From MVP to thousands of presets: the Novus Visualizers enterprise-grade build

Novus Visualizers went live on 2026-04-23 as a deliberately minimal thing: a browser-based editor where you upload a track, the page reads the audio in real time, you pick a visual style, and you export an MP4 or WebM. That was the whole MVP, and shipping it small was the right call. But across the release run to v1.20 on 2026-06-08, it grew into something much closer to a full music-video studio — and it did so without ever uploading your audio to a server. This is the update on how that happened, told as the arc of a product rather than a feature dump.

The constraint that makes it interesting is the same one that shapes the Background Remover: everything runs client-side. The audio analysis, the rendering, the lyric transcription, and the final video encode all happen in your browser. So every capability added between launch and now had to be built to run on the visitor's own machine, which is a much harder bar than calling a render farm — and the reason the exports stay free and copyright-clean.

The v1.0 launch established the core loop. The browser reads an uploaded track through the Web Audio API with a real-time 32-band FFT, detecting beats, onsets, and a loudness envelope, and uses that signal to drive a rendered animation frame by frame. A template gallery gave first-timers a starting point instead of a blank canvas, engine modes provided different visual styles, and the editor could export to MP4 and WebM with intro and outro canvases and resize handles. It was small, but it was complete: a real track went in and a real video came out, with the visuals genuinely synced to the music.

That completeness is what made the expansion possible. Because the audio-to-motion bridge worked from day one, every later release could focus on adding depth — more engines, better sync, richer export — rather than re-litigating the fundamentals. The MVP earned the right to grow by being honest about what it already did well.

The biggest structural change was the engine system. Rendering moved beyond HTML5 Canvas 2D to add WebGL through Three.js, and the engines multiplied: v1.7 added 29 pre-layered showcase templates, v1.9 added ten new engine families, v1.10 introduced the 4D engine system alongside a cinematic suite and LFO automation, and v1.12 added six-plus visualization modes per engine with a variant picker on template detail pages. The result is a very large space — engine families across 2D, 3D, 4D, and Advanced variants, each with multiple mode variants, multiplying out to thousands of distinct preset configurations and 500-plus ready-made templates.

Crucially, this was paired with parity work, not just addition. v1.16 explicitly improved engine parity across 2D/3D/4D/Advanced variants so the engines behaved consistently rather than each being its own special case, and v1.14 was a complete platform overhaul fixing stability and rendering issues. Breadth without consistency is just surface area; the parity passes are what made the breadth usable.

It is worth being concrete about what that breadth contains, because "thousands of presets" can sound like marketing until you see the structure under it. The engines span a wide vocabulary of visual forms — spectrum and waveform, particles, tunnels, terrains, vortexes, comets, constellations, inkflow, smoke, kaleidoscopes, glitch, and crystal among them — and each engine family exists in 2D, 3D, 4D, and Advanced variants, with multiple distinct mode variants inside each. That multiplication is where the large number comes from: it is not one slider relabelled a thousand ways, it is a genuine matrix of form by dimension by variant, plus hundreds of pre-layered showcase templates that assemble those building blocks into finished starting points. A user never has to confront the whole matrix at once — they pick a template or an engine and adjust — but the depth is real, which is what lets very different songs end up with genuinely different-looking videos.

More engines would have meant little if every visual were a single full-screen effect, so the run built real compositional control. A project supports up to four independent layers — a background plus three overlays — each running its own engine, so a visual can stack a calm field behind a reactive spectrum behind a particle accent rather than committing to one look. Text overlays sit on top for titles and credits, intros and outros provide ten motion presets to open and close a piece cleanly, and a set of beat-reactive behaviors — shake, pulse, glow, tilt, and drift — let individual elements respond to the music rather than the whole frame moving as one. Post effects like vignette, film grain, and chromatic aberration add the finishing grade that separates a raw render from something that looks deliberately made.

