and web
29:07 design to make something like visually consistent with AI than like images
29:12 you're generating of a person. Anyway, okay. So, we have a bunch of buttons.
29:17 This one I really liked, but it makes it makes the background hard, I think, cuz
29:24 it's not like on its own. Um, so my personal favorite ended up being
29:29 just isolating this red one from our favorite image right here. What do we
29:33 think? >> I couldn't agree more. >> Okay. >> Simple. It makes It's like, you know,
29:40 you go into like a Staples or whatever and they sell those like big red buttons
29:43 that people want to press. Like that's what this feels like.
29:47 >> Yes. Completely. I love it. It's so adorable. And I kind of like how it's
29:51 like modern but not modern and analog but not like completely analog. Um,
29:57 okay. So, we can kind of just go with this button. Normally from here, um, I
30:02 would just download this image like straight from Weevi. Let's download it. They make it so easy.
30:11 Um, and then into Figma, I would just paste it. And we're literally just starting to
30:17 like save assets that we like. And so now we know that we like this button.
30:22 Um, and we can deal with like typography and text and like all of that stuff after.
30:29 And so if we're going to go back to like, okay, what do we need to build? We
30:33 now have a record button. Um, we can deal with typography and text after. And
30:40 we have kind of this like recording um speaker system. Clearly, I'm not like
30:46 Generating Cassette Tape History Elements
30:48 an audio person. Okay. Now, the big element we're missing visually, I guess,
30:53 is the like history of after you record something, what does it look like? In
30:58 our AI generated app, it's like a little list. But if we want to lean into this
31:04 like analog whatever vintage cool system, we probably want like a metaphor for that.
31:12 And luckily, we already have a metaphor for it, which is in the name. It's a
31:18 cassette tape. So, we could just go back to one of these images. I actually
31:24 really liked one that got generated here. And I'll show you how this was made
31:30 from this color scheme, which was created the same way we did earlier.
31:35 You could say, "Generate a set of cassette tapes using this color palette.
31:39 Show the tapes from the spine view like they're stacked on a shelf." Actually,
31:44 I'm kind of curious if we did that from this color palette. What would it do?
31:48 Let's take this image. I love this image. It looks like it'd be like a cafe
31:53 in Brooklyn. Um, and let's do flex again. And then let's do that same prompt I
31:58 just pasted. Generate a set of cassette tapes. Let's see what it does. And then let's do the
32:04 same thing from this one. I'm just curious. And I just kind of copy paste the same
32:10 prompt over and over. And I also just love how we kind of tools are like meant to be
32:17 really messy. Like it's meant to be I think really just a bunch of things
32:22 everywhere. Um, and what's we haven't really talked about like the price of
32:26 something like this. Like how much is it expensive? Is it cheap? Weebi.
32:30 >> Good question. It's pretty well priced. I think I think it's like 200 credits
32:34 that are free every month. And okay, wait. I'm on the paid one, which is like
32:38 $10 a month, I think. 15. And you get 1,500 credits cuz I use it a lot for
32:41 work. >> But this you could do for free. Like you get like
32:47 >> I mean this has taken us so far probably like 30 40 credits. So, it's kind of
32:51 amazing, honestly. >> And see how good this is. Like, it
32:55 literally just created all the cassette tapes based on the two color palettes. I
32:59 think it did a pretty decent job. And then an idea is like, okay, we want to
33:03 use these cassette tapes as the history of like each recording. So maybe we take
33:09 this image, we do another flux prompt and we go um give me this same image but change the
33:22 text on the binding to be a date like January 19th, 2026 because then we can kind of keep this
33:31 metaphor going and be like, "Okay, every time you do a recording, we have like a
33:35 new cassette tape. show up on the screen. I did one earlier that I also liked. I'll show you guys in
33:42 a sec once this generates. Actually, wait, I'll just show you now.
33:46 Like this one, for example, the one I showed earlier, we took this image
33:53 and it has like random text that it generated. And then I said, okay,
33:58 replace the text with different dates. And it did it pretty nicely. So, you
34:01 could kind of imagine like you'd give this to a vibe coding platform and then
34:05 it would just like do it automatically. Let's see if it did it. Oh my god, it's
34:09 so cute. Okay, I love it. See, you can kind of see the dates and then I mean
34:12 some of them are a little off, but that's okay. These don't need to be
34:16 perfect. So, now let's download this image. I'm going to download the one
34:20 from here because it has a bit of a background color I like.
