 So very good evening ladies and gentlemen and a very warm welcome to exchange for media presents webinar. This e-webinar is powered by Miling, ladies and gentlemen, Miling Foundry and I would like to welcome all of you this afternoon for this very interesting and the need of the hour panel discussion. Now before we go on to introduce you to the panelists and our session chair, I'd like to remind all of you that we are live on our Facebook page and other social media platforms as well. We encourage you to ask questions. So please send in your questions to us so that we can address them during or in the end of the discussion. Now you can use our hashtag which is E4M webinar, E numerical four webinar and also you can tweet to us using our tag at E again numerical four M tweets. So it is E4M tweets where you can send us your tweets today. Now talking about today's session ladies and gentlemen, it is a topic that we all would definitely you know would like to know about because we are using the OTT platform so much during this quarantine time. Now how is it going to transform itself and what are the new technology or avenues that it is going to see and that is exactly what our panelists and our session chair is going to discuss today. So let me now quickly tell you what our topic of this discussion is. It's transforming viewer experience on OTT using artificial intelligence and I'd like to introduce and welcome our panelists on this webinar. Ladies and gentlemen, please join me in welcoming Dr. Gopichan Katragada, founder and CEO, Mylin Foundry. Very warm welcome, sir. Thank you. Thank you. I'd also like to welcome our panelist, Lokesh Chauhan, Chief Technology Officer, Eros now. Warm welcome to you, sir. Let us now welcome from UK, Mike Card, the former Chief Science Officer for British Telecom and also served as the President of the Institution of Engineering Technology. Mike has several patents to his name in the field of video compressing. Warm welcome to you to Mike. Hi. Thank you for joining us. I'd also like to welcome our panelist, Manish Verma, head of technology, Sony Live. Warm welcome, sir. And our panelist, Shahabuddin Sheikh, Chief Technology Officer, Alt Balaji. Warm welcome to you, sir. Ladies and gentlemen, I'd now like to welcome our session chair for this particular panel discussion, Mr. Prashant Rao, partner of Telecom Media and Technology Deloitte. A very warm welcome to you, sir. I would like you to now take the discussion forward. Thank you. Thanks, Keltie, for the warm introduction and bringing all together on this platform. In fact, I'm very happy to talk about this very interesting topic about who to meet at this point of time and then get to hear from all the learned gentlemen on the call about what are the things that's in store for us in the future. As we move more along this, the mood being COVID at this point of time, we saw Prime Minister Modi coming online and saying, this is God. That is God. Lockdown. You need to stay indoors. You cannot go to your offices. Just imagine, amongst all of this, if he said no OTT, no internet, what would life have been? How much of our things that we're doing at this point of time would have changed? So with that light note, just moving on to this topic on how technology can influence the viewer experience in the OTT sector. See, just broadly, if you look into it, there is a 87% tele-density in India in terms of penetration of telecommunications. Then as you move along and compensate with it, what do you say, about 1.1 billion customers who are having SIM cards, of which 400 of them being smartphones. And then now more and more houses are getting smart televisions. There are devices which can make your down terminal smart televisions as well, like a small streaming stick. So the multiple of these things are giving a push towards OTT's success. So as we see right now, there may be about 35 to 40 large OTT companies which are currently operating in India, catering to the demands of the customer. There may be about give and take 50 million odd customers who are currently looking into it. That means we are at the start of a big revolution and a change. And then what we need to achieve is the entire population of India going towards this. So about three and a half hours of consumption is what we see on a daily basis. So what's one other unique factor in this scenario is the OTT consumption also leaves behind the fray. You know who's consuming, what time it's being consumed, what is being consumed and various things about you, your gender, your, what do you say, demographic, etc. So how is all of this being used to give you better experience? So let's dwell into this and get to some of our speakers to understand what's happening in this area and what's being done. So quickly to jump onto this, maybe I'll just throw open a very broad question to Mr. Gopi to say, how can maybe user experience help OTT's to make them stand apart from the rest and then attract more and more customers towards them? Maybe Gopi So Prashan, thanks for the question. And if it's okay with you, I'll take it a little broader before diving deeper. So it's a fantastic juncture that we are at. And it's a great time to be a person who is interested in creativity, who is interested in technology, because they have come together like never before. Creativity has been democratized in many ways. And technology is making a lot of things happen. So today's conversation I'm looking forward to hearing from everyone in terms of how they're viewing this intersection, because there's so much happening. So the way I would like to share my thoughts is that there is a journey of a story in many of these OTT broadcasts. And that story could be video and demand. It could be a television serial or a movie. It could be sports, which is also a story. It could be news. It could be gaming. These are all things which are fair game when it comes to viewer experience. There is a journey of a user, the user and his journey between content as well as within the content. So we'll come back to each one of that. And then there is a journey of a pixel or an audio byte from where it is created all the way to where it is consumed. So each one of these are impacted and impacted in a favorable way in terms of how content is being consumed and being enjoyed. So let's come back to the journey of the story and the creative agile multi-gen needs which are now being expressed. So what is happening here? So if you look at AI, which is one of the focus areas for us today, it is looking at the ability to do plot generation. Previously, if you want to create a story, you had to do it entirely as a creative output of an individual. Even today, I believe that an individual cannot be removed from a creative activity. However, he can be made more creative and be given lot more inputs in his journey as a creative person. So there are branching narratives which are possible today. You have watched that in a few of the global channels and it was a very much a starting point. But AI can make these branching narratives much more meaningful. Beyond plots, you can have story generators. You can have short recommendations for a particular scene. You can have an action being followed in case of sports. You can have a ball being followed in case of cricket or soccer. You can create news by compiling news from various things which are happening around. So these are on one side. On the content as a media itself, you can democratize VFX. So you have seen