Voice for game dev: how we developed the voice quest "Lovecraft World"
While Google was preparing a global update for the Russian-speaking Assistant - with new voices, blackjack and built-in payments, we decided to create our own game for him. We experimented with genres, plot, easter eggs, voice acting and in-game payments. The result was a wildly interesting experience that anyone who wants to take the first step into game dev through voice can repeat. Using the Love World quest as an example , we share our thoughts on games for voice assistants: how to develop and what are the opportunities for monetization.
In January 2019, Games & Trivia accounted for a quarter of all Alexa voice assistant skills. Amazon has long sensed that games are the engine of progress in the world of voice assistants, so a couple of years ago the company declared gaming skills a priority category and introduced payment for their development.
In the Google Assistant ecosystem, gaming actions are also in the top - they are only slightly inferior to educational ones and occupy the second position in the catalog. There are not so many actions in Russian yet, but there are quite a few games among them: quite popular: weâve already talked about âYes, my lordâ with an audience of 170 thousand people, and people are also playing âLizard Spockâ ââwith Google Assistant, â Guess the capital "or a test game called" Am I a Reptiloid? "
The vast majority of all gaming skills for voice assistants are various kinds of quizzes. We decided to recall the good old text quests. Their story began in the 70s with the game Colossal Cave Adventure, where according to the script it was necessary to search for treasures in a complex cave system only with the help of text commands and text navigation. No illustrations, videos, or presence effects â just the playerâs imagination.
With the development of graphical interfaces, old-school text quests began to be forgotten. But the appearance of conversational interfaces, voice assistants and smart speakers gives the genre a second life, because this is just a story about fantasy, voice and the ability to create your own interactive radio show. Even HBO recently released a voice quest for Alexa based on the Westworld series.
In the case of the voice assistant, the key tool for creating the game world is sound. Therefore, there was a feeling that the basis for our quest should be horror or thriller - such scenes can be very cool to reveal with the help of sound effects. But in which universe to place the player? To make his goosebumps, but without a zombie apocalypse and hunting for vampires. And then it turned out that our team is full of fans of Howard Lovecraft's work (by the way, this is an ideal variant also because the rights to Lovecraftâs stories are not needed, since they are under the public domain).
So among the actions for the Google Assistant appeared âLovecraft Worldâ - an interactive drama quest in the style of âCall of Cthulhuâ and another prose of the horror king of the 20th century: 180 audio scenes and more than three hours of audio content, professional actors, more than 50 storylines, voice control and in-game shopping ... But Cthulhu is in the details, and more about them.
âWhere does madness end and reality begin?â Lovecraft wondered. Let's just say that the madness ends after a brainstorming and creativity, and the reality begins while working on UX. Thinking through the user experience is incredibly important: how the playerâs first session will go, what will happen if he leaves the game and comes back again, how to get a hint if the player gets confused ... All these cases are relevant for regular games, but in this case the task is complicated, because that all navigation is done by voice.
The scenario of the interaction of the action with the user describes what phrases the player will presumably say, how the action will react to them, what state it will switch to and what to respond to. During the dialogue, the action should guide the player so that he easily understands what is now required to be said. For example, we corrected part of the wording when we noticed that players are more successful in moving along the plot if the remarks of the characters end with a question (rather than start with it) and prompt you to answer: âYou look unhealthy. Are you sure you want to continue driving? Replace you? ".
All instructions should also be as short as possible, clear and understandable. No one will listen to a three-minute speech on how to start a game or make a payment.
Tip . Questions should be asked to the player so that his answers are predictable. This applies not only to quests, but in general all voice skills. Open-ended questions are undesirable; this can lead to a deadlock between the player and the Assistant. And you can do it in another way. For example, in our quest there is a task for a player - to repeat a character spell consisting of only consonants after the character. If correctly repeated - one story line went, wrong - another.
When you design a voice game, sound is all that you have to transmit information. Unlike console, computer and mobile games, he plays a decisive role in skills for assistants.
Therefore, we paid much attention to the quality of sound design. The characters in our game speak the voices of professional dubbing actors from St. Petersburg, Moscow and even New York. For them it was a fascinating challenge: a voice quest is a kind of play, an audio performance, and the voice here is the main expressive means.
This is how the voices of the characters sound:
In addition to recording replicas, another challenge is noise design. These are not only atmospheric, suspense enhancing sounds, but also the sounds of actions of the player himself, other characters, as well as specific sound effects. It is necessary to carefully monitor the purity and "readability" of each sound in order to avoid slurred porridge. At the final stage, the voices need to be cleaned, the best doubles selected, processed, correlated with the events taking place in the plot and brought together with the prepared sounds in order to get a finished passage. In general, our sound engineer had a lot of work.
