Thursday, June 27, 2024

Show HN: Voice bots with 500ms response times https://ift.tt/IpzE2Tg

Show HN: Voice bots with 500ms response times Last year when GPT-4 was released I started making lots of little voice + LLM experiments. Voice interfaces are fun; there are several interesting new problem spaces to explore. I'm convinced that voice is going to be a bigger and bigger part of how we all interact with generative AI. But one thing that's hard, today, is building voice bots that respond as quickly as humans do in conversation. A 500ms voice-to-voice response time is just barely possible with today's AI models. You can get down to 500ms if you: host transcription, LLM inference, and voice generation all together in one place; are careful about how you route and pipeline all the data; and the gods of both wifi and vram caching smile on you. Here's a demo of a 500ms-capable voice bot, plus a container you can deploy to run it yourself on an A10/A100/H100 if you want to: https://ift.tt/iRzxySB We've been collecting lots of metrics. Here are typical numbers (in milliseconds) for all the easily measurable parts of the voice-to-voice response cycle. macOS mic input 40 opus encoding 30 network stack and transit 10 packet handling 2 jitter buffer 40 opus decoding 30 transcription and endpointing 200 llm ttfb 100 sentence aggregation 100 tts ttfb 80 opus encoding 30 packet handling 2 network stack and transit 10 jitter buffer 40 opus decoding 30 macOS speaker output 15 ---------------------------------- total ms 759 Everything in AI is changing all the time. LLMs with native audio input and output capabilities will likely make it easier to build fast-responding voice bots soon. But for the moment, I think this is the fastest possible approach/tech stack. https://ift.tt/iRzxySB June 27, 2024 at 03:21AM

Show HN: Quick Rewire – Test your brain's neuroplasticity https://ift.tt/YxRjpMr

Show HN: Quick Rewire – Test your brain's neuroplasticity Hey HN, this is a game to test your brain's neuroplasticity. It tests how fast your brain adapts and rewires. It doesn't work with mobile browsers. https://ift.tt/CJF0TmD June 27, 2024 at 02:33AM

Wednesday, June 26, 2024

Show HN: I built a static site builder https://ift.tt/37Mqa5p

Show HN: I built a static site builder https://ift.tt/FAj20Dt June 25, 2024 at 11:38PM

Show HN: From dotenv to dotenvx – better config management https://ift.tt/vI1nqrZ

Show HN: From dotenv to dotenvx – better config management https://ift.tt/6uFKpg9 June 25, 2024 at 08:19PM

Show HN: The Tomb of Nefertari [QV 66] Guided Virtual Tour https://ift.tt/459X02s

Show HN: The Tomb of Nefertari [QV 66] Guided Virtual Tour I 3d scanned the Tomb of Nefertari and am building this guided virtual tour, trying to bring in photogrammetry of artifacts that I've made at various museums. It crashes sometimes still on mobile devices. I wasn't able to take photogrammetry photos of the artifacts from the tomb in the Museo Egizio in Turin because they were traveling while I was there (and now the museum is closed to install a new roof anyhow), so I tried to include comparanda from other museums where I had scanned artifacts. I tested the same dataset with 3d Gaussian Splatting but that had lower resolution results for great complexity in frontend code and reduced performance on older devices. [3DGS tour: https://ift.tt/FKHaY2l... ] Moving forward, if anyone's working on the same idea, I didn't find a good path to monetization through web-based 3d content, so I'll take the high resolution photogrammetry of spaces into Unreal and switch to desktop and headset builds. For the distribution advantage of web-based 3d content, I'm hoping to test next a desktop or VR build that may be easier to sell to sustain my work. https://ift.tt/WZbHqKl June 25, 2024 at 08:18PM

Tuesday, June 25, 2024

Show HN: I made a remote job board with a free CV maker tool https://ift.tt/hobZp7c

Show HN: I made a remote job board with a free CV maker tool https://benture.io/ June 25, 2024 at 02:27PM

Show HN: AI-Generated Tables for a Steampunk City https://ift.tt/TCo8hDO

Show HN: AI-Generated Tables for a Steampunk City I recently embarked on a creative experiment where I used ChatGPT to craft an elaborate steampunk story. Inspired by the rich narrative and intricate details, I decided to transform various parts of the story into demo tables, complete with text and images using https://tabulator.ai/ . The goal was to explore how AI can aid in the creative process and provide indie game developers and storytellers with some ideas quickly. https://ift.tt/DVtrkqZ June 25, 2024 at 02:34PM

Show HN: Free OSS transcription app I made and found it's faster than wispr flow https://ift.tt/jXQh9Tk

Show HN: Free OSS transcription app I made and found it's faster than wispr flow title doesn't let nuance, ofc it's not the app ...