To add initial images, we follow these simple steps:

1 - In your Google Drive, go to the init_images folder inside Disco_Diffusion in Google Drive and upload the image or images you want to use later.

You can also upload them directly from here, but I recommend the first option to have everything much more organized:

It is important to name the pics with a simple and clear title to make the images easily locatable.

2 - Once you have connected the tool with your Drive (1.2 Prepare Folders), on the left hand side (where the folders are) locate the image in the init_images folder: (content>drive>MyDrive>AI>Disco_Diffusion>init_images).

Once you have located the image you want to use as the initial image, click on the three vertical dots icon and then, click on 'Copy path'.

3 - When you get to the 'Settings' configuration, in the first point of the third block 'Init Image Settings' called init_image:, copy between the quotes the path of the image obtained in the previous point. Right there:

Activate the 'Settings' section, so that the algorithm takes into account the initial image.

4 - Play with the prompt and activate.

5 - Activate Diffuse… and that's it!

Now, let me show you a couple of examples.

Let’s start with the following initial image:

And we give it a robotic and futuristic touch:

["Robotic and futuristic redhair girl and her butterfly, 3D"]

Or this window with views to the forest:

To which we give some views of Paris:

["A beautiful painting of a window, the city of Paris is outside with the eiffel tower, photorealistic, cinema 4k, final render, detailed, amazing, masterpiece, artstation."]

Stay tuned, we'll be posting tips and tricks soon to get the most out of init images!

@joseafernandez

joseafernandez.net

Be careful not to omit any symbol and open and close square brackets. If we fail in any detail, the code will not be executed.

An example of the basic structure of prompts is:

["A beautiful painting of a singular lighthouse, shining its light across a tumultuous sea of blood by greg rutkowski and thomas kinkade, Trending on artstation.", "yellow color scheme"]

What does this mean?

First, we give instructions between quotation marks and followed by commas. In this case, we give two instructions:

1. “A beautiful painting of a singular lighthouse, shining its light across a tumultuous sea of ​​blood by greg rutkowski and thomas kinkade, Trending on artstation.“

   Description of the image we want to achieve and the artists we intend to influence our work. In this case, the description of the image is 'A beautiful painting of a singular lighthouse, shining on a tumultuous sea of ​​blood' And the artists that will influence the final result are Greg Rutkowski and Thomas Kinkade.

   Important: End with Trending on Artstation. so that the AI ​​knows where to go to get the artists style.

2. "yellow color scheme" - Color palette we want. We can include more than one color. In the case of the example, the yellow color is chosen.

And this is a final result:

Through the same prompt, we get as many unique random combinations as we want. We just have to let it act:

Tips and Tricks!

Depending on where you put words in the description (as we said before, each instruction should be enclosed in quotes) that order will have more weight or not. For example, the lighthouse painting is more important than the artists it will influence, being placed first. That's why the general description always has to come first.

Following this basic prompt structure (it can change a lot as we add more complexity but we will see this later), include some words in your prompts that can give the final result a different quality. And put it after the description in the place that you consider optimal.

• For the type of style you want:

  Add words like 3D (for 3D results), or painting, or illustration style, cartoony style, realism style, abstract style, futuristic style…

• To get more realism:

  Use words like 8k resolution, cinema 4d, unreal engine, hyperrealistic, photorealistic, detailed, cinematic, final render, octane render…

• To get light effects:

  Beautiful lighting, shadow depth, electrical light…

• To provide more details:

  Aristocratic, robotic, futuristic city, dirty, ancient, cyberpunk…

  Any adjective will help the AI ​​to better outline the work.

• Choose the artists carefully.

  The style of the artist or artists you choose will be reflected in the final artwork. It is key that you open an account in Artstation and start following the wonderful professionals that you will find there.

  And reminder!: Always end with Trending on artstation.