Attracting professional actors and announcers is cool, but there is a minus: if you want to change something in an already recorded piece, you have to go through the entire recording process from the beginning. Agree with the actor, record it, process the voice and so on. We consciously went for it, because speech synthesis technologies, although they are becoming more advanced, are not yet able to convey all the colors of intonations and feelings, make the listener fear, empathize, and smile.
Tip. You can do without actors. If the budget is small, then the voice of the assistant himself is quite suitable for voice acting. Moreover, this voice is not alone; there is plenty to choose from. Alice, for example, has six. Google Assistant has less, but itâs possible to insert your audio file into action. Although from the end of June in Alice you can also add your own sounds.
Until recently, Russian users and developers did not have the opportunity to experiment with transactions within voice skills. Since October last year, the necessary functionality appeared in the Google Assistant - digital products and subscriptions can now be purchased using voice. This feature has been available to the Russian audience since the end of March.
For games with a voice interface, the basic laws and principles of familiar game formats are fair. They relate to monetization. When we started working on Lovecraft World, we immediately wondered: what, in fact, will we sell? This point needs to be considered at the stage of script development. In voice games, just like in traditional games, you can offer the user subscriptions to update content or use certain functionality, purchase supplies (extra lives / attempts, coins, magic artifacts) or tips, for example, in quizzes.
To win, you need to find the right storyline and save the city from the forces of Evil. Frankly, none of our company completed the quest the first time - it will take several attempts. Three are available to the player for free. Each time he comes across a bad ending, the attempt burns out, and the player returns a few moves back.
If he has used up all his chances, we suggest purchasing one attempt or a âpackageâ of several. The conversion at this step is quite high - more than 14%. But, for example, in the popular Escape the Airplane game for Alexa, the conversion to payment reaches as much as 34%.
The player can make a purchase by voice - no need to leave the Assistantâs interface or perform any actions manually. This is how buying attempts to fight evil looks like:
Tip. The process of connecting transactions is simple, but it will require the creation of a development account in the Google Play Developer Console , where you will need to fill in an empty android application with BILLING permissions. Publishing is optional. In principle, only one such application is enough, where you can start all digital products for different actions. In addition, you will need to create a site, confirm its ownership in the Google Search Console and link to the created application.
To develop Lovecraft World, we used our own Just AI Conversational Platform, a professional enterprise-level solution. There are many branches in the plot that it would not be very convenient to manage in the constructor for creating bots (for example, in our Aimylogic).
The block diagram of the game "Lovecraft World" looks like this:
We tried to catch the official release of the updated Google Assistant, therefore, to reduce the development time, several people worked on the game at the same time. The platform allows this, it has the ability to use a version control system (Git). In Aimylogic, a group of people can work on one script only if the content is separated from the script (for example, it is stored in a database, in a file in cloud storage, or in a table that the script can access through the API).
If you do not have experience or do not want to go into the syntax of the platform, use the bot constructors. For example, recently the Alice Prize (the prize that Yandex gives for the most popular and interesting skills) for the second time received the Citadel created in Aimylogic: a rather complex game in structure, with parameters and even a âbackpackâ of artifacts that can be called up by voice.
This is what the script looks like in Just AI Conversational Platform:
When the game is ready, the question arises: what to do next? Everything is the same as with traditional games. You can constantly refine the action, improve the script, work off the feedback, follow the metrics (returns, the number of newcomers, the funnel of the first session and its depth, etc.), you can add new chips. If three people looked into your game, then first of all you should think about attracting an audience, and not about new twists in the plot and paid artifacts. And if the game already has thousands of fans, it is important to work not only on attracting new, but also on retaining old players - this is impossible without the development of the game.
As for the Lovecraft World, while we are experimenting with the possibilities of attracting traffic and analyzing user behavior. And also we are engaged in localization for an English-speaking audience. Over time, we plan to turn our first quest into a whole line of games assembled in one action - with general navigation and packages of tips that can be used in any quest.
The market for voice games is only emerging, there are still many questions. But this is really a promising, interesting direction and a real challenge for the developer.
In January 2019, Games & Trivia accounted for a quarter of all Alexa voice assistant skills. Amazon has long sensed that games are the engine of progress in the world of voice assistants, so a couple of years ago the company declared gaming skills a priority category and introduced payment for their development.
