I'm a game designer/game dev. Started in 2004 on The Witcher. Went indie in 2015. Bought a microphone to record audio, bought a mocap suit to record animations, etc. :-). Now you can get pretty much everything by just typing something into a text box :-)
@ai-aniverse6 ай бұрын
if only...
@mikaelasimonsen20176 ай бұрын
As a dev how do you feel about technology such as this? Really curious to hear from someone who has been in the field for a long time. P.S: Thanks for The Witcher games a reality (that goes for all the devs and people involed).
@mszczesnik6 ай бұрын
@@mikaelasimonsen2017 I do woodworking as a hobby :-). So worst case scenario, I'll be making furniture.
@ai-aniverse6 ай бұрын
@@mikaelasimonsen2017 i dont have tje same depth as the (only 12 years in) but im frankly excited. I also dont work in games. I work mainly on the touchscreen devices you see everywhere to put it simply. from medical to automotive to aviation. My 2c is that it will be useful to senior devs, i wonder how a jr dev does get a foot in the door depending on the flavor of the engineering. In my line of work, its still a little too complex to be 'automated' That being said, i was also a general contractor before this so...back to swinging a hammer or fixing plumbing which isnt exactly a bad job.
@machinosts5 ай бұрын
Give it a couple of years and you will not have a job. It's coming for you next.
@cbuchner16 ай бұрын
Mario Wahoo and ASMR had me ROFL
@Cosmic_Caravan.6 ай бұрын
This is incredible, all the pieces for video game production are almost complete. Using this with Suno + Sora + potentially GPT-5 / GPT-6 you'll have the power of a small company.
@PdWOLFG4NG6 ай бұрын
Hollywood😅 awaiting patiently 😁
@gisfdlc92106 ай бұрын
🤣🤣🤣@@PdWOLFG4NG 🤞
@markmuller79626 ай бұрын
Did you see yesterday and today news of Devin? That's the biggest piece of your puzzle
@kooostia166 ай бұрын
that sadly seems more like scam@@markmuller7962
@Cosmic_Caravan.6 ай бұрын
@markmuller7962 I did, just after posting this 😂 will be amazing to see what Devin or a similar agent could do if it's given access to all these tools
@levonkenney6 ай бұрын
this combined with sora will make the movie producers worried
@CanadaBlue856 ай бұрын
It's more of a threat to actors, voice and foley artists than producers.
@notacreativename16 ай бұрын
Going to make it aaaaaalot easier/more accessible for people wanting to visualize their imaginations. It’s going to be awesome
@HelamanGile6 ай бұрын
I'm a movie producer I am not worried I am sure this will speed up production but Sora is not yet up to standards for film production and unreleased
@Always.Smarter6 ай бұрын
why would they be worried about the job of their employees getting easier?
@CanadaBlue856 ай бұрын
@@Always.Smarter their employees jobs will be getting eliminated eventually, followed by their own jobs. Also they are the producers subordinates not their employees.
@MattVidPro6 ай бұрын
What do you guys think!? I was blown away!
@MrEthanhines6 ай бұрын
Every day, getting closer and closer to AGI
@francesco78006 ай бұрын
This technology is incredible! Speaking, however, of what I thought, I must say that it makes sense that the model is not able to generate contiguous sounds such as "train followed by the scream of a man" or "unicorn followed by the sound of glitter". After all, in any online/offline sound effects library we don't find composed sounds. They just have no purpose. If you think about it, this is a good thing, because the sounds must be taken individually and composed in post production to give the best experience. So having one sound followed by another within the same audio track would mean having to adapt the video to the sound and not vice versa, making the sound essentially useless or more difficult to deal with. To conclude, if you noticed, the model you used managed to generate single sounds very well and with high quality, but the composed ones sent him into paranoia. I think it would have been interesting to see if it could generate sounds like "storm", "drowning under water", "drone sound", "hit", "whooshes", "foleys" or any other named sound used in cinematography. What do you think about it?
@musartai6 ай бұрын
That's fantastic! However, you need to be a bit more specific. For instance, when you say punching or punch, what exactly are you punching? I believe specifying the object or material you're punching will make the sound much more closer to what you're expecting to hear.