user-generated content in various methods where really what's happening is the VFX which was only available to large media homes is available to everyone on their mobile phones using AI. You can modernize content. Content every three years is being outpaced by the display capabilities. So and you can generate content as well. You can generate trailers based on a large volume of a movie for example or otherwise put together a brief of news and so on so forth. Now the journey of the user, the need there is precise timely interactive content. And this is exactly what is made possible by AI. So recommendation engines make precise content available at the right time. You can have video X-rays which are AI based which are not manually tagged. So you can actually have an ability to look at what is happening on this screen who is there but also whose voice is playing the music in the background. But also what are they wearing? What are they carrying? Can I buy it somewhere? So many things are possible in terms of the journey of the user. And then in terms of the journey of the pixel, the story is about efficiency and quality. So buffering today you got buffered on my screen just as you are speaking. Maybe not the same for everyone. You just got stuck and your voice disappeared. It doesn't have to happen because today I can transmit a very small size file at lightning speed and only use the edge device to get the right quality. What is the quality resolution? Zero buffering, dynamic range, removing noise, removing artifacts. And AI is a tool and it should be looked upon as a tool. And it does hence one amongst the set of tools which should be looked at. At a later point I'm happy to discuss as to how AI will do all of this but I just wanted to give you that flavor of the journey of a story, journey of a user and journey of the pixel. All of which are impacted by technology in this beautiful way. Thanks, thanks Gopi for wonderfully putting across all three journeys on that side. So as people go through this journey, as various experiences are taken through this, obviously there are certain challenges that come across. One such similar challenge that we noticed recently was the massive load on bandwidth that all of the concurrent users started using it and then that put in quite a bit of load on our telecom networks. So there was a direction that was given by our accordingly changed their standards and then accordingly catered to the user environment based on access where is. So maybe it's right time to take a question on what kind of challenges our OPD work captain goes through and then what are the some of the things that we are going to overcome them. So maybe Mike can you take a stab at this challenges part of it. Yeah certainly I always look at things from my sort of previous expertise which is around image coding and image compression. I work very actively from the from the late 70s through to 2000s on image compression algorithms and it's really interesting at the time we started started doing the work the the problem was always challenged in the sense that everybody expected large bandwidths to be available. So people working on fiber optics and we were working on image compression and people said that you know was it really an interesting research activity because image compression isn't going to be needed because we've got so much bandwidth and actually if you actually look what's happened over the past 20 30 years we're still being challenged by image compression and the difficulties of doing that because the sheer volume of traffic which is on which is on the internet and the load that it creates something like a third of all of the load on the internet is internet streaming and so the costs of consequence of that are relatively high and of course there's been a whole series of coding algorithms that have been created over the time you know I work very actively on H261 in the 90s that became the basis for MPEG 2 and you know MPEG 2 could achieve what an HD picture in interlaced form 1080i probably around 40 megabits per second at that time and that was considered to be a real achievement and over a five year ten year period as the compression algorithms have got better in terms of the implementation of MPEG 2, MPEG 2 itself didn't change but the compute power available in order to do the compression increased and you know we can now using MPEG 2 probably get the same picture down something like 15 megabits per second something like that and then of course everybody came up with H264 which people know and that gave another step change from something like 15 megabits for the same picture down to something like eight or nine megabits and now we've got you know HEVCH265 which could probably fit the same picture down 1080i picture down something like five megabits per second but it is bottoming out so you use it using using that technology and one of the things that isn't talked about very much is the difference between real time content like watching sports or interacting with people and movies movies when you look at how the coding is done the the coder will typically code the picture once look at the places that didn't code very well recode them using a slightly different set of parameters in the coding scheme until they get an optimum quality across the whole of the content and as long as the average file size for a movies around two gigabits per second they don't worry because they download it and on real time you buffer for a few minutes and then you watch the movie and it all looks in good quality and the average bit rate maybe five or ten megabits per second over the channel but if you take real time content you can't do that you can't do two things you can't you can't recode it because you haven't got time you've got to get it you've got to get into the network immediately because it's a real time piece of content and you and you certainly can't buffer so you can't code it twice and you can't buffer so this means that in order to improve the quality of that content the only thing you can think about is either improving the encoder and a lot of work's gone into that over the years both in pre-filtering as well as in in the process itself but in post-filtering looking looking at ways to enhance the image that results either removing the the coding noise or increasing the resolution of it and certainly that links back to the AI issue which is if you can really learn what a good picture looks like and apply that to adaptive filtering at the output then you can improve the experience for the user. I think that was a long answer to a short question but I hope that was helpful. There's a lot of insight on that thanks thanks a lot for that. Moving on to another aspect of the experience as we know the amount of content which is getting created on OTT is enormous there are about 1600 plus hours of only would you say exclusive content which is created on this particular OTT platforms so if you look into this it is very very difficult to stand out and then give a customer the content that he wants so in this regard recommendation I understand plays a lot of role in terms of the content discovery so maybe Shahid if you can throw some light on what kind of insights and backing the processing and analytics goes into to recommend a particular episode to a particular user and then what's in store for a future. Definitely so currently see I will just give a context currently where we are in right in this corona COVID-19 as you call we have