  If you want to make realistic work, choose artists like Fabricio Rezende, Alexander Jansson, Ivailo Ivanov…

  If you want to make a video game or fantasy landscape, choose artists like Damien Mammoliti, Piotr Krynski, Marvin Wärzner…

  If you want to include colossal light effects, include artists like Rutger van de Steeg, Sylvain Sarrailh, Job Menting…

  There are dozens of them. Your choice is key to achieving a more impressive final result. We will talk about all of them more deeply in the near future.

  More examples but with classics like Van Gogh and Monet:

[A beautiful painting of a castle and tulips by Vincent Van Gogh]

[A furious and hungry shark painted by Monet]

• Finish with a touch of class!

  Add words like gorgeous, amazing, masterpiece, wonderful, magic…

In all your descriptions you can use one or several words that we propose there. The result will always be different.

For example, the image I have created of Lucio Leone (The Don of AI Mafia Club) was created with the following prompt:

[“A dirty and dangerous old man, aristocratic, storm and fog, greg rutkowski, artstation.”, “magenta and dark brown color scheme”]

And to practice just go crazy! Creativity has no rules! Like the following work of the great artist DeadbeatDadClub:

[3D ghostly astronaut lost in space made of flowers]

Many more tips and tricks soon! And we will experiment with more complex prompts that will allow us to further refine our technique.

Stay connected!

@joseafernandez

joseafernandez.net

Thanks to Disco Diffusion we will be able to create spectacular images and videos, and to do so, you can team up with the artists you love the most to help you complete our masterpiece.

You will do things like this:

O esta:

Disco Diffusion is a complex tool and we will explain it step by step in other posts, but now we are going to focus on the most important and essential stuff you have to do to start this adventure with your own creation.

Important notice: This is addictive.

1 - Changelog: First, you need to have google drive. If you don't have it, signup on gmail. Without this, you cannot use Disco Diffusion.

After that, click on the following link to access the tool (use Google Chrome):

https://colab.research.google.com/github/alembics/disco-diffusion/blob/main/Disco_Diffusion.ipynb#scrollTo=DoTheRun

And here we are. Let's skip the whole introduction and stuff for now and go directly to the Changelog section. Simply click on the play button next to it (it is essential to activate all the sections needed).

You will get an alert message warning that this notebook has not been authored by Google, but click on 'Run Anyway'.

Wait for a green 'check' icon to appear indicating that it is activated and go to the first 'Set Up' section.

2 - Set Up. Do not touch anything here, just activate each of the sub-sections by pressing play (as we have done with Changelog).

In each one you can take a look at the code by clicking on 'Show code' if you want, but it is not necessary.

In the second sub-section (‘Prepare Folders‘) it will ask you to connect to your Google Drive. Follow the instructions to connect.

I recommend that you activate one by one and do not go to the next until the previous one is active with the 'check' icon in green.

Once you have everything active, move on to section 2.

3 - Diffusion and CLIP model settings: Do not touch anything, simply activate as we have been doing until now.

Don't forget to also activate the 'Custom model settings' (last sub-section).

4 - ‘Settings‘: We are going to focus on the right hand side block.

Here also click play without changing a thing. But it is interesting to explain a couple of parameters that are going to be key for the future:

In the 'Basic Settings' we find:

Batch_name: You write the name of the folder where you want your images to be saved (once generated) in Google Drive. If you leave the TimeToDisco that appears by default, that folder will be created inside AI>Disco_Diffusion>images_out>TimeToDisco.

Steps: They are the number of partial images that will be used to make the final composition.

If you add more steps it means that the final image will be more detailed, but it will also take more time to generate.

For example:

An example of an image with 900 steps:

With practice and depending on the style you want to have, you will find the optimal number of steps for you. For now, we better leave it as it is by default (250 steps).

Width_height: The dimensions of the image. You can make them rectangular, square… It is very easy to change and they must be multiples of 64, but if you make a mistake, the tool corrects it for you. Don't worry. For now, let's leave the default size.

We also do not touch any of the following values, which we will also analyze soon. We click play, wait for it to turn green and go to 'Animations Settings' and 'Extra settings'.