@TechRenamed6 ай бұрын
Well I think it's amazing
@armondtanz6 ай бұрын
you should check out my edit AI, has great sound fx like this and u can use now...
@hansgruber34956 ай бұрын
Can we hold a moment of silence for the men falling down the stairs, I'm pretty sure none of them will ever stand up again 😢🪦
@kikeluzi6 ай бұрын
😰 That may have hurt a lot . . .
@keithprice33696 ай бұрын
A lot of those would SEEM a lot more realistic when paired with video showing the event. Hearing the can crunching by itself is sort of "Okay, I can kind of imagine it". But show a man crunching a can with the sound effect and we'd be, "Yep! That's real!" Same with the man falling down the stairs. Without seeing the man falling, the sound is a bit cartoonish. But with the visual? Absolutely. It's a bit like the talking piano, if you know what I mean.
@nathanbanks23546 ай бұрын
ASMR is definitely the highlight of this video 14:35, I'm in tears laughing at the people who expected normal ASMR. However the warp drive sounds pretty good in stereo 11:50.
@welsieinfinland13252 ай бұрын
Thank you for spoiling :/
@TerraDiASMR2 ай бұрын
Definitely had me laughing… for now, but in a few years or even months when ai gets better, non of us (ASMR creators) will be laughing 😅😬
@welsieinfinland13252 ай бұрын
@@TerraDiASMR wow you are here
@jimmysrandomness6 ай бұрын
Elevenlabs the King of AI voice. Still waiting for more emotion control and speed control👀
@willmfrank6 ай бұрын
You could try their relatively new Speech-to-Speech model; upload an audio file of yourself speaking with the emotional expression and pace and cadence that you want the narrator to emulate, and the AI will simply replace your voice with that of the narrator of your choice.
@willmfrank6 ай бұрын
Not sure; I haven't had either a reason or opportunity to try it myself; I know only what I've read on the ElevenLabs website. I am working on a video that requires narration, though, so I might give it a go this weekend@@eduardomartin8510
@tigerxplso6 ай бұрын
10:33 I laughed so hard at the mario one :D
@cbuchner16 ай бұрын
that‘s a lot of stairs 🤕
@welsieinfinland13252 ай бұрын
Infinite stairs from mario 64?????
@Dude_Wassup6 ай бұрын
Matt if you go to 16:50, you’ll hear “help me please”
@HelamanGile6 ай бұрын
Wow
@jonorgames65966 ай бұрын
@@HelamanGileomg
@amsgamingandmusic6 ай бұрын
That unsettled me...
@cesar47296 ай бұрын
11:37 too.
@arinco38176 ай бұрын
Loving this video! As I'm watching this, I'm hoping Matt is currently recording one about devin lol
@ireadclassiccomics31726 ай бұрын
I use a lot of sound effects on my channel reading classic comic books, and most sound effects come from KZbin, so this tech may come in really handy when I need one of those really hard to find sounds!
@relcnt6 ай бұрын
this is really cool, i can see this being utilized in indie games or maybe even some short movies
@ChristianIce6 ай бұрын
"mario wahoo" is the magic formula to get zombies sfx :)
@Metarig6 ай бұрын
I guess with AI sound effects, we're past the days when someone could record their fart, slap a copyright on it, and sell it for 50 bucks a pop.
@DGreen9516 ай бұрын
Lol was that vomit ASMR 😂
@darkfyrmedia69126 ай бұрын
Just saying, 11 seconds adds a great deal of humor.
@metatron39426 ай бұрын
This will see the light of day unlike Sora which I'm afraid be massively delayed maybe someday
@kuromiLayfe6 ай бұрын
think the biggest issue with sfx data used in these models is that 90% is very noisy and low quality data … unless these companies can get access to the sfx libraries from hollywood and the likes and build the models from those libraries alone, the quality will not improve as rapidly as with music or image/video. like i had a 1530 GB (1.8 million audio files) pack of sfx and had to delete 800 GB because it was unusable
@via_kole6 ай бұрын
the man falling down the stairs was kind of scary in a way. i could feel the impacts xD
@nemonomen33406 ай бұрын
6:41 Oh yeah, the creepiness has _nothing at all_ to do with the implications of a man falling hard down a set of stair for 10 seconds straight.
