https://newsnix.com/ A daily magazine covering creative technology: AI, creative coding, interactive art, music tech, XR, and the people making it. en-us Thu, 22 Oct 2026 12:40:00 GMT Astro https://newsnix.com/NewsNix400x200.png https://newsnix.com/ https://newsnix.com/articles/local-tts-tutorial-getting-started-narration-voiceover/ https://newsnix.com/articles/local-tts-tutorial-getting-started-narration-voiceover/ Thu, 22 Oct 2026 12:40:00 GMT NewsNix Running a TTS model on your own machine: what the open options actually are, how to write text a model will read well, and where the seams show. Cloud TTS is easy and sends your script to somebody else's server. Local models have got good enough for scratch narration, gallery audio and video voiceover, and you can download the weights. Here is how to choose one, what hardware you need, and the punctuation tricks that fix most bad reads. AI & Creative Tools text-to-speech TTS local AI narration voiceover tutorial open weights https://newsnix.com/articles/drum-machine-individual-outs-tutorial-getting-started-external-processing/ https://newsnix.com/articles/drum-machine-individual-outs-tutorial-getting-started-external-processing/ Thu, 22 Oct 2026 12:25:00 GMT NewsNix Why the stereo pair is costing you the mix, how to split voices to a mixer or interface, and the gain, grounding and latency traps on the way. A drum machine's stereo output is a decision someone else made about your kick. Individual outputs let you compress the snare without ducking the hats, print the analog voices clean, and reamp a single element. Here is how to set it up without a ground loop or a phase problem. Music Technology drum machine routing individual outputs mixing tutorial hardware gain staging https://newsnix.com/articles/warping-workspace-reach-remapping-humanoid-teleoperation-uist/ https://newsnix.com/articles/warping-workspace-reach-remapping-humanoid-teleoperation-uist/ Thu, 22 Oct 2026 12:10:00 GMT NewsNix A panoramic azimuth warp paired with its inverse arm transform, tested two ways on eight people: one mapping completed more trials, the other was preferred by three quarters of them. Rekimoto's lab presents coupled visual–motor remapping for first-person humanoid teleoperation at UIST 2026 in Detroit next month. Global Zoom completed 23 of 24 trials; Focus+Context completed 21 but was preferred by six of eight participants. The authors report the split rather than picking a winner. Emerging Interfaces teleoperation UIST 2026 humanoid Rekimoto remapping arXiv interface design https://newsnix.com/articles/rlmuscle-194-muscle-character-animation-reinforcement-learning/ https://newsnix.com/articles/rlmuscle-194-muscle-character-animation-reinforcement-learning/ Thu, 22 Oct 2026 12:00:00 GMT NewsNix A reinforcement-learning controller finds per-muscle activations from reference motion, an EMG prior steers it toward real human recruitment, and a volumetric solver produces the bulge. In most 3D pipelines artists specify muscle activation by hand — slow, and with no guarantee of physical realism. A Beihang and UCL team trains an RL controller on a 1D Hill-type surrogate, then drives a 3D volumetric simulation with the same constitutive law, hitting full-body simulation at 18 ms per frame on one RTX 4090. Research & Innovation character animation reinforcement learning muscle simulation XPBD EMG arXiv GPU https://newsnix.com/articles/genxr-bench-coding-agents-interactive-xr-missing-repo/ https://newsnix.com/articles/genxr-bench-coding-agents-interactive-xr-missing-repo/ Thu, 22 Oct 2026 11:50:00 GMT NewsNix GenXR-Bench puts 112 expert-authored XR tasks to ten models. Top Pass@1 is 50.9%, the failures cluster around sensor lifecycles and embodiment, and the advertised repository returns 404. An unreviewed preprint from a 16-author team introduces GenXR-Bench, running held-out tests on consumer XR hardware with 96.5% agreement against blinded expert judgement. It calls itself a reproducible testbed. The GitHub URL printed in the paper is not a repository. XR / Spatial Computing XR coding agents benchmark GenXR-Bench arXiv spatial computing reproducibility https://newsnix.com/articles/tomb-raider-raul-siqueira-30-years-3d-level-design/ https://newsnix.com/articles/tomb-raider-raul-siqueira-30-years-3d-level-design/ Thu, 22 Oct 2026 11:40:00 GMT NewsNix Raul Siqueira on why learning the architecture was the point, how camera technology reshaped level construction, and the problem of remaking 1996 for players raised on objective markers. Crystal Dynamics is rebuilding Lara Croft's first adventure as Legacy of Atlantis. Game Director Raul Siqueira told 80.lv that the team's answer to the guidance problem is to let players decide how much assistance they receive — and that not one of the original six developers is on the project. Digital Artists Tomb Raider