34:28 So, let's download this image and let's paste that into Figma. Okay, we have
34:34 these two visual references. So, now we pretty much have even though it's kind
34:38 of hard to see, if you squint your eyes, you can see that we have like all the
34:41 functionality visually. Like we have a record button, we have a speaker, and we
34:45 have the history. It's just not in an app yet, which is fine. Let's do the
34:47 Logo Generation with Ideogram V3 in Weavy
34:50 logo, and then let's come back to the last step, which is putting the app
34:53 together. Okay, this logo stuff still blows my mind as someone who has been
34:59 designing logos their whole life. Um, where am I? Okay, let's go back to
35:03 Weebi. So, now we're going to switch models. We're not doing ide. We're not
35:07 doing flux anymore. We're going to do ideoggram. Super cool model. It's It's really good
35:13 at typography, honestly. It's like amazing. So, okay, we're going to put
35:18 this into Claude. And it basically says, "Write me four logo prompts." one that's like this, one
35:23 that's like this, one that's like this, one that's like this, whatever. And
35:26 let's see what it does. >> So, it says a technical word mark version A,
35:36 handwritten, B, tape label, C, minimal word mark, D. Okay. >> Do you like these?
35:42 Maybe this one. >> Yeah. I mean, I feel like that's the
35:45 coolest. Like a version C, like a tape brand. >> Yes. Like I haven't even heard the names
35:53 TD Key or Max Cell and Bass F >> but like I I it's actually probably like
35:59 deep in my brain somewhere, you know? >> But like I I I I do like I think that's
36:04 like what we're trying to evoke. >> Okay, let's go with this one.
36:11 And then let's see. Oh, and another cool thing Logo does that Claude is good at
36:15 is giving a negative prompt, which is basically what you don't want.
36:19 >> Um, and so we don't want something 3D render glossy gradient mesh.
36:24 Unfortunately, AI is ruining gradients. Um, sparkles like the you could just
36:27 list a thousand things you don't want. So, that's kind of helpful, too. So,
36:35 let's try it. You type idiogram v3 and you're going to add the prompt which
36:39 you picked which is like a really good one I think. Get rid of the version
36:43 thing. And then the negative prompt. Let's just take clouds. I feel like it's
36:47 pretty good. Cold, vinimal, corporate. True. Definitely don't want that.
36:52 Negative prompt. Prompt. Okay, let's run it and see what he thinks.
36:56 >> Let's run it. >> I just do ide for fun cuz it's so wild.
37:03 Okay. So, flux for most things. Ideogram. Okay. >> I love it.
37:07 >> That is really nice. >> Isn't it so nice? I >> I can't even tell how to spell cassette
37:12 anymore to be honest, but it doesn't >> I think it's two S's, right?
37:16 >> I think so, but it doesn't matter. >> Doesn't matter.
37:20 >> Oh my god, I love it. Um, let's do a few more. They're so fun. Like, you could do
37:26 pretty much any. Um, and then sometimes if you only like one
37:31 prompt from Claude, you could be like, "I like this prompt. Make another one, but maybe we have a
37:38 different style, but black and white only, cuz sometimes you might need a
37:43 black and white logo." >> Yeah. And like the gasp that you just
37:47 had like oh my god is because you have been a designer your whole life and have
37:53 spent so many hours, days, nights, sleepless nights, you know, thinking
37:57 about all this. So the fact that it can get done for like the cost of like three
38:01 pennies or whatever is just like mind in like one second is mind-blowing.
38:05 >> No, it's it's so true. It's mind-blowing. Like really the only
38:08 benefit of being a designer at this point is knowing maybe a little bit
38:12 about what to react to, but that's really really minuscule. Like I think
38:18 it's so insane. Like I'm not as obsessed with this, but still cool.