In neither case do we have to touch anything. Simply, we put the mouse on it, the play button will appear, and we click it.

If we want to make sure that it is active, we display the interior and see the green check icon.

And then we move on to the most important part. The prompts.

Prompts: The instructions that we whisper to the AI ​​to start drawing. The prompts are infinte and constitute the basis of the quality of the final creation. We'll cover this in a lot more detail in the near future, but for now, let's keep it simple.

First, in quotes, say in general lines what you want to create. Continue with some of the artists that you love the most (contemporary or classic) to help you model your work, ending with 'Trending in artstation' (I strongly recommend opening an account in Artstation and follow the ones you like the most) and we finish choosing a color palette.

You can leave the one with the lighthouse you will find by default. It is spectacular. But if you want to try something different, I suggest, for example:

"A beautiful painting of a futuristic port, full of ships floating on a calm sea, surrounded by mountains, by Harry Dafereras and Petar Milivojevic, Trending on artstation.","purple and blue color scheme"

5 (last one) - Diffuse:

Here we do not touch anything either and we click play to start everything we have configured above.

And voila! Let's let the algorithm act. From point 5 (‘Create the video‘) you do not have to touch anything since we are not making a video.

If after a minute (more or less) a blurry image like this appears, everything is fine.

We wait and we will see in real time how our image is transforming. Until the final result:

If you let the tool act, it will launch more images with other combinations.

And that's it! Try it yourself, change the number of steps, the dimensions, the prompts... and you will be surprised.

I leave you some examples of works by amazing artists with great mastery of the tool:

Subscribe and you will get to do magic!

@joseafernandez

joseafernandez.net

In this case, we've used Disco Diffusion to create concept art drafts that help the writer bring their world to life. Not only applying his idea of ​​each piece of land, but also providing unique and surprising elements which can only result from the participation of artificial intelligence. So Generative Art not only gives life to ideas, but also provides inspiration. I'm sure parts of the book's story will change thanks to the AI's input.

To make a brief introduction to the world, Borhealis is a fantasy universe that we have been working on for several years.

Borhealis is composed by 6 continents. Five of them built on solid ground: Endellion, Selvha, Besuria, Allicaria and Hellatia (the ancient Humaria), and the sixth one in the sky: L'Haddes.

Each continent has different species and races, climate, resources, geography, which are immersed in endless wars with each other... and a circular city in the center of the world called New Mestiza, which constitutes a meeting point between the different species of Borhealis. 

So I shared my ideas of the different continents we want to create with the artificial intelligence, I got the help of some incredible pros like Alexander Jansson, Dennis Halbeck, Rutger Van de Steeg, Sylvain Sarrailh... I whispered to the computer some instructions and... 

Note: No init images! (is the new 'No filter' :P)

Birdcourts in L’Haddes - AI Concept Art

L´Haddes is the smallest but the most important and powerful continent in the Metaverse. It is built in the sky. Borhealis is ruled from there. It is joined to the surface by an infinite tower.

New Mestiza City - AI Concept Art

The most important city in Borhealis, where species from different continents coexist. It was rebuilt after the Massacre of the Guardians.

Lost-Lhiam in Selvha - AI Concept Art

Selvha is a continent of endless and dangerous jungles. Nature is extremely dangerous. You need a native to survive more than one hour.

Sumetra in Allicaria - AI Concept Art

Allicaria is composed by deserts, oases, cities of sand… The caverns of hell are famous because no one who has entered them has ever return.

Robacco in Besuria - AI Concept Art

Besuria is a continent made up of oceans, seas, islands and pirates. In Besuria the most paradisiacal islands coexist with the most dangerous borhealians and creatures.

The Fist of the Titan in Hellatia - AI Concept Art

Hellatia is a cold white desert. Except for the frozen forest that formed around the Fist of the Titan.

Endell in Endellion - AI Concept Art

Great valleys, trees, lakes, forests. Cold and humidity are predominant and different kinds of human species and settlements.

