icymi, here's the claude skill: https://github.com/veris-ai/veris-skills

Author here: Looking forward to discussion!

nw_wrld is an event-driven sequencer for triggering visuals using web technologies. It enables users to scale up audiovisual compositions for prototyping, demos, exhibitions, and live performances. Users code their own visual modules, then orchestrate them using the project's native UI composer.

OP here -- I work at Veris and built this. Happy to answer questions about the methodology!

SantaBench, a fun benchmark with a serious methodology. The task: play a cheeky Santa agent who researches users online and roasts them based on their social media.

excellent post -- I turned it into a video :) https://supabase.manatee.work/storage/v1/object/public/video...

Okay, this is mind blowing!

complete overview of the Model Context Protocol

Code and a recipe for XTTS_v1.1 GPT encoder training is released under the Mozilla Public License 2.0

XTTS model release (Text-to-Speech and voice cloning)

# From the release notes:

This model is trained on top of XTTS v1, using output masking. We mask the part of the output that is used as the audio prompt while training and don't compute loss for that segment. This helps us to resolve the hallucination issue that V1 experienced.

- Add Japanese - Resolve the hallucination issue (repeating the audio prompt) - Increased expressivity - Added ne_hifigan that was trained without denoising that brought some EQ and compression profile that might be unwanted for some use-cases

XTTS + Whisper + Mistral 7B

I didn't hear about xtts till now, cheers.