a lot of content thousands of hours of minutes of content we have available but the new content production has stopped now when a user comes onto a platform if we can identify the user or the universe around it you know what content he hasn't watched you know in terms of recommendations or users habits of watching you know preferences basis of that you can display him a content which is readily available with you because fresh is something with the user hasn't watched it's not necessary you produce a new content and then make it available for the user so that is where AI plays a major role where you can identify this of the user habit and preferences and then display that piece of content to the user that is one side to it coming in terms of analytics as of now like we are trying to get more and more traffic on our platforms and basis of you know different you know key metrics we are trying to identify the user patterns and then trying to promote different content pieces to them again but more or less it's more around the user habits like you know you have to identify the user habits and their preferences and then accordingly program your content. So so if I understand you right what you're saying is if a user comes onto a platform at a particular point of time maybe what is thrown at him is different then maybe when he comes at a different point of time because maybe at a morning he may want to just catch a glimpse of something evening when it's leisure time he might want to catch up on his series based on his earlier watching preference he's watched in romance so you're trying to give him another recommendation on the romance side so that's how it goes. Right and that's again from your existing catalog you know because your production is not there in place so basis of his watching habit and preferences according to the network availability like currently the network is quite congested and I think most of us have known because of the you know guidelines have primed down our you know sd content into sd. Right right so so also I understand on the same point to improve click through rates per se not only you choose which particular content should be displayed to you but also what should be the hero image that is shown across so maybe particular series may have a theme which is depicting horror a theme which is depicting romance a theme which is depicting something else based on my interest level that particular a theme hero photograph comes in to ensure that I click through on it. Exactly basically you know normally it's nowadays to you know get the first glimpse you have to have the key characters of the specific episode or show or the theme around which that episode has been built upon has to be the hero image because that is the only you know and nowadays if you have an app and most of us have that you know our corousal on the top 30% of the inventory or the real estate you have on your device that is where you display that hero image and try to attract the users and get a click through. Right right thanks thanks for that so moving on to another important aspect. Prashant if I can just add if you don't mind a little bit of interactivity since you also asked about the future I am in talks with the various researchers in the area so just wanted to give you some thoughts one is of course the individual and what he or she watches but also there is a network science element to it as to what is the peer group watching because there's a lot of content consumed based on what can be discussed and there is a viralness to the content being able to predict the viralness of a content is also another aspect but I would also point out that there are folks very early on who have taken out intellectual property on connecting biosignals to recommending content so it is not as complicated today because if you have a heart rate variability which is on most variables you can link that to what kind of content to recommend or if it is a branching content how the branch should work because I might not want to spend the time figuring out each branch so there are lots of fun stuff happening it won't come to light for another three four years but it's good to know is social media also used in the same side say for example a particular person is an influencer so maybe what is shown to him what he talks about that means that okay people in this region are now or his associates are more inclined to watch something like this it is something being done on that side as well so at at Yale we had worked with Professor Krista who does super work on network science and he was not using it for OTT kind of content as much as determining how to get a particular initiative more popular let's say but the same approach works for OTT because they use social media and social networks and an easy way to determine who is the influencer use the right word and hence what is if you know who is the influencer then you can use that as a proxy for determining recommendations to a bunch of people and then at this point also what I wanted to say is bringing in what Mr. Mike said sometime back recommendation may not only be on the side of the content to be watched it can be even whether you need to go for a plan which is a HD plan or an ST plan I had a very pleasant experience that I wanted to share on this this thing so I bought a new TV which is a HD TV and plugged in and put on a particular streaming app on it so immediately as soon as I put on and signed in it said you are on a HD plan a ST plan which is ABCD so why don't you press okay on your this thing to upgrade to a HD plan one month complimentary and I did that and it's been say two years since I've been on the HD plan that is a wonderful experience I would say and an example of how to do a upsell at a very apt point of time yeah they seem to know your bank account or the device I use yeah yeah that's a good proxy for your bank account so on the same point maybe I would like to open up another question on subscription because in OTT what's very important is subscription versus ad etc so the bunch of analytics being done to upsell cross-sell to customers move up subscription and stuff so maybe location money if you want to take this question on subscription and maybe both of you add add on these points on subscription and what can be done to what's being done to move it up sure Prashant so before getting into the subscription I would like to just take up the recommendation piece and I thought that that's really relevant for all the OTTs so yes recommendation is something which is very important for the engagement but I feel that we should take a backseat and see that how we can really create that recommendation for the user and I feel that AI which kicks in and right from the processing level of the content because it all depends on how enriched my metadata is enriched then definitely I'll be able to give a better recommendation because it all comes out of the metadata you are having in the system when we talk about the metadata there is a certain limitation in terms of how much you can put in manually right now I take the example of let's say a comedy now comedy also has the multiple variant stand-up comedy family comedy kids comedy adult comedy now if I'm not putting my metadata correctly I may start putting only comedy then the recommendation which is coming up to the consumer may not be correct similarly if I have to start putting the markers for showing the recommendation or even for that matter search if I