The piece that turns this from a set of toggles into actual direction is automation. Automation lanes and LFO modulation, introduced with the cinematic suite, let parameters change over time on their own — a value that sweeps across a drop, a glow that breathes through a verse — so the visual has an arc instead of a fixed state that merely twitches on the beat. Combined with the multi-band sync that follows, the result is a composition the creator shapes deliberately: layers chosen, motion assigned, parameters automated across the length of the track. That is the difference between a visualizer that decorates audio and a tool a creator uses to direct a piece of motion to their song.

A visualizer is only as good as how tightly it tracks the music, so a lot of the run went into the audio coupling. v1.19 added multi-band beat sync, letting bass, mid, and treble trigger effects independently rather than the whole frame pulsing on one global beat — the difference between motion that follows a kick drum and motion that follows the actual arrangement. Layers grew too: up to four independent layers, each running its own engine, plus text overlays, intros and outros with motion presets, and beat-reactive shake, pulse, glow, tilt, and drift.

v1.9 added on-device Whisper lyric transcription through Transformers.js — your audio is transcribed to captions in the browser and never leaves it — with per-word timing edits and preset caption styles, and v1.18 added a dedicated lyric-video wizard on top. The same release run added a sound-effects preset library (v1.19), album cover upload, an Album system for grouping visualizers (v1.10), Spotify Canvas vertical export, Stream Overlays with a curated catalog, and a Creator Studio for templates, analytics, and profiles. The tool stopped being just a visualizer and became a small content system around a release.

Somewhere in the run the product stopped being a single tool and became a small content system around a release. An Album system groups related visualizers; an Album Art Editor and album-cover upload handle the cover; an Audio Effects Generator ships a sound-effects preset library spanning synth pads, bass, leads, and percussion; a Lyric Video Creator with a dedicated wizard turns the on-device transcription into a finished lyric piece; and a Stream Overlays system provides a curated catalog — Now Playing, Spectrum, Alerts, Ambient — for creators who broadcast. An Audio Library, nine one-click color themes (Neon, Sunset, Ocean, Midnight, Forest, Gold, Plasma, Rose, Arctic), and a Spotify Canvas vertical export preset round out the kit, so the visual identity of a release can be built once and carried across every surface it touches.

On top of the creative tools sits the operator layer. A Creator Studio handles template, analytics, and profile management; a Community feed lets people share and discover visualizers; and a global search plus paginated browsing make a growing library navigable. The whole thing runs on a lean, edge-compatible stack — Next.js App Router, React, TypeScript, and Zustand on the front, Turso (libSQL) for edge-compatible data, and Web Audio, WebGL through Three.js, WebCodecs, Transformers.js for Whisper, and Canvas 2D doing the actual media work in the browser. The shape of that stack is the point: the heavy lifting — analysis, rendering, transcription, encoding — happens on the user's device, and the server side stays light enough to run cheaply, which is exactly what lets the whole studio stay free and ad-supported.

The reason a companion toolkit matters, rather than just being more things to maintain, is that a music release is never one asset. An artist needs the visualizer, the cover, a lyric video, the short vertical cuts, and — if they stream — the overlays, and the moment those are made in five unrelated tools they stop looking like one release. Keeping them inside one studio, sharing one palette, one set of themes, and one source track, is what lets a solo creator ship a coherent visual identity instead of a scatter of mismatched files. That coherence is itself a kind of professional polish that used to require either a designer or a stack of subscriptions, and folding it into a single free tool is the quiet point of the whole expansion: not just more capability, but a complete release kit one person can run.

The export path is where "enterprise-grade" gets real, because a beautiful preview that cannot be exported reliably is worthless. Export runs through WebCodecs, entirely client-side — no upload, no server render queue — producing MP4 (H.264) and WebM (VP9) up to 4K (3840×2160) at 24, 30, or 60 fps, with platform presets for YouTube, TikTok and Reels, Instagram, Spotify Canvas, X, and Discord. Because it is local and copyright-clean, you own your exports outright, with no attribution required and no watermark.

