wczekalski
Joined 13 years ago
Karma
237
Submissions
43
Member since
May 5
Recent submissions
Comment · 9 months ago
I was recently migrating a large amount of data off of MinIO and wrote some tools for it in case anybody needs that https://github.com/dialohq/minio-format-rs
Comment · 1 year ago
We moved to Seaweedfs around one year ago and I couldn't be happier. It also fixed all of the performance problems we had on MinIO.
Comment · 1 year ago
Dialo | ONSITE | Warsaw, Poland | https://dialo.ai | Full-Time
We build voice AI agents for customer service over the phone. We are currently conducting hundreds of thousands of calls every day. Dialo is profitable.
We develop and self host every component across speech-to-text, text-to-speech, LLMs, real-time audio processing, and voip.
Our strategy centers on developing an open-source foundation and propriatery vertical-specific products on top of it.
tech stack: AI (wav2vec2, llama, styletts2), OCaml (systems eng), Nix, k8s, TypeScript, React, Node.js
Roles:
- Applied AI research (Speech LLMs)
- AI engineer
- UI Engineer (a full stack role with frontend focus)
How to apply: mail me at [email protected] with the coolest thing you build and tell me what you want to work on! (github link and/or resume optional
Comment · 1 year ago
It is very useful for long lived (bidirectional) streams.
Comment · 1 year ago
dialo | ONSITE | Warsaw, Poland | https://dialo.ai | Full-Time
We build voice AI agents for customer service over the phone. Our goal for 2025 is to open source a SOTA speech llm for customer service use case as opentalk.org. We currently operate on the Polish and US market focusing on financial services industry. Our business is growing rapidly but we are a small team (3 devs) so you can have a huge impact.
tech stack: AI (wav2vec2, llama, styletts2), OCaml (systems eng), Nix, k8s, TypeScript, React, Node.js
Roles:
- Applied AI research (Speech LLMs)
- AI engineer
- Front end/full stack developer
- Systems engineer
- SRE + any of the above hybrid
- Go to market engineer (entry level role)
How to apply: mail me at [email protected] with the coolest thing you build and tell me what you want to work on! (github link and/or resume optional)
We are also offering $2500 referall reward for successful referalls. Just drop me a line if you want to recommend somebody exceptional and describe what's their most impressive achievement.
Comment · 1 year ago
My friend and I used to build a cross platform UI library in OCaml called… brisk. We didn’t make it production ready so i am sure the naming is a coincidence https://github.com/briskml/brisk.
Comment · 2 years ago
it simply stores objects as files on the disk. Then it distributes the chunks around the place (so you need to reassemble it when reading) and lastly, when you read the file, it's not O(1). There is some "discovery" process to locate the objects where the servers chat with each other rather than have the location stored somewhere and be O(1).
Comment · 2 years ago
Loki is not bad but PSA, don't use it with anything else than real AWS S3. The performance with Minio is awful (and can't be good, because of how minio works). Might be a bit better with Seaweedfs.
Comment · 3 years ago
What's your basis for the claim that they are based on TorToiSe? I have seen this claim made (and rebutted) many times.
Comment · 3 years ago
One thing I've seen done for style cloning is a high quality fine tuned TTS -> RVC pipeline to "enhance" the output. TTS for intonation + pronunciation, RVC for voice texture. With StyleTTS and this pipeline you should get close to ElevenLabs.
Comment · 3 years ago
have you tested longer utterances with both ElevenLabs and with StyleTTS? Short audio synthesis is a ~solved problem in the TTS world but things start falling apart once you want to do something like create an audiobook with text to speech.
Comment · 3 years ago
It depends on hardware but IIRC on V100s it took 0.01-0.03s for 1s of audio.