i asked fable 5.2 to write audio samples by hand to see if it could synthesize speech. this is not actual formant synthesis, this is literally claude writing the audio samples directly. pretty wild
AI interfaces nowadays are still pretty much turn-based.
In my PhD thesis, I thought about what interaction would look like if we could go beyond taking turns.
One example is intra-turn interaction:
Most of us won’t be part of a band, but we'll all be musicians contributing to albums.
The AI conversation is mostly about the technology layer, so I'm starting @YetAnotherMag to understand what's changing at the human layer.
https://t.co/epZNboZOT3
For the past year, my favorite workflow for writing has been: walk around outside speaking into a voice note, sometimes for up to an hour, collecting my thoughts and structuring the piece out loud. Then, ask an agent to turn the voice note into an outline of bullet points. Finally, sit at a computer and turn the bullet points into prose.
I think this two-phase approach plays to the strengths of both modalities. Walking helps me get deep into thought and avoid distractions. Working at a screen is better for being sharply critical of the ideas and getting the wording right.
If you squint, it’s a bit like “write drunk edit sober”.
I made a full original game in 24 hours using AI
Stack: Opus 5.5 + Flux 3 Max + GPT Image + Tripo + Blender + Three.js
Anyone can make games now. Full workflow thread and prompt below:
(1/7)
Ixi’s album breakdowns are so informative and emotional for those of us who have these pieces etched into our DNA. Here’s her recent analysis of ones one my all time fav albums: Aphex Twin’s SAW2 https://t.co/mBlss27wdj
Wow, I've been waiting for this day to come for years. A widely available, low latency phone assistant with an intuitive UX on top that gives the human real-time conversation monitoring, steering, and smooth call hand-off back to the human. Amazing work @bot team @benjitaylor
over the past few months we've been testing a new app with designers
with @mainframe you can design directly in your codebase on a canvas alongside your agents and team
we're actively rolling out invites. join the waitlist
Built a physical simulation of a record player with WebGL2 and Web Audio. It's *shockingly* accurate: https://t.co/MBNaTfKoai.
Instead of playing a song, it works by converting it into the physical shape of a groove. Then a simulated stylus rides those walls 384,000 times a second, and you hear the voltage that motion would induce in a cartridge.
So the crackle, the pops, the wow/flutter, the dull inner grooves are natural byproducts of the simulation. You can also "DJ" the record.
(The converted songs are in the public domain or CC-licensed. Code is linked for you to remix and re-deploy.)
my vocal synth models the vocal tract as a network of branching acoustic tubes, unlike vocaloid which stitches together and pitch-shifts phoneme recordings. still working on improving naturalness by tuning it against recordings of real singers, but I'm already impressed at how cool it sounds!