@Ether8206 ай бұрын
11:01 Feel like I can almost hear the samples on some. Goldeneye Theme, CSI Miami Theme. Probably others. Mario 3 start (crowd in the dark, not as close but maybe a Mario link for why it generated that).
@matthewdignam73816 ай бұрын
This would be sick for sound effects in a horror game
@blindstreet6 ай бұрын
Not only in stereo, sounds like it's binaural too. About the man falling...that's looong staircase. I'm an audio editor so I know what to hear in a sound effect. Can someone try...'neighbors having a kinky session while your roomate is snoring in the background
@zippythinginvention6 ай бұрын
Temporal AI like video/audio don't like past-tense prompts. Humming works better than hums. Talking works better than talks.
@DumPixels6 ай бұрын
Neither hums nor talks are past tense words…
@zippythinginvention6 ай бұрын
@@DumPixels I'm pointing out two things.
@DumPixels6 ай бұрын
@@zippythinginvention Ah, ok my bad
@chaserock46756 ай бұрын
This episode had me rolling! LOL!
@markmuller79626 ай бұрын
The mario ones made me laugh out loud 😂
@diamonx.6615 ай бұрын
Those bizarre car ones feel like they belong in a Game Theory video. The AI's alive and trapped, what's it trying to say?
@TeleviseGuy3 ай бұрын
I love how "Viewers" is its own chapter in the AI-generated chapters lol
@GS1956 ай бұрын
This could be good for sound design for music production
@Oneirio6 ай бұрын
What my brain decides to produce right before I fall asleep 16:35
@official_meelees3 ай бұрын
put quotation marks around parts you want to be spoken and it speaks them.
@LoneBagels6 ай бұрын
Dudee.... I cannot wait to make my next horror trilogy :D
@bobhawkey37836 ай бұрын
As usual for these things it works well about 20% of the time. And they ding you 10 characters for each of 5 generations out of which 1 may be useable. Cool though! Maria wahoo is amazing!
@HexOverride6 ай бұрын
Freesound is shaking in their boots right now 😂
@chad_usa6 ай бұрын
Someone didn't watch the video
@mihailniagolov36746 ай бұрын
I wonder if you can cover the new self-teacing humanoid robots. I see it's huge topic but it'd be cool to talk about it
@blackshard6416 ай бұрын
8:09 Eleven Labs starts breaking out into a rendition of Low Rider
@simonstrandgaard55036 ай бұрын
Listening to this. Last time I toyed with audio was more than 20 years ago. Impressive quality. Mindblown.
@Jeal0usJelly6 ай бұрын
Video on consistent characters? Honestly can't wait 🤩 I think soon we'll be able to do consistent everything, for example just today I read about a new LLM called Command-R, which apparently is more accurate than anything else in answering "needles-in-a-haystack" type of prompt, which is promising. I'm hopeful GPT-5 will also be more than an incremental change and put pressure on the whole industry so we'll eventually see it trickle down to open-source models too.
@MrTk34356 ай бұрын
Matt! another home run review Thank a lot 🔥🔥🤟🔥🔥
@TheGeeMaster13373 ай бұрын
10:38 Mario mating call
@hipjoeroflmto47646 ай бұрын
15:59 oh no it's ocean gate all over again
@adisatrio38716 ай бұрын
I really hope that the developer don't patch out the Mario Wahoo and ASMR generation. Let those be a nightmare fuel hidden prompt lol
@GaryJr5306 ай бұрын
Pov: hoping you try the "falling down the stairs while eating an apple" sound effect
@Deckardb253646 ай бұрын
It's giving me some impressive sounds related to speculative biology. Like, non existent animals. Made a python project that like; no mans sky's a bunch of text based creatures and uses a shitload of generated noises i made with it. It definitely, definitely is already aboce AudioLDM 2.