38:21 >> Yeah. Yeah. >> Cool. And then the other ones that got
38:25 made earlier were also really cute. Like all of the I mean this one was bad, but
38:28 this one's really cute. And you can kind of just feel out like what fits and what
38:32 doesn't. And I mean the benefit of making a mood board and all the like
38:35 earlier stuff of like how should it feel is it can let us look at this and then
38:40 this and kind of decide okay this feels more the vibe. >> Yeah. Cool.
38:45 >> Okay. >> So we have a logo. >> Okay. We have logo. We're good. Um let
38:52 Compositing the Final App Screens in Figma
38:53 me download this logo. Now we're pretty much ready for the final situation which
38:59 is making the screens in Figma. Um, so in Figma, if you press F, you can get
39:05 the frames. Let's just click iPhone and make a few iPhone screens. Add some
39:12 border radius to be pretty. And I'm going to paste in a little top
39:18 bar. Let's call it cassette. Just spell it differently everywhere.
39:21 And then >> how did you paste the top bar for people who don't know how to do that?
39:27 >> That's a good question. I have a really bad answer which is I have it in a that
39:34 I used forever. But you can also Figma community search iOS and like every iOS component on the
39:42 planet shows up that you can copy paste. >> Exactly. >> Yeah. And it's honestly more for it's
39:47 like you don't need to do it here, but it just makes it feel so much more
39:51 legit. Um and then let's bring these down so that we have like a bit of a side by
39:57 side. Okay, what do we need to do? Record audio history of what you record
40:03 timer. Okay, these don't need to be perfect. Let's first do record.
40:08 Actually, let's just like add our logo in. Cute. The problem is that it's a
40:15 black background and we don't want that. You can very easily in Weebi. Here's
40:20 another tip in Weebi. You take the result, you type in background remove.
40:27 >> Oh wow. And it just does it right there. And we can redownload that.
40:30 >> That's handy, >> right? >> Yeah. >> So cute. Logo background remove. So,
40:46 Very cute. And then we're going to want to color. Let's just use
40:51 something like this for now. and we can change it cuz that doesn't quite fit.
40:56 Something another tip that's Figma that I like to do a lot is anytime you have
40:59 an image and you're trying to figure out your color, go here to appearance and if
41:04 you go down and type overlay or hard light, you can kind of play with these
41:08 blend modes and you see like it changes the color based on the background. So
41:12 like even this for example feels a little better than that. Um but we can
41:16 change our color after based on our color scheme. Let's just get the
41:21 functionality down. So we have this like splash screen. Now let's do the record
41:26 audio button screen. Let's just go ahead and paste this directly for now.
41:32 And we can fix it later in terms of spacing and whatnot. Let's change the
41:36 background. And now we need the history of what you record. Let's just paste
41:41 this in and see how it looks. We kind of have the basics, but it looks
41:50 like we need to change a couple of the backgrounds. There is actually one more
41:54 cassette tape I liked better in our weevi that might work better. And now
41:58 you only see it once you're in Figma. cassette tape here. This one I ran into the same thing
42:10 earlier. Simplified cassette tape. And just for reference, I made that by just
42:13 saying like generate a new row of cassette tapes. And so you you really
42:19 just like iterate on your exact weebi um the more you go cuz in this one it's
42:24 like a little hard to see the spacing and stuff. So let's make another one
42:31 with simplified cuz that tape's adorable. >> Yeah, >> right.
42:36 >> Yeah, that looks really good. >> It's so cute. Okay. And maybe they can
42:41 like horizontal scroll it. But we also need to unfortunately remove the
42:45 background. Remove the background. And we're also going to remove the
42:51 background of our favorite one, which is this one. Um, and that's kind of like a thing you
43:02 should do all the time is make your favorite asset and then remove the
43:05 background at the end once you're putting everything together.
43:11 um to make it all cohesive. But while that's running, let's download this.
43:17 Call it clean. Go back to Figma. Paste in clean. Paste in our logo
43:29 at the top. Cool. So, this is our new version of history. This is our current little splash
43:37 screen. And then let's get the one that we removed the background
43:45 of down here. Oh wow, it removed everything. That's okay. I'll show you how to fix that cuz that
43:54 happens really often. Now we have this little button. Sometimes when that
43:57 happens, I actually just take this into Figma and remove the background.
