[Twitch Cloud service idea]Cloud assisted streaming of realtime games by netcode or replays with voiceover/webcam overlay to let people with low bandwidth able to stream cloud rendered ultra quality content over twitch.

I know I propose 2000 ideas a week some could be bad or needing more than a few engeneer to built, but I just finished to see your stream on twitch and I’m very excited about what you are doing and really wanted to contribute in some way and come up with this idea.

I live in Italy and in my area internet has no viable upload to make any decent quality streaming possible so I come up with this idea and I hope you will love it as I’m.

What if I just need to stream multiplayer net data(even for single player games the streamer really play as a client connected to the game running on the server in the cloud) or few mb of replay and a server render and stream the result for me while I send a voiceover or low definition overlay video with the few bandwidth I have?

The streamer then could also stream at an highter resolution than the one it uses to play if its hardware it’s not that good too. it could democratize a bit the streaming and you always have the latest top notch streaming quality without any hardware, software or internet upgrade.

In perspective it will make it possible a day to stream at 4k hdr at 60fps(when twitch will support it) from just an ldr tablet with 3g internet access playing the game at 25 fps.

People pay a lot for their streaming setup costs and Amazon can then compete with screen recording software and client hardware updates offering a better service for the price.

Naturally every Lumberyard game will support this by a gem and naturally if you do a few bucks with this idea rember about me.

Some Additional consideration about the possibilities.

  • Using a multigpu instances the quality of the game could be superior to the one shipped for the pc / console. So developers could develop asset for the cloud and then scale down them for the consumer version of the game.take just in consideration how easy is to just scale up procedural content es speedtree or substance textures to highter resolution.
  • Watchers can have a totally different view of the game than the streamer,es look by cinematic angles, different/more dramatic post processing with more or less or different ui elements.so games could be developed to be a greater streaming master or just offer this possibility to watchers on vote base or entitling temporarely one as a game master.
  • VR streaming rendering spherical views of a non vr game people can then watch using cardboard.
  • Server can have available a huge list of overlays the streamer access with the service without to buy them elsewhere making the amazon always the middleman between streamers and overlay makers.
  • In some games the streamer also need to upload it’s savegames(so uploading them directly on progress on the streamer account could be an option and he will be always ready to stream).
  • The game may also register automatically data replays and always ask at the end of a session if the player want to upload them.
  • Cheats are a bit limited by the fact the player is connected to an authoritative server even when playing single player games.
  • As the game runs on a server streamer can setup the server to arbitrate watchers joining the game with their client streamer can at any time convert the single player game to a multiplayer one as the server is just there.
  • imagine watchers upload their ghostcar by an interface in twitch and get them voted or randomly picked and added to the stream on the next lap of a racing game.
  • I stop there as possiblities are quite infinite when you don’t have the bottleneck of the streamer hardware and software setup and the power of the cloud.

Great, Thanks Binky I’m happy you liked it.please let me know if there are any news.

As for the tecnical concerns I know that it’s surely a non trivial matter.

To leverage the multi gpu instance will require some work and other work if you want to offload encoding to specifically optimized instances and mainly for whatever other aspect you want to offload.

Microsoft for example has successfully offloaded the physics for a first party game so it is totally possible.

Imagine if whatever game made with lumberyard could display in streaming even a single game aspect just 1000% but also just 20% better than the non cloud enhanced version just checking a box who will be able to resist to the temptation?

I believe that athe most sensible aspect is probably the business model someone have to make a proper plan to evaluate how much you can push with these things and still be sustainable who have to pay for it between ads,devs,streamers,watchers etc?

I believe that this could be the major issue way more than technical ones .I can just speculate about how complex is in a company like Amazon just to start a project like this.

However I believe that a huge aspect to take in consideration is the effort of third party developer to make something engaging for their games using these new possibilities it will be a race to make the best out of the cloud in a everyone wins scenario.

One again you come through with a great idea. I’m not a technical guy, but this sounds great to me. I’ve pushed it to the team. Thanks for all your ideas!

Yeah, it’s a very interesting idea. I personally believe we’ve barely begun to scratch the surface of what we can do with cloud computing power. Same goes with VR.