@SMmania1236 ай бұрын
12:09 woah
@darkwing_the_spacecat6 ай бұрын
11:08 What the FUUUUCK, lock that thing in the basement and set it on fire, holy hell! XD
@Fustercluck066 ай бұрын
I lost it on the Mario wahoo 😂 😂 😂
@user-jv7ig6ie5b6 ай бұрын
They need to integrate this directly into their projects UI so you can insert them quickly
@SuMiTMeshram256 ай бұрын
we needed something like this
@ArnoSelhorst5 ай бұрын
Love the Mario, ASMR and car eating you sounds!😅 Bonus: your facial expressions while listening to these abominations.😂
@limeflashlight41016 ай бұрын
I like computer .
@wedontexist3696 ай бұрын
Let’s hope computer like us back
@MattVidPro6 ай бұрын
@@wedontexist369thats what im sayin
@virtualfg6 ай бұрын
Ye
@Pumpkin5256 ай бұрын
I wonder how well it would do sounds for aliens or monsters in a horror game.
@deadplthebadass213 ай бұрын
It gave me a bunch of ideas immediately
@noop-chair6 ай бұрын
Tbh if I am making a game I would rather use meme soundeffects.unless they become better
@agnesslovehealz6 ай бұрын
Mario wahoo reaction matt epic live for this lol priceless and the stereo futuritic sound effect love
@MrJaggy1236 ай бұрын
"This is such nightmare fuel! How can we dial it back a notch? I know : 'Man shrieking in terror'" 😉
@stepphun5 ай бұрын
oh... mh ... i thought this would take a video as input an then lay out a sfx layer timed to the animation/video. describing every single soundeffect is a bit ridicoulus isnt it because its will take the same time to generate and evaluate each soundeffect to record and edit it myself, exept the recording is way more fun.
@Ethanmyertrains100Ай бұрын
Train horn 13:15
@suzannecarter4456 ай бұрын
Your expressions while listening are hilarious!
@sybervisions6 ай бұрын
This is an insane features, I cant wait to try them on my projects!
@HikingWithCooper5 ай бұрын
I have purchased ~100GB of AFX and the problem is that there are SO many files, it's impossible to find what I'm looking for. I would gladly use this instead, but maybe in another version or two.
@ArnoSelhorst5 ай бұрын
Without a doubt one of your most fun videos. Never laughed so hard. Thank you so much!
@nevets56 ай бұрын
Some high quality ASMR...
@Kavokane6 ай бұрын
Now I regret watching this at night 😳
@natecodesai6 ай бұрын
This is pretty cool. If I want to suggest something for your AI audio video series how do I go about that?
@EmilyNilsen6 ай бұрын
That was a lot of stairs
@chariots8x2306 ай бұрын
It would be cool if we could pick the exact number of seconds & milliseconds that we want our sound effect to be.
@harnageaa6 ай бұрын
Good stuff, not on the suno AI level, , feels more like a suno AI v1 type of thing. SO maybe in 2 generations it'll be totally usable.
@Razumen6 ай бұрын
I think some parts of your prompts aren't needed, like the "coming out of a faucet" isn't really needed when you could just have "water being poured into a cup." So that may be messing up the generation, especially when it seems like you had more success the simpler your prompts were. Plus, usually sound effects are made separately and then combined later because it's more flexible. For instance, the sound of a train engine would be separate from the horn, and the sound of a man falling down stairs would be separate from the sound of a hundred marbles dropping.
@bendydave6 ай бұрын
they sound amazing in surround sound
@mrrfyW6 ай бұрын
I’m already holding my phone Matt as I’m watching this video on it Edit: Nobody laughed at my joke? Pathetic.
@iiijgciii6 ай бұрын
Typing ceramic cup will get better results its about how u say things
@Rapscallion20094 ай бұрын
I want to know when the equivalent of Automatic1111 is going to surface for SFX. :-)
@janwalter766 ай бұрын
Mario wahoo 😂😂😂
@zippythinginvention6 ай бұрын
"Wahoo" in the style of Mario
@cerebrumexcrement6 ай бұрын
sound quality is pretty nice. signed up.