level design Crystal Dynamics retro game development Raul Siqueira 3D https://newsnix.com/articles/recreate-apollo-dsky-raspberry-pi-virtualagc/ https://newsnix.com/articles/recreate-apollo-dsky-raspberry-pi-virtualagc/ Thu, 22 Oct 2026 11:30:00 GMT NewsNix Michael Newton's Recreate Apollo project runs the Apollo 13 Command Module software on yaAGC behind a 3D-printed Display and Keyboard — GPL-3.0, 30 stars, and faithful down to the 1201/1202 alarms. The MIT Instrumentation Lab's engineering drawings were released under a public records request and archived by the Virtual AGC group. A Calgary avionics engineer turned them into printable CAD, then put a Raspberry Pi behind it running an emulator so exact that he argues it is not an emulator at all. Creative Hardware Apollo Guidance Computer Raspberry Pi retro VirtualAGC 3D printing preservation open source https://newsnix.com/articles/roland-tr-1000-firmware-1-50-keyboard-mode-analog-routing/ https://newsnix.com/articles/roland-tr-1000-firmware-1-50-keyboard-mode-analog-routing/ Thu, 22 Oct 2026 11:20:00 GMT NewsNix Step entry without motion recording, the step buttons playable as a keyboard, new MFX, and a routing option that sends every analog element out through AFX. Roland posted its own 1.50 walkthrough on 21 October. The two changes that alter how the machine fits a studio are keyboard mode on the step buttons and a new analog routing that lets you record the analog voices with nothing digital in the path — plus USB audio routed to the individual outs, which turns the TR-1000 into an interface. Music Technology Roland TR-1000 drum machine firmware analog routing Rhythm Creator https://newsnix.com/articles/mistral-voxtral-3-api-only-no-open-weights/ https://newsnix.com/articles/mistral-voxtral-3-api-only-no-open-weights/ Thu, 22 Oct 2026 11:10:00 GMT NewsNix Three transcription models, a TTS model and a LiveKit-based voice agent SDK — all in public preview on the API, with no weights on Hugging Face. Mistral published open Voxtral weights as recently as 8 October. Voxtral 3, announced on 20 October, ships as an API preview and self-deployment-via-sales. There is no Voxtral 3 repository on Hugging Face, and the announcement never uses the word 'weights'. AI & Creative Tools Mistral Voxtral speech-to-text TTS open weights voice cloning LiveKit https://newsnix.com/articles/pollinator-pathmaker-berlin-ginsberg-algorithm-not-ai/ https://newsnix.com/articles/pollinator-pathmaker-berlin-ginsberg-algorithm-not-ai/ Thu, 22 Oct 2026 11:00:00 GMT NewsNix Alexandra Daisy Ginsberg's Pollinator Pathmaker now has an edition in every district of Berlin, and a wild-bee count well above the city's pollinator-friendly average. LAS Art Foundation curator Sophie Korschildgen and artist Alexandra Daisy Ginsberg spoke to designboom about the project's first international commission. The number that matters: a wild-bee specialist recorded 83 species at the Natural History Museum edition last year, against a 30–50 average for areas of Berlin classed as pollinator-friendly. Interactive Art Alexandra Daisy Ginsberg Pollinator Pathmaker LAS Art Foundation Berlin generative art ecology algorithmic art https://newsnix.com/articles/uist-2026-detroit-preview-maes-shneiderman-debate/ https://newsnix.com/articles/uist-2026-detroit-preview-maes-shneiderman-debate/ Thu, 22 Oct 2026 10:45:00 GMT NewsNix Eleven workshops, a record 1,259 submissions at a 21.4% acceptance rate, and a keynote debate thirty years in the making. The ACM Symposium on User Interface Software and Technology runs 2–5 November at the Detroit Marriott at the Renaissance Center. Pattie Maes and Ben Shneiderman return to the direct-manipulation-versus-interface-agents debate they helped define in the late 1990s — a question that reads very differently now that the agents actually exist. Community UIST UIST 2026 conference HCI Detroit Pattie Maes Ben Shneiderman interface agents https://newsnix.com/articles/processing-foundation-maintenance-fund-endowment-200k/ https://newsnix.com/articles/processing-foundation-maintenance-fund-endowment-200k/ Thu, 22 Oct 2026 10:30:00 GMT NewsNix An endowment that yields $200,000 annually, split evenly between Processing Java and p5.js, with two community committees deciding how it is spent. Twenty-five years after Casey Reas and Ben Fry started Processing at the MIT Media Lab, the Processing Foundation has endowed its upkeep. The Maintenance Fund pays for the unglamorous work — keeping sketches online, keeping the reference current, keeping Processing Java running on new operating systems — and 60% of the Project Leads' compensation. Creative Coding Processing p5.js Processing Foundation open source funding maintenance Casey Reas https://newsnix.com/articles/tod-machover-already-and-not-yet-bso-live-ai-symphony/ https://newsnix.com/articles/tod-machover-already-and-not-yet-bso-live-ai-symphony/ Wed, 21 Oct 2026 12:25:00 GMT NewsNix already and not yet premieres tonight at Symphony Hall — three movements on democracy, with an AI that first isolates every player and then goes looking for common ground. Machover's new BSO commission runs 22–24 October under Kent Nagano. A Live AI system built at the MIT Media Lab listens to the orchestra during the performance: in the second movement it pulls each instrumentalist's sound apart into an ensemble of soloists, and in the third it tries to connect them, pushing sound out into the hall through loudspeakers installed for the piece. Music Technology Tod Machover MIT Media Lab Boston Symphony Orchestra AI composition live electronics Kent Nagano https://newsnix.com/articles/worlds-upon-the-wind-ukiyo-e-godot-public-domain-deckbuilder/ https://newsnix.com/articles/worlds-upon-the-wind-ukiyo-e-godot-public-domain-deckbuilder/ Wed, 21 Oct 2026 12:05:00 GMT NewsNix Worlds Upon The Wind remixes public-domain ukiyo-e in Godot, draws 200-plus structures by hand to match, and releases the whole thing under CC0. Max Shawabkeh spent 18 months on a non-violent deckbuilder about restoring a shattered world. Faced with the art problem every solo developer hits, he chose to remix public-domain historical prints rather than generate anything — cutting his procedural map out of a 1600s travel guide by Hishikawa Moronobu. Digital Artists Godot ukiyo-e public domain indie games procedural generation shaders CC0 retro https://newsnix.com/articles/esp-mosaico-esp32-s31-modular-dev-kit-espressif/ https://newsnix.com/articles/esp-mosaico-esp32-s31-modular-dev-kit-espressif/ Wed, 21 Oct 2026 11:45:00 GMT NewsNix A 480×480 AMOLED, haptics, battery, two Bosch magnetometers and a new dual-core RISC-V ESP32-S31 — and the vendor's own documentation never mentions AI once. ESP-Mosaico splits into a CoreBoard and a BaseBoard over a 60-pin connector, with a square touch AMOLED, an ES8311 codec, an IMU and dual magnetometers. Coverage has framed it as a platform for AI-assisted hardware creation. Espressif's own documentation calls it an expandable smart interaction development kit and says nothing about AI at all. Creative Hardware Espressif ESP32-S31 ESP-Mosaico RISC-V dev kit AMOLED hardware https://newsnix.com/articles/live-camera-performance-tutorial-getting-started-stage/ https://newsnix.com/articles/live-camera-performance-tutorial-getting-started-stage/ Wed, 21 Oct 2026 11:15:00 GMT NewsNix The screen always wins. Here is how to put cameras in a performance without accidentally making a film with actors standing underneath it. A close-up beats a body at forty metres, every time. That one fact decides whether live video in a performance reads as a second layer or as a distraction. This covers the camera positions that work, why operators belong inside the set, latency budgets, keeping a projector and a stage light from fighting, and the rehearsal order that saves the tech week. Interactive Art live video performance theatre cameras projection staging tutorial https://newsnix.com/articles/godot-tutorial-getting-started-creative-coding/ https://newsnix.com/articles/godot-tutorial-getting-started-creative-coding/ Wed, 21 Oct 2026 10:45:00 GMT NewsNix You do not need the game engine parts. Here is the minimum Godot — one node, one script, one shader — that makes it a better sketching environment than it looks. Godot is usually introduced as a game engine, which buries what makes it useful for generative work: a real scene tree, built-in shaders, a fast export to desktop and web, and the ability to drop to C++ for the one subsystem that needs it. This walks through the smallest viable setup and the four things that genuinely differ from p5.js. Creative Coding Godot GDScript shaders creative coding generative art tutorial https://newsnix.com/articles/dancing-diffusion-continuous-programs-sana-conditioning/ https://newsnix.com/articles/dancing-diffusion-continuous-programs-sana-conditioning/ Wed, 21 Oct 2026 10:10:00 GMT NewsNix Instead of conditioning every denoising step on the same instruction, each span of timesteps gets its own learned condition block — so you can rewrite a later operation and keep what the earlier ones did. A diffusion model's trajectory is a sequence of steps that all receive the same prompt, which makes individual operations impossible to target. This preprint compiles text into a latent program of per-interval condition blocks, runs it on a frozen SANA, and reuses the executed prefix instead of regenerating from noise. Creative Coding diffusion generative AI research image editing SANA conditioning programming https://newsnix.com/articles/phonobridge-emg-silent-speech-phoneme-llm/ https://newsnix.com/articles/phonobridge-emg-silent-speech-phoneme-llm/ Tue, 20 Oct 2026 22:50:00 GMT NewsNix The trick is