want to if I'm not able to put the markers at the right time by using the right system because putting the markers manually may not be possible I need to use a AI system which can put the marker based on the character based on the emotion based on various key moments in the in the in the sequence right so I should be able to pre-process the content using a sort of either manual and the combination of manual and AI systems which are able to put either the markers or the metadata efficiently at the time of pre-processing which I can then use for either recommendation and on top of that I would say even more important which is becoming now is the personalization of the content or the experience for you because it is all about experience right and given that that the consumption is gradually moving towards the prime time consumption on the bigger screen from a smaller me moment on the mobile devices which is still the the prime which is the main consumption for the consumers on on the mobile which is wherein I am interacting one on one with the content the moment it moves to a big screen time time consumption is happening on the bigger screen you need to create the experience based on the profiles right now you can't just watch a content on a bigger screen having a four or five of your family members sitting so you then start creating the personal experiences by having those profiles I think these are all emulated using the metadata enrichment of the metadata based on the metadata then I'll be able to either create the experience or recommendation or the personalization so that is what I think we have been working on and we are trying to create a lot of personalization for the consumers inside the application besides of course the recommendation and in the same breath I would take up your your subscription also so we offer both avod and swad content right while avod is good which is primarily our catch-up content and some of the premium content also is available in the aim but but a lot of content is in the in the premium segment originals have original acquired content and some of the other premium all the sponsors behind the paywall right so we need to first ensure that while the user is most of the users when they are onboarding they onboard by consuming the free content available free content either it's a full catalog full choice fee or part of it is free for sampling or whatever how and when I should be asking the user for the subscription that is something which is very key right it's not that the moment you land on to the platform I start bombing you saying hey you subscribe you subscribe you subscribe that's a sort of a very uh not a very good consumer experience so we need to profile the consumers and ensure that I am I'm asking him to subscribe at the right content at the right time when he's trying for and engaged enough on to my platform and once he's engaged how I upsell because my content is having multiple types of content how when I should say that hey you know you have been paying on a monthly basis month on month now it is time for you to upgrade to maybe a six months or a 12 months back because you are going to save I we saw that you are consistently consistently paying and this brings back to your experience of SD and HD if a user sees that I am paying X amount every month on month and I'll end up paying 12x if I upgrade to a annual plan probably I'll end up saving 50 percent of the cost that's brings in the brings a wow factor wherein consumers feels that this product is for me takes care of of me and it really values what I am spending on to the platform and the third aspect is how do I ensure that users are not churning out when they are on my platform so when I need to kind of profile whether the consumption pattern is there for example if I find that you are not consuming something it could be because the content is not available or it could be because of the quality of service also if I am able to profile and understand that what sort of a quality of service I am able to offer it could be the last mile issue it could be ISP specific issue it could be device specific issue but if I am able to capture the quality of service data and recommend or suggest a user hey you know you look like you are on a lower bandwidth you may want to shift to alternate connection to have a better consumer experience or whatever users are definitely going to engage and kind of consume more so subscription and that is when once they engage with your platform they consume more they consume more they get a satisfaction they then likely to pay on continue to pay on your platform so I think there is a complete funnel which we need to manage and at each and every level we need to make sure that we are really capturing the right kind of data using that data to use it in a way which really brings into a better experience and convert the behavior into habit essentially when the user is landing there is a behavior change which is happening so I am consuming either data on a smaller device moving to a bigger screen or I am consuming it on broadcast moving it to the digital now this behavior should convert into habit by bringing in the better consumer experience and the consumption behavior onto the platform. There is a lot of info, useful insight there. Rukish you want to add something onto it? Sure thanks Manish sir, you have covered most part of it leaving me very less to contribute to but the way I see it this is a mix of what you were getting from earlier experience of using family you know this television show so all of you and your family will sit down and watch the show because it was entertaining to everybody the angle of personalization which comes into picture with OTTs now we can selectively identifying what you consume on large screen versus small screen where you consume what you would say family friendly content versus where you are open to more edgy stuff so you can identify based on location based on devices identify the time of the day day of the week also these metrics available to us so these intelligence recommendation system pieces that we talk about they are all evolving with these newer signals we are evolving on part of how we do process any specific user data so the mood of the day and all those factors come into picture the recent events come into picture like Rishi Kapoor died we have a catalog of movies which belong to his golden era it was rather easy to surface it any of those signals that we get okay we do demography we understand which other perspective users for which we surface this content and we'll get you know a larger click through rate or a larger playback so that engagement metric that money sir was talking about just now that is the key factor for any subscription business if you don't have enough engagement if you are not able to add newer content if you're not able to resurface all the content if you're not able to identify the taste of the user based upon when it makes sense to him or her then we will not be able to engage him or her normally that engagement piece is a constant drive you use social as you talked about earlier to get the influencers to talk about your content produce high quality show then you can market it better all of this goes as a journey so media is all about telling you know a story we are in the business of being storytellers how best we can tell a story is by creating a great show surfacing it to write audience and then ensuring the last my delivery is super the user