@I-Dophler6 ай бұрын
🎯 Key Takeaways for quick navigation: 🎵 11 Labs presents AI-generated sound effects, showcasing impressive text-to-speech technology. 🎧 Sound effects range from soda can crunching to water dripping, with variations and nuances in each generation. 🌟 Impressive audio quality and stereo effects enhance realism and usability for various applications. 💥 Explosion sounds demonstrate significant improvement in AI sound generation capabilities. 🤖 While capable, AI struggles with more complex prompts like overlapping sound effects or abstract scenarios. 👾 Models excel in typical sound effects, such as punch sounds, suitable for video games and media production. 🚀 Early access limitations acknowledged; AI still evolving, not yet mastering intricate or elaborate prompts. 🎮 Overall, 11 Labs' AI sound effect generator shows promise for practical use in gaming and media production. Made with HARPA AI
@pinchopaxtonsgreatestminds95916 ай бұрын
You could have said "Man Hums" on the later sound effect, because it would be easy to layer the sounds anyway.
@TheLoreLabs4 ай бұрын
Ahhh I missed the give away 😅 only running rtx 1080
@floraphonic6 ай бұрын
Not great but still cool, my job is safe for now :)
@AlphaProto6 ай бұрын
Nice way to get license free sound fx.
@mylittleheartscar6 ай бұрын
The foley industry is gonna shake!
@Fresh204486 ай бұрын
Bro what do you think about Devin? Will you make a video about it?
@desu386 ай бұрын
They're awful as sound effects, but they make some neat samples.
@jdwrink6 ай бұрын
I was thinking the same thing. These sound like they could be in music by Aphex Twin or Venetian Snares.
@chariots8x2306 ай бұрын
It would be interesting if ElevenLabs learns how to do overlapping sound effects. I guess it’s probably complicated to have multiple different things happening in one sound effect.
@user-on6uf6om7s6 ай бұрын
It's also a balancing act of whether training for that introduces it where you don't want it which may be more of an issue. Like the "falling down the stairs" audio, it's all a bit uncanny without any screaming or grunting but I may have a specific idea for what I want with that and if I have those human sound effects mixed in, it would either make the sample unusable or I would have to use techniques to extract the bits I wanted. Ideally, you could ask for something happening while something else happens but unless they can really eliminate concept bleeding, I think this way is better.
@kc-jm3cd6 ай бұрын
You can overlap, episodes and post and just make single sounds video Game makers will or should embrace this. I can’t wait until LTX is more of a thing then we can start making actual movies.
@flagshipbowtie6 ай бұрын
What other sounds effects generators we have now?
@AdrienLatapie6 ай бұрын
AI haters will say "Just learn to do foley"
@FRANKENSTEIN-tn1ml6 ай бұрын
12:27 Cloverfield vibes
@hakaisauce6 ай бұрын
Have You seen about Devin! By cognition labs it actually blows my mind.
@stevesm20106 ай бұрын
Provided you get a useable effect…. LOL What’s the copyright and usage status?
@user-on6uf6om7s6 ай бұрын
You get the IP rights insomuch as Elevenlabs can provide them. The US copyright office's policy is that a prompt isn't sufficient for a copyright but that's probably not much of a concern for sound effects anyway. But yes, you have commercial usage rights.
@stevesm20106 ай бұрын
@@user-on6uf6om7s Thank you for the info.
@ydmoskow6 ай бұрын
They should add thumbs up thumbs down to help train the model
@aceyage6 ай бұрын
These sound like shit as if recorded with a phone, not anywhere close to professional sfx and 192kbps is not good quality. LOL. The enshittification works. Good taste is hard to come by these days.
@PdWOLFG4NG6 ай бұрын
Ffing amazing
@dffeqq6 ай бұрын
When the man starts talking
@tubestreamkyki6 ай бұрын
Let me tell you, the electronic music from now on can be hugely different! Those composers love experimenting.
@JulienMatthey5 ай бұрын
Isn't there anyone thinking that it sounds pretty bad, really?
@hakarthemage6 ай бұрын
It's cool but i don't think the quality is that good. They sound like a recording with a poor mic in a room with hard reflections