not better sensing — it is using phonemes as an intermediate layer, so LibriSpeech transcripts can train the decoder without a single new EMG recording. Silent-speech interfaces are starved of data: every hour of training requires someone wearing electrodes and mouthing words. PhonoBridge routes EMG through phonemes first, which lets a language model learn from ordinary speech transcripts instead — and cuts word error rate from 28.27% to 24.77%. Emerging Interfaces EMG silent speech interfaces research phonemes accessibility wearable https://newsnix.com/articles/caesar-mixed-reality-cognitive-attack-ismar-2026/ https://newsnix.com/articles/caesar-mixed-reality-cognitive-attack-ismar-2026/ Tue, 20 Oct 2026 21:45:00 GMT NewsNix An ISMAR 2026 testbed for collaborative mixed reality that builds four classes of perceptual attack into a shared search-and-rescue scenario — and measures what added latency does to a team. If a team coordinates through virtual content, distorting that content distorts the team. CAESAR is a multi-user MR framework for evaluating those attacks systematically rather than one at a time, and its first study takes a latency attack apart in detail. XR / Spatial Computing mixed reality security ISMAR 2026 research collaboration latency XR https://newsnix.com/articles/agent-micro-waits-sidecar-too-long-to-watch/ https://newsnix.com/articles/agent-micro-waits-sidecar-too-long-to-watch/ Tue, 20 Oct 2026 20:40:00 GMT NewsNix Google researchers name the dead minute while an AI agent works, and build a second assistant to fill it by reading the agent's run over your shoulder. An agent takes two minutes on one turn. Watching is tedious; switching away costs you the context you need to resume. Twelve interviewed users described both horns of that. The authors' answer, Sidecar, sits in the same interface and tells you which decisions the agent quietly made on your behalf. Research & Innovation AI agents interaction design research attention waiting Google https://newsnix.com/articles/eleven-realtime-v1-max-mini-speech-to-speech/ https://newsnix.com/articles/eleven-realtime-v1-max-mini-speech-to-speech/ Tue, 20 Oct 2026 19:30:00 GMT NewsNix Eleven Realtime v1 Max and Mini are speech-to-speech rather than text-to-speech — and Mini can tell the difference between you finishing a sentence and you saying 'mm-hmm'. A 0.57-second response time and 98.8% on Big Bench Audio puts Eleven Realtime v1 Max at the fast-and-accurate corner of the speech-to-speech leaderboard. For anyone building a talking installation, the backchannel detection in the cheaper Mini model is the more interesting feature. AI & Creative Tools ElevenLabs speech-to-speech voice AI latency real-time audio https://newsnix.com/articles/bob-moog-foundation-2026-online-auction-closes-26-october/ https://newsnix.com/articles/bob-moog-foundation-2026-online-auction-closes-26-october/ Tue, 20 Oct 2026 18:20:00 GMT NewsNix A Keith Emerson-signed caricature, Dorothy's Minimoog prints, and an acrylic painted live at the Moogseum on Bob Moog's 92nd birthday — bidding ends 26 October, 12 p.m. ET. The Foundation is running an online auction of synthesizer art and memorabilia through its own Shopify store, with proceeds funding Dr. Bob's SoundSchool, the Archives and the Moogseum. Bidding opened 19 October and closes Monday 26 October at noon Eastern. Some oversized lots are Asheville pickup only. Community Bob Moog Foundation Moogseum auction open call synthesizer Asheville archive https://newsnix.com/articles/lukasz-twarkowski-oracle-onassis-stegi-turing-thermal/ https://newsnix.com/articles/lukasz-twarkowski-oracle-onassis-stegi-turing-thermal/ Tue, 20 Oct 2026 17:15:00 GMT NewsNix Łukasz Twarkowski's Turing piece films its own actors live from inside a modular set, and turns them into thermal-camera colour fields on the screen above. At Onassis Stegi in Athens, ORACLE splits the stage horizontally: live action below, live footage above, shot by operators standing in the set. Thermal and infrared imaging renders the performers as red, blue and green figures — a surveillance aesthetic applied to a play about Alan Turing. Interactive Art Łukasz Twarkowski Onassis Stegi theatre live video Alan Turing thermal imaging Athens https://newsnix.com/articles/roland-n-vision-boutique-wavetable-synthesizer/ https://newsnix.com/articles/roland-n-vision-boutique-wavetable-synthesizer/ Tue, 20 Oct 2026 09:55:00 GMT NewsNix Bi-timbral, up to 56 voices, two wavetable layers per part and a new wave shaper — in a line that has spent a decade reissuing its own back catalogue. Every Roland Boutique except the SE-02 has been a miniature of a vintage Roland. The n/vision is the second original design in the series and the first wavetable instrument in it: two layers per part, 56 voices, a 64-step polyphonic sequencer, and the K-25m keyboard