should not feel that he was short charged this was not worth his money and he or she is facing any issues the RCA should be quick thorough and we should be able to identifiably you know bridge that gap in that experience what Manisha was talking about is churn specifically how churn actually happens if you're giving somebody the wrong content he will not be engaged if you're giving somebody bad quality on part of his tv not being able to you know played nicely as his phone is because he's trying to do you know a full hd sort of viewability on his tv with a different isp versus his mobile phone he's doing an sd watch which is very easily available in being on a 4g network half a billion people right now in india are on 4g network that gives us a huge huge opportunity it parser passes the tata sky and until reach and that's where i see that this is a habit that is going to be much more consolidating if you are able to deliver the premise of you know this is worth your money so i here i see a lot of comments coming in on on the specific topic here where people are saying are indian audience ready to pay are will you be worried if a person is paying but still not watching it so will you go back and reach out to him and say hey you were watching not paying not watching but paying for it and kinds of things so maybe we can take some of this in terms of is the indian audience ready to pay golden question that that you are coming again and again to start off i would say that given the right content given the right price points and the amount of things which are available at say for example to consume oddity you need a multiple things right you need right content you need right device you need the right pipeline to send it to you in terms of a good bandwidth network if all of those are available in the right price point the right personalization experience why not is what might take this maybe or the panelists want to take a go at it i mean it's it's not breaking the bank i mean all of us are pretty pretty economical as part of buying a subscription for so definitely there's a bank for buck approach that all of us are pricing our backs with obvious case very certainly earlier talked about if you understand if user is unable to watch then how to validate that if it is our fault how to ensure that his money is not going to waste that is a very key factor and part of customer engagement and satisfaction that he is not being you know taken for a ride so the trust factor is very very key for any of us to be competitive in the business the moment people believe that it's not worth their money then people will not be bothered about you know next month's subscription and isn't our population our boon at when it comes to this so while someone else if he creates a particular content spending it's somewhat of budget recoup that there is a small base to charge and they need to charge a particular amount and we with our billion class population if everyone contributes a little bit the content creator in the platform all get their bank for money and there'll be lot more to share and then this particular content which is created can be reused again and again so that's that's one of the secrets and that that is the primary reason we are all very very aggressively believing that India is a growing market and the cost of creation versus the money that can be made over that particular asset that balancing act right now both seems a little off but in the longer run we believe that will come in picture you should be making money that's what we are doing here sorry Manish go ahead okay thanks Gopi so Prashant I see it actually twofold when we're talking about willingness to pay and the propensity to pay right now there are two parts to it one is the behavior and second is the whether I'm I have the propensity and capacity to pay what I feel that take the example of music right it's not that I can't pay for it but lot of people due to high piracy people are still side loading and it is there is a behavioral shift which started happening when people realize that it's better to pay for music because I can create my catalog I can create my playlist and I can consume it better so they started getting a better consumption and the experience when they started paying for the content and organizing their music so I think that behavioral shift is very important wherein users are instead of using the pirated content to start using the the the premium content and start willing to pay for it what we have seen that it's a journey itself right it's not going to happen overnight what has happened now it has expedited that journey wherein now users have started making the payment for the premium content it started and that's what we are seeking besides all the free content consumption whatever is happening people are actually paying and consuming the premium content also and I think we need to just continue with this journey and and create this behavior into habit wherein users are now paying for the content and consuming it of course what location mentioned about trust actor they need to have that trust when I am making the making the payment I am really going to consume that across the devices right it's not the services are going to be available for me always and I should be able to watch the content as per my convenience and again it comes back to the experience so people are going to pay for content and experience both if we expect users to pay for content only probably know they will continue to side load and and kind of consume the content and that is where again I am talking about coming back to AI and ML to know your content to know your consumer to engage him give him the experience which actually which actually give him not only the content but experience both across the various devices are going to be the key and that is where users will be willing to pay yeah so Prashant I just wanted to add that while you're right in terms of India is a large market and you know even if there is significantly right now AVOD consumption which is advertisement based even the percentage which is subscriber can be attractive however it gets segmented also in terms of content by vernacular and so while the population is large and there is a large group who can understand and look up to Hindi content and English content but there are other content also so there are there are challenges which can be solved because your vernacular content or you're making your other content available for people who only understand vernacular is also possible both from a transcription kind of services as well as various other means to make sure that the content which is produced is available also the remote location problem exists as soon as you go out of tier one city to tier two even a place like chikma gloar you can't even get your whatsapp messages leave alone play a video so that that also is a challenge in India which needs to be addressed in order for the subscription based to continue to increase better people than me in the audience I'm sure that they will have comments if they agree or disagree but Manish, Lokesh, Shahbuddin and I think you might have been experiencing these as well in terms of remote areas and vernacular. So a point on the payment thing before I just jump on to the other thing with regard to vernacular remote area and then how do we manage on that side on the payment side of it the other challenges that we are seeing is the digital payments and acceptance of digital