Hacker News Reader

Comment by dawsonwhitfield | Hacker News Reader

Hey guys - Dawson here (founder) let me know if you have any questions about Logojoy / the story.

Have you looked into whether this is legit on legal usage? You appear to be using noun project icons in these logos, and afaik most logos on the noun project are creative commons with attribution (and I confirmed by selecting a few logos with icons which matched this license on the noun project site).

It doesn't appear that you provide any attribution at all to the icon creators wherever you display the icon, and I have no idea what license you deliver to the end user (I haven't went through purchasing a logo - knowing that you're just sticking some text with a noun-project icon doesn't strike me as worth it, honestly). Are you in the clear to be displaying these icons with no attribution, and are the users that buy your icons clear to be putting it on websites and business cards without attribution?

I haven't even looked into whether your font usage is appropriate, can you clarify that as well?

The Noun Project has an API plan which allows for usage like this I believe: https://thenounproject.com/developers/

Squarespace, Tailor Brands and some other tools are using the API for logo design (without the attribution) much like what Logojoy is doing.

nounproject has a paid plan with royalty free and no attribution needed.

> I haven't even looked into whether your font usage is appropriate, can you clarify that as well?

I did a test and the first result it gave me was just my company name in the Disney font.

You talk about the ML being reinforced over time by user actions so you must have suffered from the 'cold start' problem in the beginning. How did you address this? Did you initialize it with your own expert curated rules/weightings? Something else?

Have you done any automated mining of the elements in existing "good" company logos (e.g. Nike, AirBnB, etc.). I could see doing something manual too like Music Genome Project or Netflix's internal curated movie characteristics tagging. You'd go through assessing these gold standard logos for qualities like contrast, amount of whitespace, if brand name intersects the artwork, etc and then augment your model with these style rules.

Any insights you'd be willing to share would be appreciated!

Yeah, the ML is still in it's infancy, so we're almost in that cold-start right now. To address that, I built in some more randomness into the algorithm so the initial logo designs can just get the user thinking and then we can get a better sense of their taste as they favorite logos. Without that randomness, the algorithm tries and fails horribly at creating logos.

I would love to analyze 'good' logos. I would imagine it's realllllyyy hard to analyze actual pixels. We just analyze actions and pre-created objects (ie. a font choice).

Actually, maybe not, if you have a large dataset that you are willing to share.

Analyzing pixels is something ML has gotten much better in the near history :) Even if you have a very simple recommendation engine running behind the scenes, you can throw in some convolutions to utilize your data to create a feature space just like word2vec, which would give you great power to generalize. Even putting in pre-trained networks like VGG16 (instead of training one from scratch) could give you a great headstart on this.

So, I went through the whole logo creation process and got stuck at the sign up screen. I don't mind signing up, but when I used last pass to generate a password, I got told it was too long. I only tried to generate one that was ~9 characters long. If you are checking length, might want to put the requirements on that screen

I got past the signup screen but was then presented with a scrolling list of empty logo options (testing on my iPhone 7).

I thought the same, but it turns out you need to type a word into that input box to see the logos. The words scrolling are just examples you could type.

Oh really? I ended up skipping it because I thought they weren't loading. Should at least have a blinking cursor.

> So, I went through the whole logo creation process and got stuck at the sign up screen. I don't mind signing up ...

Yeah, I stopped here too. I'm not going to hand over my details sight-unseen ...

In Chrome, Right-click the edge of that "Sign-up" div, inspect, find and select the containing div if you didn't hit it already, and hit the delete key. I was able to use the site and browse more styles and customize from there.

Circumventing access controls, probably worth mentioning that's unlikely to be lawful in the UK.

Really you should not be checking length at all. There is no such thing as a password which is "too long."

From a password security standpoint, I largely agree. You can open yourself to system attacks by allowing arbitrary length input. You should have some kind of limits on any user input.

Sure, but that limit should be closer to hundreds of characters—not 9.

Agreed :)

Hey! Just a heads-up, there's a small typo on your landing page-- It says "They chose Logojoy because it's of how easy it is to make a great logo." That "it's" is superfluous. Anyways, great product! Congrats on your success.

Oh damn. Thanks. How did I not see that. Well, can't update it right now :\

I tried it, but got a Firefox is not supported message. Why not? What's the missing feature?

Otherwise looks pretty cool! Nice business, it's inspiring.

came here to ask the same question

Their is a login form on the website on a page without HTTPS. You should change that.

Also, no HTTPS when you are supposed to enter the credit card details. This stopped me from buying the logo.

Actually it is an iframe with https (using stripe)

But if the host traffic is compromised you can serve malicious payment gateways/forms.

If the host is compromised HTTPS isn't going to save you.

it will if it's the host /traffic/ that is compromised

Hey, awesome work! Your story is inspiring :)

Having trouble accessing the site though: "logojoy.com" does not work, but "www.logojoy.com" does. Unfortunately Indiehackers links to the one without the "www". Just wanted to let you know.

I was quite concerned the checkout page didn't use any encryption at all, which caused me to cancel my purchase.

What can you tell us about the algorithms you used or where did you start on learning A.I. ?