included at $549.99. Music Technology Roland wavetable synthesizer Boutique hardware n/vision https://newsnix.com/articles/mituhashi-edo-kiriko-vrchat-glass-cutting-world/ https://newsnix.com/articles/mituhashi-edo-kiriko-vrchat-glass-cutting-world/ Tue, 20 Oct 2026 09:25:00 GMT NewsNix Mituhashi Glass Crafts gives you a rotating cutting wheel, coloured glass that turns transparent where you cut it, and an export button that hands back your piece as code. Edo Kiriko is cut by hand against a spinning abrasive wheel, and a cut is permanent. Mituhashi — a 3D artist who makes it as a hobby — rebuilt the craft in VRChat with adjustable wheel angle and depth, optional cut assist, and pieces that export as code or as 3D objects for your own world. Digital Artists VRChat Edo Kiriko Japan craft Unity virtual worlds artist spotlight https://newsnix.com/articles/reflashing-smart-devices-tutorial-getting-started-esp32-ota/ https://newsnix.com/articles/reflashing-smart-devices-tutorial-getting-started-esp32-ota/ Tue, 20 Oct 2026 08:50:00 GMT NewsNix How to identify what chip is actually inside, get your own firmware on without opening a mains enclosure, and leave yourself a way back. A smart bulb or plug is a microcontroller you already own, sold below the price of the parts. This covers how to tell an ESP32 from an ESP8285 before you buy, the over-the-air route that avoids opening a mains device, how to keep a rollback path, and the four ways people brick these things. Creative Hardware ESP32 firmware OTA reflashing smart home safety tutorial https://newsnix.com/articles/substance-3d-designer-tutorial-getting-started-nodes/ https://newsnix.com/articles/substance-3d-designer-tutorial-getting-started-nodes/ Tue, 20 Oct 2026 08:20:00 GMT NewsNix A procedural material is a small program that outputs textures. Here is the mental model, the six nodes that do most of the work, and the habits that stop a graph becoming unreadable. Substance 3D Designer is not a painting tool with extra steps — it is a node-based program that emits texture maps, and people bounce off it because they approach it as the former. This walks through the graph model, the handful of nodes you actually need, how to wire a basic material end to end, and why you should resolve everything to grayscale before you touch colour. Digital Artists Substance 3D Designer procedural texturing PBR materials nodes tutorial https://newsnix.com/articles/kling-4-0-stereo-audio-multilingual-lip-sync-flash/ https://newsnix.com/articles/kling-4-0-stereo-audio-multilingual-lip-sync-flash/ Mon, 19 Oct 2026 21:05:00 GMT NewsNix Two-channel audio, lip sync in more than ten languages, up to ten keyframes and an 8,000-token prompt window — against rivals it no longer clearly beats. Kling 4.0 shipped on 19 October with the features a video model needs to be used in production: stereo audio, multilingual lip sync, 30-second single-pass generation, ten keyframes, MCP and CLI access. The more interesting detail is that the side-by-sides against Wan 3.0 and Seedance 2.5 are a draw. AI & Creative Tools Kling AI video generative video lip sync audio text-to-video https://newsnix.com/articles/etherbulb-esp32-c3-smart-bulb-rf-spectrum-receiver/ https://newsnix.com/articles/etherbulb-esp32-c3-smart-bulb-rf-spectrum-receiver/ Mon, 19 Oct 2026 20:30:00 GMT NewsNix Peter Holzhauser's GPL firmware uses an ESP32-C3 bulb's Wi-Fi radio as a 2.13–2.73 GHz receiver: blue for Bluetooth, green for Wi-Fi, brighter when the signal is stronger. A smart bulb already contains a 2.4 GHz radio and a light. Etherbulb rewires the relationship between them — the bulb listens to the band, classifies what it hears, and shows it as colour. No app, no cloud, no transmitting, and a hardware interlock so the RGB and white channels can never run together. Creative Hardware ESP32 SDR radio firmware open source ambient display RF https://newsnix.com/articles/splatxr-unity-toolkit-on-headset-gaussian-splatting/ https://newsnix.com/articles/splatxr-unity-toolkit-on-headset-gaussian-splatting/ Mon, 19 Oct 2026 19:55:00 GMT NewsNix A Princeton Unity toolkit with splat physics, splat-to-mesh reconstruction and portal-style blending — announced as MIT-licensed open source, with no code published yet. XR pipelines are built on polygons, so splats arrive without the surfaces that shadows and collisions need. SplatXR, at UIST 2026, supplies the missing building blocks — collision meshes generated from splats, shader effects, portal blending, on-headset capture. The repository it promises does not exist yet. XR / Spatial Computing gaussian splatting XR Unity mixed reality research UIST 2026 open source https://newsnix.com/articles/open-form-geometric-style-transfer-lora-disentangle/ https://newsnix.com/articles/open-form-geometric-style-transfer-lora-disentangle/ Mon, 19 Oct 2026 19:20:00 GMT NewsNix