payments by the consumers at all and then ability to pay through a credit card or the wallet wallets have come as a bone in that scenario and then also what we are seeing is a combination and then newer payment models are reviving and coming up in terms of operator based career based billing which is seen elsewhere in multiple other countries in India we are seeing that it started to happen also bundling of products say for example we've seen three four ODD products being bundled in and being offered together as one subscription to the user and then paid along with his other normal subscription that he's used to so these are some things which are also trying to convert users into the paid ecosystem as such maybe. So one other thing I'm just seeing here is on the point that Gopi mentioned as well a small remote area that you go in you may not get network as such and I'm also seeing one other question which is coming in how do we ensure that best in class video quality experience is achieved irrespective of the content that is being used or what kind of encoders that we use that's one part of it the other part is a lot of the small scale cities may not have the best in class network so do we have a different strategy for them so Mike maybe you can take both of these questions together and put some light on it. Well looking at the sort of the fundamentals of it the quality of the encoder is pretty well everything and essentially what the encoded stream is is a set of instructions on how to decode the picture so the decoder itself has very little control over the quality that's achieved it's on the encoding side within the bandwidth that's available there are some things that can be done though which is in the post-processing section of the decoder in terms of being able to enhance the image after it's been decoded so you're basically looking at technologies which adaptively filter the filter the picture to replace and improve where the encoding errors have occurred and indeed at the same stage possibly increase the resolution of the of the image but really it's up to the encoder in order to in order to make the choices in terms of what it can do in terms of the compression and you're left with the result of the decoder which becomes more and more complicated the worst the worst the image is but certainly AI technology at that stage is certainly beginning to need to help because you can apply very clever adaptive filtering which has learned what a good picture looks like and apply that post and that that makes you somewhat independent of the of the encoding quality but the ideal situation of course is that you've got you know the best quality code you have you can have for the for the bandwidth and then you can think about upscaling the image quality in resolution terms rather than just worrying about the coding artifacts so on the same line there is one more question might if you want to just take that as well can AI based scaling deionization can be performed streamlining without affecting the latency is a question that's coming from one of the viewers yes because the the filtering is per frame and is a spatial filtering per frame so we might argue technically whether it's a millisecond or less than a millisecond or something but effectively in in in reality compared to the end-to-end delay of the system you can do the improvement within a tiny space of time you wouldn't normally worry about if I can add to that so the challenge is doing it real time like Mike said and doing it at a minimum of 30 frames per second and with the abilities that we have today at the edge you can do it real time and it has been demonstrated so on another question which is coming up looks like someone is really missing going to the theaters they're asking do we see a trend going forward that there'll be 40t first release of movies happening and what kind of a thing that's in store for us so you want to take take a shot at or maybe location one of you sure I'll take this one it makes a lot of sense for situation like current where there is no way people can go to theaters and watch it but these are very early indicators on part of how does it transpires into what you'd say monetization strategies I mean movie production is much much expensive business than producing an original show the amount of money that goes into a you know high quality movie affiliate drama or you know something which is set up in some remote location as an action movie or something like that it costs a lot how do you basically monetize on basis of that is very difficult to answer there has been some great experiments like trolls which basically made enough buck but I am not very sure every movie that goes through a digital premiere route is going to basically have the same impact and it's actually very difficult to uh quantifiably measure the financial success of any of these properties directly uh there's a so for movie it's very simple it's it's essentially first two weeks where you basically make 95 or 97 percent of your overall money rest of it is pretty much very small incremental updates on week on week basis of business uh but that's money in the bank that's directly correlating to that particular movie for an OTT platform it would for a subscriber based OTT platform it's a very difficult task a t-watt definitely will make most sense in that scenario because then you can directly attribute and correlate okay this movie actually made this much money right so so a related question on that side now with a lot of analytics available like we understood there are two kinds of things that you look into 360 of a customer 360 of the content itself by uh breaking it up into multiple metadata like Manish was explaining uh by doing all of this is some kind of analytics being done for uh curating content as well is is the technology dictating what is to be procured what is to be uh produced etc so any any thoughts on that side well definitely definitely we understand we are getting much more insightful about what content works with what segment of people how we create bigger segments by overlapping multiple factors together would make uh a show or a movie which will have wider audience and this is this is now getting you know backed by technology but this is traditionally being the approach of creating content how do you create a family friendly show which is clean and is fun is because you understand this for uh you know a living room experience the entire family needs to sit down and enjoy it so you are basically creating wider audience space for yourself if you're creating a content that can be consumed with time right right very cool so in fact that can be one of the future revenue strings also I can hope that as you collect more and more and more into data of people watching and the viewership habits you can more and more predict what will sell and what will not sell so producer before going putting his money on the bank he'll come and ask one of you guys uh you know what whom should I take as the hero what kind of story should I take if I'm putting this many crores as my budget location will save me that happens now itself I mean what you're talking about is pretty much what happens today if you take a salman khan uh star which is action driven having really amazing action sequences I mean it's sure to get enough crowd uh in the theater just to add to this uh you see there are a lot of uh decision-making going when you might have seen there are multiple seasons of a content of