In the Google Assistant ecosystem, gaming actions are also in the top - they are only slightly inferior to educational ones and occupy the second position in the catalog. There are not so many actions in Russian yet, but there are quite a few games among them: quite popular: weâve already talked about âYes, my lordâ with an audience of 170 thousand people, and people are also playing âLizard Spockâ ââwith Google Assistant, â Guess the capital "or a test game called" Am I a Reptiloid? "
Finding an idea for a voice game
The vast majority of all gaming skills for voice assistants are various kinds of quizzes. We decided to recall the good old text quests. Their story began in the 70s with the game Colossal Cave Adventure, where according to the script it was necessary to search for treasures in a complex cave system only with the help of text commands and text navigation. No illustrations, videos, or presence effects â just the playerâs imagination.
With the development of graphical interfaces, old-school text quests began to be forgotten. But the appearance of conversational interfaces, voice assistants and smart speakers gives the genre a second life, because this is just a story about fantasy, voice and the ability to create your own interactive radio show. Even HBO recently released a voice quest for Alexa based on the Westworld series.
In the case of the voice assistant, the key tool for creating the game world is sound. Therefore, there was a feeling that the basis for our quest should be horror or thriller - such scenes can be very cool to reveal with the help of sound effects. But in which universe to place the player? To make his goosebumps, but without a zombie apocalypse and hunting for vampires. And then it turned out that our team is full of fans of Howard Lovecraft's work (by the way, this is an ideal variant also because the rights to Lovecraftâs stories are not needed, since they are under the public domain).
So among the actions for the Google Assistant appeared âLovecraft Worldâ - an interactive drama quest in the style of âCall of Cthulhuâ and another prose of the horror king of the 20th century: 180 audio scenes and more than three hours of audio content, professional actors, more than 50 storylines, voice control and in-game shopping ... But Cthulhu is in the details, and more about them.
Quest script is about UX
âWhere does madness end and reality begin?â Lovecraft wondered. Let's just say that the madness ends after a brainstorming and creativity, and the reality begins while working on UX. Thinking through the user experience is incredibly important: how the playerâs first session will go, what will happen if he leaves the game and comes back again, how to get a hint if the player gets confused ... All these cases are relevant for regular games, but in this case the task is complicated, because that all navigation is done by voice.
The scenario of the interaction of the action with the user describes what phrases the player will presumably say, how the action will react to them, what state it will switch to and what to respond to. During the dialogue, the action should guide the player so that he easily understands what is now required to be said. For example, we corrected part of the wording when we noticed that players are more successful in moving along the plot if the remarks of the characters end with a question (rather than start with it) and prompt you to answer: âYou look unhealthy. Are you sure you want to continue driving? Replace you? ".
All instructions should also be as short as possible, clear and understandable. No one will listen to a three-minute speech on how to start a game or make a payment.
Tip . Questions should be asked to the player so that his answers are predictable. This applies not only to quests, but in general all voice skills. Open-ended questions are undesirable; this can lead to a deadlock between the player and the Assistant. And you can do it in another way. For example, in our quest there is a task for a player - to repeat a character spell consisting of only consonants after the character. If correctly repeated - one story line went, wrong - another.
Sounds and voices
When you design a voice game, sound is all that you have to transmit information. Unlike console, computer and mobile games, he plays a decisive role in skills for assistants.
Therefore, we paid much attention to the quality of sound design. The characters in our game speak the voices of professional dubbing actors from St. Petersburg, Moscow and even New York. For them it was a fascinating challenge: a voice quest is a kind of play, an audio performance, and the voice here is the main expressive means.
This is how the voices of the characters sound:
In addition to recording replicas, another challenge is noise design. These are not only atmospheric, suspense enhancing sounds, but also the sounds of actions of the player himself, other characters, as well as specific sound effects. It is necessary to carefully monitor the purity and "readability" of each sound in order to avoid slurred porridge. At the final stage, the voices need to be cleaned, the best doubles selected, processed, correlated with the events taking place in the plot and brought together with the prepared sounds in order to get a finished passage. In general, our sound engineer had a lot of work.
Attracting professional actors and announcers is cool, but there is a minus: if you want to change something in an already recorded piece, you have to go through the entire recording process from the beginning. Agree with the actor, record it, process the voice and so on. We consciously went for it, because speech synthesis technologies, although they are becoming more advanced, are not yet able to convey all the colors of intonations and feelings, make the listener fear, empathize, and smile.
In addition to recording his character (and there are 16 characters in Lovecraft's World), each actor had the task of recording cues from the crowd - a street crowd, according to the script, aggressively opposed to the player. The text was harsh, rude, with insults and obscene language. When the script was written, it didnât occur to us that it would be awkward to ask honored artists to scream obscenities. And we already wanted to abandon this idea. But it turned out that the actors themselves were not averse to "discharge" after an intense recording. So the extras material turned out to be very emotional and believable.