Open-Form Geometric Style Transfer splits structure from style across two LoRAs, so a stylised Latin alphabet's design language can be applied to shapes it has never seen. Geometric style transfer usually only works inside the shape family it learned on — fonts to fonts, icons to icons. A new preprint from Daniel Cohen-Or's group separates the two concerns into a universal Geometry LoRA and a style-specific one, and tests it by pushing alphabet styles onto Greek glyphs and human poses. Creative Coding style transfer typography LoRA research generative AI graphics diffusion https://newsnix.com/articles/beyond-explicit-intent-artistic-intent-rl-co-creation/ https://newsnix.com/articles/beyond-explicit-intent-artistic-intent-rl-co-creation/ Mon, 19 Oct 2026 18:45:00 GMT NewsNix Beyond Explicit Intent argues artistic intent is latent until you see options — and then feeds your choosing back into the model as a reward signal. Every text-to-image tool assumes you can say what you want before you see anything. A new preprint builds the opposite loop: widen exploration, let the artist pick, treat the picking as reinforcement, and let the model drift toward one person's taste. Three user studies, and one structural risk the paper does not name. Research & Innovation generative AI research reinforcement learning creative process human-AI collaboration text-to-image https://newsnix.com/articles/trinketry-trinkets-traceable-design-histories/ https://newsnix.com/articles/trinketry-trinkets-traceable-design-histories/ Mon, 19 Oct 2026 18:10:00 GMT NewsNix A canvas where the saved unit is an element rather than a generation — and where those elements double as search queries and as anchors back into your own process. Generative design tools give you branching trees of whole images, which turn out to be nearly useless for reconstructing how an idea developed. Trinketry saves fragments instead. In two studies with 18 concept designers, element-level history beat whole-image history at recovering turning points, goals and overlooked material. Emerging Interfaces design tools generative AI interaction design research creative process canvas https://newsnix.com/articles/museum-ai-situated-generation-four-wave-study-iss2026/ https://newsnix.com/articles/museum-ai-situated-generation-four-wave-study-iss2026/ Mon, 19 Oct 2026 17:35:00 GMT NewsNix Nineteen Chinese science-museum practitioners, interviewed annually from 2023 to 2026, describe content that is made during the visit rather than before it. An ACM ISS 2026 paper follows 23 museum practitioners through four annual waves of interviews and names the shift it found: from prefabricated display to situated generation, where visitor-facing content is produced during the interaction. The four design propositions it offers are mostly about restraint. Interactive Art museum AI exhibition design research ISS 2026 interaction design longitudinal study https://newsnix.com/articles/elektra-biennial-friction-art-games-montreal/ https://newsnix.com/articles/elektra-biennial-friction-art-games-montreal/ Mon, 19 Oct 2026 17:05:00 GMT NewsNix F(R)ICTION: Art and Games fills Arsenal Contemporary with about forty artists and 35 playable works — more than a hundred hours of gameplay, hung as art. The 8th ELEKTRA Biennial runs to 15 November at Arsenal Contemporary Art in Montréal. Curators Lynn Hughes and Ida Toft have assembled roughly forty artists and 35 game artworks, and built the ticket around the one thing a games exhibition actually needs: unlimited free re-entry. Community ELEKTRA BIAN Montreal games biennial exhibition digital art retro https://newsnix.com/articles/image-model-evaluation-tutorial-getting-started-geneval-dpg/ https://newsnix.com/articles/image-model-evaluation-tutorial-getting-started-geneval-dpg/ Mon, 19 Oct 2026 16:45:00 GMT NewsNix What GenEval, DPG-Bench, FID and CLIPScore each actually measure — and why a model can win three of them and still be worse. Benchmark numbers in model cards are easy to read and easy to misread. This walks through what the standard text-to-image metrics measure, what each one is blind to, how to run a comparison that is actually fair, and how to build a small private eval set that tells you more about your own work than any public leaderboard. Research & Innovation evaluation benchmarks GenEval FID text-to-image diffusion tutorial https://newsnix.com/articles/installation-documentation-tutorial-getting-started-festival-submissions/ https://newsnix.com/articles/installation-documentation-tutorial-getting-started-festival-submissions/ Mon, 19 Oct 2026 16:35:00 GMT NewsNix Juries watch two minutes of your three-minute video. Here is how to build documentation that survives that. A festival jury meets your installation as a video file, three stills and 1,000 characters. This walks through what to shoot, how to