a show is coming right now when to decide uh what to I mean what all shows are to be uh season based and when the season should be launched there are a lot of uh uh data which is supporting that right I should create the when it should be created and when it should be launched it should be based on really the facts on the data depending on what sort of a subscription churn and uh consumption is happening so that my timing is right if you are launching let's say you have hundred thousand users and they are consuming a type of show and most of them on an annual plan and you launched another season on let's say six months I mean they are going to consume in the next six months right whereas you should be launching it on 13th month or on 12th month so they are they are likely to renew the subscription because these are all the users who are engaged to that particular show so there is a lot of data which anyways in the OTT space and the traditional media also being used when the content is produced so if I understand right what I what I also heard from a few people is it's just not you became a paid subscriber from being a a word subscriber what route did you take did you click on a maybe an icon which is a sports property and come in and become a member or did you click on a series which is an english series or a desi kind of a series which is basically giving you a tag on what you are looking for and basically as soon as that is getting over you're looking at how do you replenish something in terms of content to keep this guy on that's right so another question that comes in maybe to the panelists here is how is audience measurement done in OTT and then how reliable is it and how are people using it maybe Shahgudin you want to take this sure basically there are like quite a few key factors like number of you know shows he has won video minutes he has consumed and type of genre he's consuming based on that you know we try to segment the audience and based on which we tried to then you know display a program content specific audience right so there is there is another question which is coming in here saying that are there any regulations for OTT in India and how is it affecting you in your day-to-day operations you want to take anyone of them want to take this question we have we have worked with other partners to create what you would say proactively more family friendly content we believe that this OTT premium experience is something that you're going to enjoy in your living rooms profanity nudity all these things are if presented towards the sense of art it makes a little sense but just for the sake of it it does not make sense to me and hence you know we have tried very proactively to be on the right side of things right add on to this while the government has not maybe come up and then said you know what you follow these these very strict guidelines to go in on the side the the federation of all the OTT players are coming together and putting in a self-governance kind of a law to say that before someone comes and says you know what you don't go out of your house any point of time for covid you just make a rule which says you know what I'll go out of my house only from 7 a.m. to 7 p.m. and then later on during the period I'll be staying inside my house so similar to that there is some bit of self-regulation which is happening on this side to make make things better so another question on this side of it is there is some bit of activity in this entire ecosystem which is manual and it takes a lot of time and effort and is error prone is there some kind of technology which are being used to make this better in terms of error free and use of lesser time per se maybe some examples on this side would help anyone who want to take this so let me go first Prashant so few things one of the areas where there's a lot of manual work is content creation as an example if it is animation work vfx work significant amount of manual effort is required and there is increasing amount of automation coming in in as we go along some of it is technologies other than AI but recently AI has played a role so today you will find many examples where you can take up one picture and use a substitute actor who is not well known but can provide facial movements and make that picture act so you can have a person who is no longer alive or you can have an animation or you can have an animal so these are areas where you are seeing a lot of automation coming into play this was just an example the other areas would be things like taking old content and making it let's say you're taking interlaced content of low resolution and you want to make it progressive and high resolution it used to be done manually a lot of color correction and various other tools being used today you can use AI definitely to go from a particular quality to a current quality just wanted to give you two examples to get it started just to add presenting what then I was talking about the cue point so there are two big use case which probably would be required is one is skip introduction so when you are in the binge watch you definitely would not like to get into the introduction again and again those two point tagging of the skipping of when to skip the introduction and similarly the skip credit right in the binge watch definitely would not like to see that credit which comes at the end of the show or the movie that is the one use case the second is the creation of the clips are especially in the sports wherein four and six and the key moments and the creation of the clips and posting it onto the various platform or even onto your platform these are all can be automated because these are all really required to be done in the real time you can't do it post match getting over and things like that so these are all some of the things which are happening to automated system which can really help and expedite the overall process which least amount of error because it's all based on some learning and once it is there you can kind of keep on doing it. Okay are subtitles to the what do you say the voice also one one example of something similar and maybe also grading user what do you say recommendation whether it's AI whether it's U rated or it's adult rated based on maybe a blood splash which is coming on the screen or a lot of abusive language which automatically AI picks up and recommends to a particular user or not recommend. See from a technology standpoint Prashant the ability to transcribe is very much there and the value is when you can do it with vernacular language but translate is a difficult thing because there is a cultural element to it so let's say that we want to translate Hindi to English that is very difficult because you can see even in your Google Translates or otherwise that you don't get the cultural context at all and in a movie where there are songs and there are you know gullies and there is local nuances you'll just not get it right so transcription is possible translation will take some more time to automate but even transcript so there are this will also help where instead of translating by listening to the video you have a transcription and somebody can be a good person can be used for translation without having to sit in front of the video and go through and that itself saves a lot of time. In fact right after this I saw a particular image yesterday it said dosa batter was translated by Google to say dosa ballebas. And as I'm saying even if you put it in a language like