Interesting approach for logo creation!

I don't know about Logojoy in particular, but every AI begins with some variation of rand(). :)

Not sure why you were downvoted, in this case that's a real possibility. It's hard to beat rand() for MVP.

From https://news.ycombinator.com/item?id=13067481

> Yeah, the ML is still in it's infancy... To address that, I built in some more randomness into the algorithm so the initial logo designs can just get the user thinking and then we can get a better sense of their taste as they favorite logos. Without that randomness, the algorithm tries and fails horribly at creating logos.

Some variation of rand() indeed!

From the Logojoy How it Works page.

    The last step isn’t actually AI – it’s a genetic algorithm. As you ‘favorite’ logos, we apply a natural selection type process to generate more accurate logos as you scroll. Each logo is made up of hundreds of ingredients (ie ‘funky font’, ‘bold color’, ‘logo has a symbol’, etc.) – these are called traits in the context of genetic algorithms.

As others in the thread have already mentioned, the AI referenced in the previous steps seems more like a rules-based algorithm.

Dawson, just came here to just say that I am super impressed with Logojoy, and especially how the onboarding process is both so fantastically smooth AND it actually IS the product! Very cool.

On the flipside, this surely has the potential to put thousands of cheapo logo designers out of (that line of) work.

Fantastic work! Landing experience (result first, registration later), logos seen working in context (blanks, packages, etc), overall smoothness... even the support guy notifying about possible outages during to HN hug of death, everything is so perfect!

You are a natural talent, and I envy you. :)

Haha thank you! It's been a lot of work. PS. how bout that support guy - what a cool dude

I tried "lûd" as the company name, and that works but many of the logos show "ld". It looks like some of your fonts don't have accented characters.

Please stop having Javascript take over the scroll event in the logo generator. I am having to throw my mouse across the room to even barely scroll down the page.

My back-of-the-envelope calculation says that you are losing money on adwords, unless about 50% of all sales are enterprise level, in which case adwords breaks even.

1800 visits 25% signups * 0.5% sales = 2.25 sales* for $250 spend on adwords.

He says he gets 5000 visits a day and the 25% / 0.5% stats are related to the 5000, not the 1800 from AdWords.

The AdWords visitors likely look different and sometimes it takes experimenting time and money to find an effective way to make a profit on AdWords traffic, so taking a small loss or breaking even in the short term isn't so bad. Acquiring customers has some value besides what they purchase that day as well -- word of mouth, repeat orders, etc.

I'm more worried about him thinking he can 2-4x conversion rate "within the next month"!

Really well done! I was just playing with it and noticed that when I combine an inline icon with a name, the icon hangs way lower than the middle line of the next. Is that a bug? A design choice I don't understand? As another point of feedback, some spinners when searching icons and loading logos would be really nice. Maybe because of the HN effect or maybe for some other reason, but it took a few seconds for lots of actions and I wasn't sure if stuff was supposed to be happening or not. But the product is fantastic and I hope to use it soon.

hey Dawson - would love if you could go into more technical depth on how things work. Cool product!

Absolutely. I talk a lot about the tech in the IH article, but I'll re-iterate here with as little jargon as possible, focusing on the machine learning part, as I assume that's what you're more interested in.

The logo generating algorithm is still in its infancy, so right now it's mostly acting as an idea source for most users.

So our logos are just combinations of ingredients — fonts, colors, layouts, symbols, etc. Logojoy uses machine learning specifically to learn which ingredients go better together.

It starts with tracking basically everything our users do. We track things like: the inspiration they selected, the logos they favorite, the changes they make to logos (e.g. changing the font), the time spent looking at certain logos, the commonalities between all of their favorited logos, the logos they purchase, and more. We currently track about 80 types of actions.

Every day, the learning algorithm reads all these actions and weights each one by how many times it occurred. Because of the structured way it reads actions, it's able to define rules based on "heavy" actions.

For example, an action might be defined as "user changed font weight from 100 (light) to 600 (bold)". This action object includes the number of times it occurred, every other preceding action, and all of the logo's ingredients. In this case, let's say the algorithm concluded that every time this happened, the color of the logo was classified as "light". The algorithm would presumably define a rule that says "if the color of the logo is light, do not use a light font weight".

Is it just rule-based, or are these "weights" being used to train a classifier, ANN, or some kind of genetic algorithm? Nothing wrong with the former, if it works for ya - but the latter would be interesting, too!

Based on my previous experience, my guess is that the latter would be much too blurry, because generative models tend to only output blurry images.

thanks for your response!

Could you say anything more about the ML portion? I'm curious how you account for the temporal/chaining aspect of actions, what kinds of learners worked well, etc. On the implementation side, did you find a usable ML library for PHP or roll your own implementations?

You may want to try using an ML algorithm called association rules, which produces rules automatically. Though acccounting for the sequence of events would be harder.

This is great. Sorry if this is a noob question but what did you use? Is this all coded in raw php? Or is there some sort of combination of tools/libraries you used to make this happen like tensorflow?

..know anyy good ML tutorials? this all sounds very interesting.