film a dark room without ruining it, how to show interaction honestly, and the technical rider that decides whether a selected work actually gets installed. Community documentation festival open call installation video tutorial https://newsnix.com/articles/newimages-festival-2027-open-calls-xr-fulldome-virtual-worlds/ https://newsnix.com/articles/newimages-festival-2027-open-calls-xr-fulldome-virtual-worlds/ Mon, 19 Oct 2026 16:15:00 GMT NewsNix Three competitions, two deadlines, and eligibility windows that differ by three years depending on which one you enter. NewImages Festival's 10th edition runs 21–25 April 2027 in Paris, and its open calls are live until 22 December 2026, with an Early Bird cut at 10 December. XR and Fulldome cost €75 early or €100 late; Virtual Worlds costs nothing. The eligibility rules are where entrants get caught — XR wants work completed in 2026 or 2027, Fulldome accepts anything after 2023. Community NewImages Festival open call XR fulldome virtual worlds Paris festival https://newsnix.com/articles/aarav-sharma-ssvep-bci-six-target-speller-ar-glasses/ https://newsnix.com/articles/aarav-sharma-ssvep-bci-six-target-speller-ar-glasses/ Mon, 19 Oct 2026 15:55:00 GMT NewsNix Aarav Sharma moved his brain-controlled robot arm project into communication — six flashing targets, a word-suggestion model, and AR glasses next. Flicker six targets at six different frequencies, read which one the visual cortex locks onto, and you have six commands without a single muscle movement. Aarav Sharma — who started on a 3D-printed brain-controlled arm at 14 — has turned that into a speller, with a language model proposing words and the user keeping the final say. Reported response times are under five seconds for a full sentence. Emerging Interfaces BCI SSVEP EEG OpenBCI accessibility AR young makers https://newsnix.com/articles/sensory-comfort-multisensory-xr-autistic-adults-iss2026/ https://newsnix.com/articles/sensory-comfort-multisensory-xr-autistic-adults-iss2026/ Mon, 19 Oct 2026 15:35:00 GMT NewsNix Twenty autistic adults rated five complete multisensory configurations of the same content. The most stimulating one was not the least comfortable. Sensory-friendly design usually means reducing input. An ACM ISS 2026 study ran the same popular-science content through five full XR configurations for 20 autistic adults and found comfort ranged from 2.97 to 4.43 out of 5 — with discomfort tracking abrupt transitions, simultaneous demands and an inability to disengage, rather than the amount of stimulation. XR / Spatial Computing XR accessibility autism multisensory ACM ISS user study https://newsnix.com/articles/hooktheory-explainable-harmonic-decisions-frozen-model-corpus/ https://newsnix.com/articles/hooktheory-explainable-harmonic-decisions-frozen-model-corpus/ Mon, 19 Oct 2026 15:15:00 GMT NewsNix Running an explainable harmony model unchanged against HookTheory annotations separates what it cannot represent from what it simply did not pick. Josef Pavlíček and Martin Pelikán took an explainable melody harmoniser built for single songs, froze its rules, and ran it against 19 HookTheory excerpts — 439 harmonic segments of human-annotated chords. Exact agreement was 20.5%. The useful number is the split underneath it: in 46.24% of segments the human chord was not even in the model's candidate set. Research & Innovation music information retrieval explainable AI harmony HookTheory evaluation arXiv https://newsnix.com/articles/open-embedded-probe-oep-spec-interoperable-debug-probes/ https://newsnix.com/articles/open-embedded-probe-oep-spec-interoperable-debug-probes/ Mon, 19 Oct 2026 14:55:00 GMT NewsNix Open Embedded Probe has 660 commits, seven repositories, a COBS framing layer and a pid.codes VID:PID — and zero stars. Open Embedded Probe specifies how debug probes and host software can interoperate without knowing about each other in advance: a two-layer protocol, CRC-16 COBS frames on serial ports, length-prefixed frames on USB bulk and TCP, and discovery through confirm/list/describe. The whole ecosystem is a month old, and until v1 freezes the authoritative text is Japanese. Creative Hardware Open Embedded Probe debug probe open standard USB embedded open source Arduino https://newsnix.com/articles/iris-3b-pixel-space-diffusion-no-vae-apache/ https://newsnix.com/articles/iris-3b-pixel-space-diffusion-no-vae-apache/ Mon, 19 Oct 2026 14:35:00 GMT NewsNix A 3B pixel-space text-to-image model under Apache 2.0, with a 4-step turbo, depth and 4× upscaling heads, and an unusually honest negative result. SperidLabs' Iris-3B generates pixels directly: no VAE, no latent space, 665K training steps across a 256→512→1024 curriculum. The model card leads with that. The accompanying paper reports that pixel-space pretraining gave no significant benefit on the downstream tasks it was