Tip. You can do without actors. If the budget is small, then the voice of the assistant himself is quite suitable for voice acting. Moreover, this voice is not alone; there is plenty to choose from. Alice, for example, has six. Google Assistant has less, but itâs possible to insert your audio file into action. Although from the end of June in Alice you can also add your own sounds.
In-app monetization
Until recently, Russian users and developers did not have the opportunity to experiment with transactions within voice skills. Since October last year, the necessary functionality appeared in the Google Assistant - digital products and subscriptions can now be purchased using voice. This feature has been available to the Russian audience since the end of March.
For games with a voice interface, the basic laws and principles of familiar game formats are fair. They relate to monetization. When we started working on Lovecraft World, we immediately wondered: what, in fact, will we sell? This point needs to be considered at the stage of script development. In voice games, just like in traditional games, you can offer the user subscriptions to update content or use certain functionality, purchase supplies (extra lives / attempts, coins, magic artifacts) or tips, for example, in quizzes.
According to the plot of our quest, the heir to the mysterious estate (that is, the player) comes to a small town to enter into the inheritance rights, meets with various characters and answers their questions. The further development of events depends on the answers. One wrong answer is a loss: a hero can go crazy or get a bullet in his head, die in a car accident, be beaten to death by local residents, or even sacrificed to the Great. There are more than 30 such finals in the quest.
To win, you need to find the right storyline and save the city from the forces of Evil. Frankly, none of our company completed the quest the first time - it will take several attempts. Three are available to the player for free. Each time he comes across a bad ending, the attempt burns out, and the player returns a few moves back.
If he has used up all his chances, we suggest purchasing one attempt or a âpackageâ of several. The conversion at this step is quite high - more than 14%. But, for example, in the popular Escape the Airplane game for Alexa, the conversion to payment reaches as much as 34%.
The player can make a purchase by voice - no need to leave the Assistantâs interface or perform any actions manually. This is how buying attempts to fight evil looks like:
Tip. The process of connecting transactions is simple, but it will require the creation of a development account in the Google Play Developer Console , where you will need to fill in an empty android application with BILLING permissions. Publishing is optional. In principle, only one such application is enough, where you can start all digital products for different actions. In addition, you will need to create a site, confirm its ownership in the Google Search Console and link to the created application.
Useful numbers
- 2-3 days action moderation without transactions lasts.
- 4-6 weeks may take moderation of the action with built-in payments.
- The audio file used in the script should last no more than 120 seconds. If it is longer, the player will start, and this is inconvenient.
- 70% of the developerâs revenue - this is the action monetization scheme for Google Assistant.
Game Creation Software
To develop Lovecraft World, we used our own Just AI Conversational Platform, a professional enterprise-level solution. There are many branches in the plot that it would not be very convenient to manage in the constructor for creating bots (for example, in our Aimylogic).
The block diagram of the game "Lovecraft World" looks like this:
We tried to catch the official release of the updated Google Assistant, therefore, to reduce the development time, several people worked on the game at the same time. The platform allows this, it has the ability to use a version control system (Git). In Aimylogic, a group of people can work on one script only if the content is separated from the script (for example, it is stored in a database, in a file in cloud storage, or in a table that the script can access through the API).
If you do not have experience or do not want to go into the syntax of the platform, use the bot constructors. For example, recently the Alice Prize (the prize that Yandex gives for the most popular and interesting skills) for the second time received the Citadel created in Aimylogic: a rather complex game in structure, with parameters and even a âbackpackâ of artifacts that can be called up by voice.
This is what the script looks like in Just AI Conversational Platform:
Rolled out the game, what's next?
When the game is ready, the question arises: what to do next? Everything is the same as with traditional games. You can constantly refine the action, improve the script, work off the feedback, follow the metrics (returns, the number of newcomers, the funnel of the first session and its depth, etc.), you can add new chips. If three people looked into your game, then first of all you should think about attracting an audience, and not about new twists in the plot and paid artifacts. And if the game already has thousands of fans, it is important to work not only on attracting new, but also on retaining old players - this is impossible without the development of the game.
As for the Lovecraft World, while we are experimenting with the possibilities of attracting traffic and analyzing user behavior. And also we are engaged in localization for an English-speaking audience. Over time, we plan to turn our first quest into a whole line of games assembled in one action - with general navigation and packages of tips that can be used in any quest.
The market for voice games is only emerging, there are still many questions. But this is really a promising, interesting direction and a real challenge for the developer.
All Articles