So I tried it. It seems like almost anything is going to require designer interaction. The logo I ended up with has such low contrast with the background I would be embarrassed to even present it as a mockup, much less treat it like final product. I couldn't get logos in containers, which seems obvious, and very simple to do from what you're already doing. The color picker is suuuuper dumbed down, it'd be nice to see some true color palette options in the vein of Coolors (many other examples exist but that one is representative of what is doable pretty easily and light years better than picking shades of yellow). I could imagine giving this to someone and having them play with it for 30 minutes to get an idea of what they like, but right now unless they want a company name in a single color format (and even then many color options have no legible results) you aren't going to finish anything there.

Do you plan to allow editing the stroke color independent of the background? This is the most immediately problematic issue for me, because it means I cannot get a high contrast logo in any light color. Maybe I missed a way to do this?

Looks great! Made a logo in like 5 min that looks amazing. Letting me share it before buying lets me discuss with other people on the team before making a decision.

You mention trying to compete with designers "on the same level", but this seems like a different space than design. I do logos for my clients and most of the work is helping them make choices about what they should do with their logo, not actually drawing it. Have you considered partnering with designers instead of seeing them as competitors?

Designers could easily leverage this as a starting point, then refine the idea more accurately, down to the pixel.

Hi Dawson - What is your background? What's your team consist of?

It's just me right now! Hoping to bring on a developer and someone to help out marketing it ASAP.

I've been a designer for about 12 years (doing lots of logos) and developing for about 8 years.

Trying the site with chrome on android, some frustrating issues:

* as soon as logos appear the signup form covers all of them, so I can't see the logos. Is it by design? I tried signing up and the wheel keeps spinning, I'll retry later.

* Since I tried multiple times, the browser allows to autofill forms. When I do that the js validation fails thinking I haven't entered any char yet.

* not sure what to do on the Selectr-icons-screen, I've tried clicking here and there with no results. Luckily I can skip the step :-)

* Reloading the page (eg I'm wondering if the server is overloaded and I want to give it a chance) the logo generation restart from scratch. Pls consider keeping state in the URL.

Anyhow, great idea :-)

Hey Dawson. I'm trying to create an account on logojoy, but it looks like you're blocking emails with a + (plus) sign as invalid. This bugs me, as I like to give each app a unique email address.

I would love to know more details on the tech side. Did you use a machine learning framework or api? Just would love to know your process as a dev looking to incorporate some machine learning.

Is it mostly a solo effort?

I am releasing my first larger project soon - any tips in terms of release, first places to post to spread the word etc? Also top X things you would have done differently?

Also one step that did not quite work for me was picking the icons (it said pick 3 icons you like or something). There were three cutouts and words flying by, but no images and there were no selections available. Here is my browser version: Version 55.0.2883.44 Ubuntu 16.04 (64-bit)

You are expected to enter text in the text-box above where the text flies by.

As a usability feature, I believe that it should auto-type suggestions (which fly by) in the box when you click on them.

It doesn't help that the text box is white (#FFF) and the background is very light grey (something like #DDD). Hard to see the box.

Ohh, thanks that was it :). I thought it had something to do with my browser...

This is really cool -- my major issue has been that it doesn't preserve case sensitivity. So logos like eKitten don't seem to work :(

Only works in Chrome...

tried it in firefox and it worked fine.

I simply ignored the "chrome only" warning

If true, that just makes the "chrome only" warning even more obnoxious.

The tooltips over the color selector's bottom row are cropped in Firefox but not Chrome. The page seems to have more whitespace at the bottom in Chrome.

Might be caused by all the traffic, but received this error when clicking "Make Your Logo"

>Connection Failed: 1226

I received the same error, but refreshing the page took me straight there.

Hey Dawson, small world! I worked at RHINO for a bit last year. Were you working on this at that time and what was the time commitment to it back then? You mention two and a half months but given the freelancing I’m guessing it wasn’t full-time effort?

Hey! Small world for sure. It was 90% of my time, I did a little freelancing for existing clients. I started work on it in September 2016.

Dawson, how did you decide what were the simplest things that needed to be done first?

Great question. I've been designing logos for 12 years so I have a good sense of what's important:

1. Amazing fonts. This is the low-hanging fruit, and the core of every logo. Some logos are just nice fonts :P

2. Brainstorming and idea exploration. Logojoy is cool because it gets you thinking, and makes it a joy (haha) to explore different fonts, symbols, colors, etc.

3. Presentation. If you look at how the top agencies present logos to clients, it's always with a ton of whitespace, real-life mockups, etc.

Fonts are expensive. Are these free fonts?

Signup didn't work for me. I was stuck there.

Love what you guys are doing, I tried it out and ended up with way better logos than any of the designers we found.

FYI, I thought I'd give it a try today, but it appears to be broken (syntax error in the console).

Might be a little finicky due to all the HN traffic, please try again, I promise it will work eventually.

Is there a way up prevent the design mock-ups on the preview tab from capitalising the first letter?

just tried it and it's stuck at only two logos but it says it's generating more logos :/

would love to know if there's free alternatives like logojoy, I think it's nifty

Click into a logo, then click back and it should work. Gets stuck sometimes :(

Reply on news.ycombinator.com