We will also be able to write novels with the help of an artificial intelligence and the best writers in the history. 

And that will be possible very soon!

If you want to know more about Borhealis, I strongly recommend you to visit the writer blog, Bretton Waterhouse, by clicking here.

@joseafernandez

joseafernandez.net

Talking about art, we are now experiencing extraordinary moments that can change the way of creating in the incoming times. In my opinion, we are living the beginning of what is going to be the art of the future. And by that, I mean AI Art / Generative Art.

What is AI Art?

AI art is art generated with the help of an artificial intelligence.

As you know, artificial intelligence is a branch of computer science that deals with creating algorithms that mimic human intelligence. This discipline has progressed exponentially in recent years, reaching unimaginable milestones.

In the field of art, you team up with an artificial intelligence to do teamwork whose result is unique that provokes emotions (what Art does) and in some cases surprising and beautiful.

There are many platforms where you can create AI Art. I particularly work with Disco Diffusion and the basic way of doing it is:

1 - You think about the work you want to draw.

2 - You configure a series of simple technical options.

3 - You whisper to the machine a series of instructions

... and voila:

(Examples of images that I have created with Artificial Intelligence with the prompts included)

@joseafernandez

joseafernandez.net

And I insist, these actions are based on my personal experience. If you have any other actions I've missed, I'd be happy if you share them!

The use case that I am going to show you is again from Chekin, and it is about Upselling.

Upselling is one of our most promising features and we can define it as the skill or the art to persuade a customer to purchase something extra, something additional or something more expensive.

In the hospitality sector,  we can see Upselling when a vacation rental or hotel prompts additional services to guests, such as an earlier check-in or late check-out, a session at the spa, a breakfast buffet, taxi to airports or train stations... etc. Vacation rentals or, more commonly, hotels, can also upsell rooms (otherwise known as room upgrade) to guests with the same intention as with Upselling. 

Actions

I am going to expose a battery of actions that are easily exportable to your company. The concept goes through generating interest and intrigue thanks to an attractive communication design and arousing the feeling of exclusivity.

So, to include an early access to anything always catches the attention of any user, regardless of the sector.

With the following actions, we not only intended to generate interest, but also measure it (that is really important).

First, we have to differentiate between current clients and potential clients, with personalised actions for each of them.

Current customers

-Dashboard

When the hosts enter his dashboard (the reservations management platform they visit on a daily basis), we showed them the following modal in which we urge to get an invitation to participate:

Also, we created a landing page within the dashboard that explains the Upselling feature more deeply, including links to video, ebook and post. Also the main CTA directs you to get an invite.

-Chatbot

Within the dashboard, we configured the chatbox with the following message:

Hi there! I have a secret…

We have Upselling ready for early access! Would you like to get an invite to try?

-Mailing

It is difficult for people to open promotional emails from brands (even when it comes to big ones like Amazon, Google, Apple...). For this reason, we must generate interest from the 'Subject' to try, first, to get the highest number of openings of the email possible, and then, the highest number of clicks on the proposed CTA.

The email must be personal, signed by someone real with name and surname.

Subject:

'Generate more money from each reservation. Upselling is here 🎉 !’

-Marketing campaigns

It is important to offer a series of incentives which will help us a lot in this first stage. All accompanied by a reward of course.

For example:

• Do you want 6 free months of Chekin? Show us the most original deal we've ever seen!

• We give 6 months free for all your properties to those who sell 20 deals in a month.

Potential customers

-Web

To measure and generate interest among potential customers, gain more registrations and we added a popup to everyone who enters the web for a period of one month.

Also, in the Upselling section, we included an alert message 'We have the early access ready' the CTA will be 'Get an invitation':

-Blog

I did write and signed 10 original and useful posts. They went published from 8 weeks before the launch (1 per week).

This not only increased the obtention of new clients interested in the feature, but it helped (and keep helping) to the current clients on their Upselling activities.

These are the titles we published:

• What is Upselling for hotels and vacation rentals?

