61–70 of 383 posts
I don't mind so much the size in MB, the fact that it's pure CPU and the quality, what I do mind however is the latency. I hope it's fast. Aside: Are there any models for understanding voice to text, fully offline, without training? I will be very impressed when we will be able to have a conversation with an AI at a natural rate and not "probe, space, response"
(it does however explain how many of these comments are older than the thread they are now children of)
What's a good one in reverse; speech to text?
Web version: https://clowerweb.github.io/kitten-tts-web-demo/ It sounds ok, but impressive for the size.
Does anybody find it funny that sci-fi movies have to heavily distort "robot voices" to make them sound "convincingly robotic"? A robotic, explicitly non-natural voice would be perfectly acceptable, and even desirable, in many situations. I don't expect a smart toaster to talk like a BBC host; it'd be enough is the speech if easy to recognize.
We've had formant synths for several decades, and they're perfectly understandable and require a tiny amount of computing power, but people tend not to want to listen to them:
https://en.wikipedia.org/wiki/Software_Automatic_Mouth
https://simulationcorner.net/index.php?page=sam (try it yourself to hear what it sounds like)
Web version: https://clowerweb.github.io/kitten-tts-web-demo/ It sounds ok, but impressive for the size.
Does anybody find it funny that sci-fi movies have to heavily distort "robot voices" to make them sound "convincingly robotic"? A robotic, explicitly non-natural voice would be perfectly acceptable, and even desirable, in many situations. I don't expect a smart toaster to talk like a BBC host; it'd be enough is the speech if easy to recognize.
Well sure, the BBC have already established that it's supposed to sound like a brit doing an impersonation of an American: https://www.youtube.com/watch?v=LRq_SAuQDec
I'm surprised that phone manufacturers do not include good TTS models in their browser APIs for example. So that websites can build good audio interfaces.
I for one would love to build a text editor that the user can use completely via audio. Text input might already be feasible via the "speak to type" feature, both Android and iOS offer.
But there seems to be no good way to output spoken text without doing round-trips to a server and generate the audio there.
The interface I would like would offer a way to talk to write and then commands like "Ok editor, read the last paragraph" or "Ok editor, delete the last sentence".
It could be cool to do writing this way while walking. Just with a headset connected to a phone that sits in one's pocket.
I don't mind so much the size in MB, the fact that it's pure CPU and the quality, what I do mind however is the latency. I hope it's fast. Aside: Are there any models for understanding voice to text, fully offline, without training? I will be very impressed when we will be able to have a conversation with an AI at a natural rate and not "probe, space, response"
Average duration per generation: 1.28 seconds
Characters processed per second: 30.35
--
"Um"
Average duration per generation: 0.22 seconds
Characters processed per second: 9.23
--
"The brown fox jumps over the lazy dog.. The brown fox jumps over the lazy dog.."
Average duration per generation: 2.25 seconds
Characters processed per second: 35.04
--
processor : 0
vendor_id : AuthenticAMD
cpu family : 25
model : 80
model name : AMD Ryzen 7 5800H with Radeon Graphics
stepping : 0
microcode : 0xa50000c
cpu MHz : 1397.397
cache size : 512 KB