supposed to help — and says so in the abstract. AI & Creative Tools Iris-3B SperidLabs diffusion text-to-image open weights Apache 2.0 pixel space https://newsnix.com/articles/mutracer-webgpu-path-tracer-openpbr-three-js/ https://newsnix.com/articles/mutracer-webgpu-path-tracer-openpbr-three-js/ Mon, 19 Oct 2026 14:15:00 GMT NewsNix WebGPU first with a WebGL 2 fallback, a two-level BVH built across workers, and 4.6 million triangles denoised in about 12 seconds. MuTracer is eleven days old, MIT-licensed, and already at v0.1.3. It renders physically based path tracing in the browser with OpenPBR 1.1 surfaces — coat, thin-film interference, dispersive transmission, subsurface scattering — and ships a Three.js sync layer plus Intel Open Image Denoise running on WebGPU. The README publishes its own numbers. Creative Coding WebGPU path tracing rendering Three.js OpenPBR open source JavaScript https://newsnix.com/articles/shear-electronics-relic-eight-analog-voice-cores-production/ https://newsnix.com/articles/shear-electronics-relic-eight-analog-voice-cores-production/ Mon, 19 Oct 2026 13:55:00 GMT NewsNix Two SEM-topology VCOs and seven discrete VCAs per voice, a four-parameter supervariable filter, and a design that has been in gestation since 2017. The Relic gives each of its eight voices its own analog signal path and its own digital processor: 16 SEM-topology oscillators in total, seven LM13700-based VCAs per voice, and a filter with cutoff, resonance, bandwidth and profile as independent parameters. Shear Electronics says the first production run is under way — a run that follows a prototype first shown in 2017. Music Technology Shear Electronics analog synthesizer polysynth Oberheim SEM hardware retro https://newsnix.com/articles/armon-naeini-entroplay-glitch-motion-interactive-artist-spotlight/ https://newsnix.com/articles/armon-naeini-entroplay-glitch-motion-interactive-artist-spotlight/ Mon, 19 Oct 2026 13:35:00 GMT NewsNix The artist who works as Entroplay treats the camera as a feedback loop, not a sensor — and writes the code that becomes the texture. Armon Naeini makes interactive installations in which a depth camera watches you and hands back a data-moshed version of your own movement. Across Resonance, Self at CPR, ID Pt. III at The Music Center and the 270-degree Entrome, the throughline is self-reflexivity made physical — and in Entrome, the visual texture is literally his own JavaScript source. Digital Artists Armon Naeini Entroplay glitch art TouchDesigner interactive installation motion tracking creative coding https://newsnix.com/articles/refik-anadol-winds-of-hong-kong-mplus-facade-data-painting/ https://newsnix.com/articles/refik-anadol-winds-of-hong-kong-mplus-facade-data-painting/ Mon, 19 Oct 2026 13:15:00 GMT NewsNix Winds of Hong Kong runs nightly until 7 January on one of the world's largest media screens, built from publicly available wind and environmental data. M+ has turned its south face over to Refik Anadol until 7 January 2027. Winds of Hong Kong reads publicly available wind and environmental data from the city and renders it as an evolving data painting across thousands of LEDs — and the museum is framing it explicitly against the energy cost of generative AI. Interactive Art Refik Anadol M+ data art projection Hong Kong generative art LED https://newsnix.com/articles/quantization-tutorial-running-image-models-low-vram/ https://newsnix.com/articles/quantization-tutorial-running-image-models-low-vram/ Mon, 19 Oct 2026 12:24:00 GMT NewsNix What fp8, int8, GGUF Q-levels and Hadamard rotations actually do to your weights, and which trade you are making when you pick one. Every current image model ships in several sizes, and the labels — fp8_e4m3fn, Q4_K_M, int8 ConvRot — look like they need a maths degree. They do not. Here is what each format does to the numbers, why outlier channels are the thing that breaks naive quantisation, and how to pick a build for the VRAM you have. AI & Creative Tools quantization ComfyUI GGUF VRAM diffusion fp8 local AI tutorial https://newsnix.com/articles/fm-synthesis-tutorial-getting-started-operators-ratios/ https://newsnix.com/articles/fm-synthesis-tutorial-getting-started-operators-ratios/ Mon, 19 Oct 2026 12:16:00 GMT NewsNix Subtractive synthesis removes harmonics from a rich wave. FM builds them out of nothing, and the whole instrument is governed by two numbers. FM synthesis has a reputation for being unintuitive, which is mostly a labelling problem — the two controls that matter are the frequency ratio between operators and the amount of modulation. Learn what those actually do to the spectrum, why integer ratios sound musical and non-integer ones sound metallic, and how to get a bass, a bell and an electric piano from the same patch. Music Technology FM synthesis synthesis sound design DX7 operators tutorial