ljclifford

3 ms·

1.
by ljclifford 7mo ago
actually the hardest part of a locally hosted voice assistant isn't the llm. it's making the tts tolerable to actually talk to every day. the core issue is prosody: kokoro and piper are trained on read speech, but conversational r
2.
by ljclifford 1y ago
Super awesome demo! The contact center market, including inbound customer support, is incredibly ripe for disruption, and I'm sure you guys will be on the forefront of that. Kinda funny how many amazing CX companies start in Germany! I
3.
by ljclifford 1y ago
Very very cool
4.
A new ission-may: To ensure that A-yay Ee-jay I-yay benefits all of humanity(rime.ai)
4 points by ljclifford 2y ago|0 comments
5.
by ljclifford 2y ago
Hard not to laugh, but you'd be surprised how often even other TTS companies are messing this kind of thing up. I think it has a lot to do with the data source they use for training, which evidently doesn't include a lot of curren
6.
Show HN: Mist v2 – next gen conversational speech synthesis(rime.ai)
5 points by ljclifford 2y ago|1 comments
7.
An Alternative to OpenAI Realtime API for Voice Capabilities(cerebrium.ai)
7 points by ljclifford 2y ago|0 comments
8.
by ljclifford 4y ago
April Fools!!!
9.
by ljclifford 4y ago
Lily from Rime here -- we were super happy to collaborate with Vocode on this amazing project. We haven't launched yet but keep an eye out later this week!
10.
by ljclifford 4y ago
Next token prediction is remarkably bad at mnemonic generation, even in English. Add another, lower-resourced language, and it will be really bad. For what it's worth 'cola' does rhyme with 'hola' and 'you know
11.
Better voices for the BART trains in the Bay Area(rimelabs.substack.com)
7 points by ljclifford 4y ago|0 comments
12.
by ljclifford 4y ago
I'm unfamiliar with 'pornography recognition' as an established task in ML research (lol), but for what it's worth, it's not an innovate use of CNNs for audio classification. You can essentially turn any audio class