• Upselling Tips and Tricks: Tricks to create a winning offer

• Upselling Tips and Tricks: How to select the perfect image for your offer?

• Upselling Tips and Tricks: Why a Smart Availability Strategy Means Success

• Upselling Tips and Tricks: What channels to use to Upsell in the different stages of a reservation?

• How to improve your KPIs thanks to Upselling?

• Upselling: Product launch

• Upselling Tips and Tricks: Some Creative Deal Ideas

• How does Upselling work on Chekin?

• How does Upselling increase my direct bookings?

You can take a look to the posts by clicking here.

-eBook

I also created the content and style of an ebook very useful for the current and potential customers. The ebooks are really appreciated by the users.

You can see or download the eBook by clicking here.

-Social media

Following the Social Media Marketing Plan, we released actions in LinkedIn basically, that is the main social network where our clients are.

-Adwords

Nowadays it is important to configure an Google Adwords campaign. Specially when you are launching a product or feature.

-Webminar

We organised some webminars with clients to explain them what is Upselling and the best practices.

-Press

Depending on the budget, it would be great to get mentions from influencers, early adopters, specialised media…

And that’s all. To guarantee an optimal product launch, the operation must start at least a month before and then measure the actions and act accordingly.

Thanks for reading!

@joseafernandez

joseafernandez.net

What’s is conversion about?

Imagine a funnel. A specific number (X) enters the wide end and that number is modified inside the funnel until another number (Y) comes out of the small end. The conversion is the percentage of X that comes out the small end.

For example, in the case of our guestapp, one of the funnels we analyse is related to the people who start de online check-in process and how many finish the process with their registration (in this case, the conversion is 38,52%):

Traffic > Clients

I like to say unique funnel, to the one who convert traffic into clients. So, we measure the number of new users we receive in our website and how many of those users are converted into clients.

Depending on the type and complexity of the company, that single funnel should be broken down into other sub-funnels to address improvements individually.

So for example, at Chekin, we divide the unique funnel into 3 sub-funnels:

Funnel 01: From traffic to lead

How many new users who enter our website show interest by leaving their email addresses, becoming a lead. The users can leave their email in the register form, bot, contact sales, more forms...

Funnel 02: From lead to trial

What number of leads finish the registration process and start the 7 days free trial.

Funnel 03: From trial to client

What number of users who started the trial, finish it and become customers through an active subscription.

So first, we have to measure the different funnel properly and once we have collected that data, we plan to attack each funnel by designing and proposing some changes that could improve that data.

Once the design is ready, we set-up A/B tests to understand, with real users, if our proposal improve the conversion or not.

If it does, we can then open that design to everyone, with the certain that it will improve the experience.

In the past, we designed what we liked, what we saw as beautiful from our point of view. But in a monitored world, we can take advantage of A/B testing tools to understand if our designs, even if we love them, really improve the conversion of the company or on the contrary, they worsen it.

Many times we get surprises, because what we think is better is not what users prefer. And at the end of the day, the user rules!

This does not mean that we should kill creativity at all. The other way around. Let's propose creative ideas and show that those ideas, well implemented, improve conversion. At first it's a bit difficult, but like everything, with training, our concept of design changes completely...

…for the better, obviously.

Thanks for reading!

@joseafernandez

joseafernandez.net

I can understand that it may seem unpleasant, but if you learn to enjoy bad reviews, you are on your way to becoming a great professional.

Well, how can we do it?

1: Build a good predisposition

Be honest with yourself, understand that your work can be improved and open your arms to those improvements.

2: Cultivate the humility.

That is essential to learn and progress. If you think you are the smartest person in the world, you are wrong. There is always someone smarter than you. In fact, if you think like that, there will be hundreds of thousands of people smarter than you.

3: Discover your limits.

Be honest with your limitations and be clear that thanks to enjoying the bad reviews and learning from them, one day you may not have limits. Deep in our hearts, we all know what we do right and what we do wrong. It is key to understand this.