44:02 Background removal, it's like so common that you just kind of mess with it until
44:06 it comes out right. Like honestly, if we just did it in Wavy again, it' probably
44:11 work too. Um, now let's get the color scheme we want. We already have the color scheme from
44:19 Weevi. Let's go back up to our color scheme. Just kind of copy rain shot, paste it
44:32 here, and kind of eyeball a better color scheme that fits the
44:37 vibe. Like something like that is nice to feel a little more warm.
44:49 And now I guess we need text. >> We'll literally just look at the vibe
44:53 coded one to kind of copy what text we need. Um, or we can just look at what we need to
44:59 do. So, you need to record audio, history, and timer. So, record audio.
45:05 Oh, well, we're missing a record button. Now, we know from the history screen.
45:14 Let's go ahead and add a cute little record button here. Maybe some T type. Some a helpful tip
45:21 for mobile apps is like I would just keep the typography very simple and the
45:27 same oops 20 same tip as earlier. A helpful thing for typography is for colors use blend
45:42 overlay and then it automatically is the color that fits well
45:49 or that screen. See, like this automatically looks like it fits really
45:54 well. And I just duplicated it. Something like that. >> Do you ever think we'll get to the step
46:00 where like the step before this, like the pre-Figma step where you'll just be
46:03 able to put into Google AI Studio and it'll just like figure it out for you?
46:07 >> Honestly, you could do that now. Like I think part of me likes compositing in
46:12 Figma so that when I am going back and forth with Google Studio, I have a north
46:14 star. >> Yeah. But like you could very much just literally get to this step, paste in
46:22 these reference images to Google Studio and kind of see what it does. It might
46:26 take a lot more similar to vibe coding like back and forth. >> Yeah,
46:30 >> I think it's similar to like how much back and forth do you want to have? How
46:33 much of a vision do you have for it? Like this isn't a necessary step by any
46:38 means. At the end of this, what I want to do is just compare what
46:45 >> output versus Google AI Studios output. >> Yes, I think that's a great idea cuz I'm
46:49 kind of curious like if we paste this in this in and like the logo, there is a
46:54 logo I kind of liked better that was generated here that I do think we should
46:58 use. Um, it's this little I think it's just that we need like a
47:05 black and white one honestly for it to be really simple. But honestly, either
47:10 would work. It's really just personal preference sometimes, but just so we
47:16 have some less colors going on. Let's remove that background.
47:21 And then the last text we'll add is library. And then paste it at the top. And there
47:33 >> Download it at the app store right now. >> Rate it. Fun stars.
47:36 >> Yeah. >> But I like that idea. Should we paste it
47:38 Pasting the Design Back into Google AI Studio
47:39 into Google? >> Let's paste it into Google. And then at
47:44 the after that, let's tell people if they actually wanted to make this
47:48 fully functioning app, like how would they actually do that?
47:54 >> Okay, let's do it. Remember you design the interface based on the logo color
48:02 image of cassettes record button attached. There's probably a much better
48:18 in-depth prompt we could use and make this work really well to be honest.
48:23 >> Right. And we would go to a code or or another LLM to go and optimize that for
48:26 us. >> Exactly. Like you could finesse it to the point where
48:34 you have pretty much everything. Or you could honestly just like upload
48:37 the Figma files too, but you really don't need to. And then you kind of see
48:41 what it does. >> Yeah. And then we wait. >> I wonder what it'll do.
48:47 I mean, if it could do that, it would make a, you know, the lives easier for a lot of
48:54 people. Um, the way I see it is it's kind of like, you know, do you want to
48:59 There's like you can drive a manual car, you can drive an automatic car, or you
49:03 can drive a self-driving car, aka you're not driving out at all, full
49:05 self-driving. >> Um, and you're just like managing the
49:10 wheel. And like now it's kind of pick your poison based on like your level of
49:16 comfort and all the and all these different tools. it's just like whatever comes up with
49:24 the best output. Like you could design the whole thing in Figma and then
49:27 suddenly like you get a one shot that's just better or vice versa.