Hindi, Hindi has so many dialects some people speaking in western Uttar Pradesh would be very different from how we speak in Rajasthan it's as simple as that when you try to understand and transcribe it but I don't think anybody is able to do it 100% it's anyway you know around 80% accuracy you have to put a little bit of manual effort to identifiably say how to improve on that the models are improving but I do not believe 100% transcription or 100% accurate translation is happening anytime. Other example for you know manual labor now shifted to AI was identifying good key frames while you are trying to scrub through a video earlier used to basically do you know create screenshots and identify markers and put it while you are trying to see what is happening. Now AI has enabled that pretty much you don't have to do anything. Yeah I see one other question which may come up on time here is is the application of AI and ML only limited to recommendation in thumbnails there are a bunch of things I know maybe anyone want to summarize a bunch of things that we are doing through AI and ML and not just recommendation in thumbnails this context. I know that I've taken some time if anybody else wants to do it happy otherwise I'll happily do it. So maybe anyone want to give it a go or Ruby you can take it. Okay so we discussed it early on in the discussion today there is a lot AI can do but I want to caveat that AI is a tool and shouldn't be looked upon as silver bullet but it's a very important tool that we should leverage the ability for it to contribute is in even from creativity which includes the journey of the story to making the content more user friendly to making sure that the quality of the content is the absolute best so right from zero buffering increasing resolution removing delay improving dynamic range removing noise removing artifacts to recommendation engines video audio x-rays you know being able to participate in a show so I can use my face and paste it over the hero's face and this is the future though not right now but it's very much there and then in terms of creativity plot generation story generation short recommendation action following ball following news compilers democratizing vfx modernizing content and generating content I know there is much more but really today where we are is that we can take your content we talked about sd as a reaction to the present situation we do not have to view in sd you can transmit in sd but we'll can still view in hd that we can do today so so in in very precise summary what bhopi has put across is there are lot more users than just recommendation and hero short when it comes to ai and ml as such whether it is content creation at one part of it the other one is when you get to consumption of it the device that you're using based on the device that you're using automatically scaling up and scaling down in the kind of experience that you get picking off where you left use talk on one device you pick it up in another device that's also a bunch of technology which is helping you make happen seamlessly as as you go the language barriers when you look into it the entire menu may be customized to you in your localized language may not be the content itself but at this point of time but the language of the application itself is customized to your needs the interest based recommendation is always there htst like like how bhopi mentioned then then basically your interactive content the content it's a there are certain shows where you can choose the outcome you want to have a happy outcome or you want to kill the heroine in the end as such so you it's in your hand the producer is and the director is no longer taking those decisions whatever appeals to you you can watch that accordingly so these are some basic examples of how things are going in this direction so as i see the clock is hitting uh five three any any closing coming from anyone then maybe we can just close up on this one any closing comments i'll just add to the last one you're progressively moving towards the cost optimization you are using the AI nml techniques uh when do we need more servers how do you optimize your cdn cost all of these are now based upon signals that we process and have predictive analysis at hand do we need to scale up for a weekend seem straightforward but with how much factor very interesting yeah how do you basically identify the right cdn partners uh that is also something that we are putting in models and trying to understand it uh because you know if a partner a delivers a 99 percent uh versus a partner that serves for 100 percent that is that one percent difference uh worth 1.5 bucks money so on a conclusive note if i'm putting on my imaginative hat to slightly hide my lockdown haircut as well uh so what what we see here is uh maybe in the future uh as you open an app and then start pressing to view a particular uh show of your choice maybe your lights will dim as is your blinds will come down uh automatically your app will place an order for a coke and a popcorn why from your nearest vendor uh it will automatically post inviting a few friends to watch along with you and that may be as as small as your or as big as your imagination can go go buy so that's that's what i would like to say and conclude and hand it back to jati and this scenario that's 1984 all in one we don't want really to learn that word when somebody else decides what should happen other than popcorn you can always give me popcorn fine i love that idea so i think we have decided it says you want to order a popcorn along with it you just say yes to alexa and then it'll do it for you thank you thank you mr ral and uh all our panelists thank you for your time and i could see that by the end we were becoming really imaginative on what uh artificial intelligence can do to us in this quarantine but thank you so much for your valuable time i think i've taken some uh highlight points from this discussion which makes me really happy and excited that someday i can be the face of the heroine and watch the movie as i would imagine otherwise but thank you so much gentlemen for your time uh to all of you who are watching us you can rewatch this because the live will be available on our facebook page for you to rewatch it as well you can hashtag e4m webinar and uh you know tell us your comments your views and you can tweet to us using our tag e4m tweets i would once again like to thank all of you gentlemen for spending this time and discussing this really interesting topic because all of us are all the time on our otd platform so you want to know what is new that is happening there so before moving forward and concluding this webinar i would like to tell all of you that myelin foundry has partnered with us to organize this webinar myelin is a deep tech company and they are launching a product called four view stream which deploys artificial intelligence on users mobile devices to enable streaming at disruptive costs and even in very low networks so here is a short video to demonstrate and highlight their innovation let's take a look so once again thank you everybody for watching this e-webinar and thank you once again to our session chair and our panelists for your valuable time stay tuned to exchange for media page we'll be coming up with more interesting contents and webinars and discussions that will definitely enrich your lives while we are here in lockdown at home so thank you very much i'm chati kawai who's signing off we'll see you again