4: Expose your work to the great public.

Set-up feedback and survey tools, monitor user sessions and be brave. Repeat yourself that bad reviews will make you stronger. Because it is 100% true.

5: Separate constructive and destructive critics.

And learn from the first ones. Don't waste a single second or energy on second ones because it's not worth it.

6: Listen, learn and design.

Analyse the bad reviews and do act consequently. When you've solved them, you won’t receive those reviews again.

7: Celebrate every bad constructive review like you've won a trophy.

Because be sure you've won one and that will help you to enjoy this priceless process.

So, don’t be hard to yourself. All of us make mistakes on a daily basis. No one is perfect nowadays, but if you embrace your mistakes, you will not make them anymore. So you will be closer to perfection.

By the way, I strongly recommend to practice meditation. Meditation always help. Not only with this exercise, but in all aspects of your life. A great app for beginners is headspace. Meditation, like everything in life, can be trained. You only need 10 minutes/day to improve, so be patient and work on it. It will change your life.

Listen to your users and have fun!

Thanks for reading!

@joseafernandez

joseafernandez.net

Hey there!

It is a great pleasure for me to start my blog, where I want to share the experiences in creativity, art and technology that I have been accumulating over the last decade, along with the new trends of the present and the future.

I hope my posts are useful, interesting, instructive, and possibly in some cases, funny.

If you want to know more about me and my work, click here to take a look to my portfolio.

Thank you so much for joining me on this journey!

Selur on sah ytivitaerc.

@joseafernandez

Building a website, an app or a functionality is not as simple as eating a chocolate and cream waffle. But it is an exciting and rewarding process for the entire product team.

Let me share the product roadmap I have learnt according to my own experience over the last 14 years when building tech products:

Step 1 - Scope and discussion:
Before starting any technical action, it is important to talk about the project, the bases and the requirements. At this point, we have to create an initial specification document related to the design.

Step 2 - MVP definition:
With all the above information, a minimally optimal version of the product can be defined to put it in the hands of users in the shortest possible time.

Step 3 - Wireframing production (UX):
The entire user experience is designed and the first interactive prototype is generated.

Step 4 - UX discussions with the design team, product managers, developers and real users:
This experience is discussed as much as possible with the agents involved, quickly and efficiently, and the conclusions are implemented to obtain a final UX agreed upon with real users.

Step 5 - Final design production (UI):
Having the definitive UX, the final design of the interface is prepared. Before doing so, it is important to have defined a style guide that is consistent with the company's branding guidelines.

Step 6 - UI discussions with real users:
As with the UX, before having the final design, it is important to discuss it with the agents involved. Two rounds of feedback at most is optimal.

Step 7 - Dev specs document:
Once we have the final UX/UI, we can enable the specifications for the developers, so that they can have the style guide, distances, measurements, sizes, fonts, colors, assets... along with the front-end and back-end functionalities required to bring the project into life.

Step 8 - Development:
The design team is always alert during the development phase to quickly support everything that is necessary, provide them with new requirements that may arise and help correct possible errors that have been overlooked in the previous phases.

Step 9 - First quality testing:
After the developers themselves, the participants in the UX/UI design have to do the first serious testing to ensure quality control of the MVP.

Step 10 - Drivers:
With few, but qualified users (5 maximum and getting their consent first), the pilots are launched, which is the real product working and errors and successes are analysed.

Step 11 - Release:
Product Launch.

Step 12 - Monitoring and A/B Tests:
On a daily basis, real user sessions are monitored to discover blockages, frustrations, bugs, UX or UI failures... and they are analysed weekly, taking actions to fix the detected issues.

Similarly, no UX or UI design should be implemented without having previously validated it through an experiment (A/B Test), which should last at least two weeks.

And that’s essentially, how we built our Guestapp at Chekin that has been used by more than 4 million guests.

If you like to take a look to my portfolio, click here.

Thanks for reading!

@joseafernandez

joseafernandez.net