49:32 >> Yeah. Like I found that shaders especially in Google AI Studio, it's like the one
49:39 thing where I found that it just oneshots it so well that I'll almost
49:46 never start with an existing thing. >> And then have you played or have you
49:52 seen you know any of these mobile vibe coding app platforms like
49:57 vibe code app or there's anything.com. Have you seen these? >> Yes, it's a good point. I have seen
50:02 them. I haven't played around with them in extreme depth. I've like tried a
50:07 prompt or two. >> Yeah. >> And I know they're getting better at
50:11 backends and stuff like that, but that's usually always my um hold up with them
50:17 is they'll feel like a little almost like the old version of no code
50:20 sometimes. But honestly, they're getting so much better that I'm sure it'll
50:24 evolve a lot. >> Yeah. Yeah. It's on my to-do list to like try it out. But I have seen like
50:31 some of them are built on cloud on Opus 4.5. So it's like you got to think it's
50:35 pretty good, right? >> Yeah. No, they're probably really good.
50:40 And I feel like in the past 6 months like Claude has gotten so I mean I guess
50:44 all the models but at least from what I use Cloud's gotten so much better at
50:48 Swift and iOS whereas I feel like web used to really be like
50:55 substantially better. Um, and now it's like really good at iOS, too.
50:59 >> So, the for the mobile apps that you're building now, you're using cursor.
51:03 >> I'm using Claude Code is my number one love of my life. My number two love of
51:08 my life, cursor with Gemini model and plan mode. And then for oneoff shots,
51:13 one-off prototypes, I'll use like Google AI Studio directly.
51:16 >> You can only have one love of your life. You can't have two. That's That's
51:19 blasphemy. >> No, you can't. But yeah, it's Claude. It'll always be Claude.
51:25 Okay. What do we got? What am I looking at here? >> Let's see. I mean, it got the logo. It
51:30 got the orange. >> Okay. Honestly, it's like much cuter.
51:36 >> Yeah. Oh, 100%. >> Wow. Oh my god. Life is so crazy. I
51:41 can't believe it does this. >> I mean, it's not perfect.
51:47 And I still I think the Figma version is a lot better, but I I think that you
51:52 probably can pingpong this to get to a point where it's almost if not Figma
51:57 Figma level >> 100%. Also, like our prompt was so short.
52:04 >> Yeah. And we're doing this live, so you could point is people listening, you if
52:09 you want to go back to AI Studio, feel free. You don't need to be in Figma.
52:14 >> Yes, 100%. Like if you spent one more minute than we did on the prompt, I
52:18 would have done it. >> Amazing. Um, Teria, anything else you
52:20 Final Advice: Spend Time on Inspiration, Then Build
52:23 want to leave people with? Anything else that you didn't cover that you wish you
52:25 did? >> Um, good question. No, I would just say that spend lots of time on Cosmos. Spend
52:32 lots of time gathering your own inspiration. Um, that's like actually
52:37 just a lot of it. And then >> yeah, >> you can pretty much make anything you
52:41 want. Yeah, it sounds like it feels like what I've learned is it's
52:46 a game of kind of like collecting the right inspiration and then using
52:54 like maybe sleeping on it and being like, okay, this photo, this picture,
52:59 this and then getting the insights from there and then and then just bringing it
53:03 back to AI Studio or Figma and and that's what it is >> 100%. It's so simple. It's like really
53:10 you can take one visual reference and then just like make everything based on
53:13 that. >> Well, thank you so much for for for coming and showing us all this.
53:21 I think it's going to make a huge difference in people's products lives
53:27 and uh and I'll include links for where to follow you on your journey in the
53:32 show notes in the description. highly encourage people go and follow Sariah
53:37 and her journey. She's criminally underfollowed, criminally unknown, and
53:44 but people who know you know that you're an incredible designer. Uh, and I'm just
53:49 honored that you came on the show, that you shared a lot of sauce, and that I've
53:55 gotten to know you over the last five or six years. >> Nice. Thank you. You, too. I'm so happy
54:02 on. I I hope you come on back on again and uh people listening, I hope you go
54:07 and build something beautiful and and I know you will now. So, thank you so
54:09 much.