The most recent release, v1.20 on 2026-06-08, is a reliability and sharing pass in the same spirit as the Background Remover's v1.5.0: it fixed durability for saved albums, templates, and projects; fixed the export progress display so it reads a clean 0–100% instead of going negative during MP4 conversion; improved advanced-engine export performance by eliminating per-frame canvas read-backs; unified audio unlock behind a single gesture; corrected Open Graph and Twitter link previews for shared visualizers; added a native mobile share sheet; and shipped the public changelog itself. That is the through-line of the whole v1.0 → v1.20 arc — a small honest MVP grown into a deep studio, kept free and client-side, and hardened release by release into something you can actually ship a release with.

It is worth dwelling on that last release, because it is the clearest signal of what stage the product has reached. Early releases add capability; mature releases protect it. The durability fix for saved albums, templates, and projects — corrected through a configuration fallback — addresses the worst possible failure for a creator tool, which is losing work you thought was saved. The export-progress fix, taking the bar from a confusing negative percentage during MP4 conversion to a clean 0–100, is the kind of thing that does not add a feature but removes a moment of doubt at the exact point a user is waiting on their finished video. Eliminating per-frame canvas read-backs on the advanced engines is a real performance change that makes the heaviest visuals export faster, and unifying audio unlock behind a single gesture removes a class of "why is there no sound" confusion that browser autoplay rules create.

The sharing half of the release is just as telling. Correcting Open Graph and Twitter link previews so a shared visualizer resolves to the right domain, adding a native mobile share sheet, and shipping a public changelog are all about the work leaving the tool cleanly and being seen. A studio that produces great videos but botches the preview when one is shared has not finished the job; v1.20 finished it. Taken together, the pass is the same move the Background Remover made at v1.5.0 — stop adding surface area for a release and instead harden the surface that exists — and seeing both apps reach that stage at once is the strongest evidence that the ecosystem is past the MVP era and into the keep-it-dependable era.

There is one guarantee here that is rarer than all the rest combined, and it deserves its own section because it is the one an independent artist feels most directly: you own what you make. Many free creative tools quietly take something in exchange for the zero price — a watermark stamped on the corner, a usage license buried in the terms, a resolution cap that nudges you toward a paid tier, a credit line you are not allowed to remove. Novus Visualizers exports are yours outright. They are copyright-clean, carry no watermark, and require no attribution, and you can use them commercially or personally without restriction.

That ownership is not a separate generosity bolted onto the product; it is the same architectural fact as the privacy, viewed from the other side. Because the render happens on your own device through WebCodecs with no upload and no server queue, there is no point in the pipeline where a service inserts itself between you and your video to take a cut, stamp a mark, or assert a claim. The audio that drives the visual and the transcription behind the captions never leave the browser, and the finished file belongs wholly to the person who made it. For a working artist weighing tools, that combination — professional 4K output, full ownership, real privacy, and no cost — is not a feature list to compare line by line; it is the entire reason to choose this one.

The v1.0 to v1.20 arc is a foundation, not a conclusion, and the most useful thing to take from it is the proof it offers rather than the feature count. The product grew from a single reactive editor into a layered studio with thousands of preset configurations, on-device transcription, a companion toolkit, and a reliable 4K export path — and every one of those additions obeyed the constraint the MVP started with: it runs in the browser, on the user's device, and the audio never leaves it. What that demonstrates is that a deep, professional creative tool does not require the centralizing, data-hungry, subscription-shaped architecture the industry treats as inevitable. The same depth can be built on a client-side foundation that keeps the work private and free.

That is the claim worth holding onto, because it generalizes past visualizers. The next addition — another engine family, a deeper automation system, a new companion tool — will have to pass the same test before it ships, which is what keeps the growth coherent rather than letting the studio sprawl into something that betrays its own premise. If you want the mechanics behind any single piece — how the audio becomes motion, how multi-band sync follows the arrangement, how the in-browser WebCodecs export runs without a server — the linked posts cover each in turn, and the tool itself, free and needing no signup to try, is the fastest way to confirm that the studio does what this update describes.