A room that responds.
Person segmentation and pose landmarks can turn movement into visual control. Combine these signals with a beat detector to build responsive installations and event visuals.
Try gesture-driven particlesEXPERIENCE LAB / THE AFTER-HOURS WORKSHOP
Curiosity,
in motion.
A place for experiments that connect images, sound, stories, and intelligence. Play with a visual sketch, then explore the public projects behind this direction.
EXPERIMENT 004 / SIGNAL PLAYDUOTONE
Remix the
landscape.
Change the look. Turn up the signal. Add a pulse and let the image move with it.
Original artwork. Ready to remix.
A real-time graphics and synthesized-audio sketch inspired by our media experiments. These pixel treatments are not model inference. Your selected image stays on your device; no microphone access is needed. For live AI perception, try the camera experiments.
THE NEXT EXPERIMENT COULD BE…
Serious engineering.
A playful instinct.
A recurring character.
Curate authorized character footage, review captions, and train a video-model LoRA to explore visual consistency across short generated scenes. Dataset quality and GPU evaluation remain central.
Explore the training scaffoldA story you can step into.
Synchronize images and video with a reader’s position. Combine careful sequencing with locally generated or licensed media to explore immersive learning and interactive storytelling.
Explore the readerThese are technically grounded directions for exploration, not additional claims of completed enterprise deployments.
FROM SAMRAT’S PUBLIC WORKSHOP
Open the hood.
Follow the curiosity.
Selected public repositories span model adaptation, video tooling, and immersive interfaces. Project scope and maturity are described from their public documentation, reviewed 2026-09-10.
LIVE VIDEO × SOUND
Starburst VJ
Camera or RTSP input, beat-reactive visual effects, and human segmentation experiments in a macOS Python application.
Experimental project. The browser sketch below is inspired by this direction; it is a separate implementation.
VIDEO × CHARACTER LORA
LTX Studio
A WebUI scaffold for LTX-2.3 inference, caption review, dataset organization, character LoRA jobs, and generation history.
Development scaffold with explicit dry-run and runner boundaries. Requires an appropriate GPU environment; not running inside this page.
GPU × MULTI-STREAM MEDIA
Fast Video Player
A native Windows player built around libmpv, hardware decoding, independent borderless windows, zoom, pan, and looping.
Media infrastructure rather than a generative model. README performance observations are machine-specific.
STORIES × SYNCHRONIZED MEDIA
PictureStory Reader
An immersive reading application that arranges documents, images, and video widgets, including scripted image changes tied to reading position.
A public desktop project exploring how media can follow a narrative.
DATASETS × LOCAL MODELS
LLM Fine-Tuning Collection
Dataset-preparation and Unsloth training scripts, with LoRA adapters and GGUF export for local inference workflows.
Public experimental tooling. Validate current dependencies and model compatibility before using it for an engagement.