VidNexusVidNexus
>> Join Waitlist
Public release log · 26 releases

Changelog.

Every shipped version since the first commit on April 10, 2025. No marketing fluff - just what changed, when. Lifetime updates means every entry below stays free for you, forever.

v1.0.0-rc.18
2026-10-30

Release candidate 18: public-domain archives and your own footage

Latest
  • FeatureFootage from public-domain archives: Internet Archive (Prelinger and FedFlix), Wikimedia Commons, NASA and DVIDS (with your own free key), searched next to Pexels, Pixabay and Coverr
  • FeatureLong archive films: VidNexus finds the right minute for the scene and downloads only that part
  • FeatureMy footage: clips you bought or downloaded (Storyblocks, Artlist, Envato) go in one folder and are matched to scenes automatically
  • ImprovementOnly public domain, CC0 and CC BY clips are used (CC BY-SA only if you allow it), and their credits are added to the YouTube description
v1.0.0-rc.17
2026-10-28

Release candidate 17: AI models from Settings

  • FeatureSettings > Compute downloads the recommended AI models with live progress, no terminal needed
  • ImprovementThe Lab's Rewrite with suggestions runs on Claude with your Anthropic key, because local models' rewrites rarely passed the check
  • ImprovementUS English spelling throughout the app
v1.0.0-rc.16
2026-10-27

Release candidates 12 to 16: voice edits, retries and the installer

  • ImprovementRegen voice shows its progress in the editor and picks up the new voice and its cost
  • FixThe live preview plays a new voice straight away, without stutter
  • ImprovementThe speech-speed sliders are gone: ElevenLabs barely changed the pace, so every voice reads at its natural speed
  • PerformanceA scene's voice is reused wherever its file is, so a moved folder never pays for the same voiceover twice
  • FixRetry repeats a Render as a Render instead of a full build
  • FixThe build-status popover stays readable over the video, and the installer keeps every file motion graphics need
  • FixOne footage download that fails no longer fails the whole video
  • FixCaptions stay off while a full-screen motion graphic is on screen, Shorts keep their length, and titles of topic videos stay on the topic
v1.0.0-rc.11
2026-10-26

Release candidates 10 and 11: the known defects fixed, footage that fits the era

  • ImprovementHistory videos get period footage: the search wording follows the years in the script, modern-looking clips (screens, cars, skylines, banknotes) rank lower, and scenes that name a place or modern finance get a second search for period footage
  • FixProjects keep working after the assets or renders folder moves, a backup is restored or you switch PCs: file paths re-link to the same files in today's folders
  • FixFull-screen motion graphics render over the footage, like the editor's preview, instead of on a black card
  • FixVideo titles only pair a number with the names it belongs to in the script
  • ImprovementWhen Windows Smart App Control blocks the footage-matching model, the build carries on in search order and Settings explains why, once
  • PerformanceMotion-graphic files nothing uses any more are removed a day later (Undo still works)
  • FixStat callouts and headlines keep their look on any theme; tweet and phone-conversation graphics no longer fail to render
  • FixHighlighted text wraps at spaces and never starts a line with a comma; the real Playfair Display italic is bundled
  • ImprovementVideo settings can change the graphics blocklist for one video; the license check explains itself in plain words when it can't reach the server
v1.0.0-rc.9
2026-10-25

Release candidates 5 to 9: the full test pass, our own FFmpeg, live map previews

  • FixFixes from a 179-case test pass: renders with background music no longer hang; brand-kit LUTs render on Windows and keep the right colours on AMD graphics; beat detection is no longer about 70 ms early; Undo no longer reverts later saves; deleting a video mid-build stops its voiceover
  • PerformanceOur own FFmpeg build: the install is about 370 MB smaller, and its full source is published at releases.vidnexus.app/source
  • FixCaption styles in Montserrat, Inter or Roboto no longer fall back to Arial: the fonts are bundled
  • FeatureMap graphics show the real map in the editor's live preview, with the Mapbox and OpenStreetMap credit on screen
  • FeatureSettings shows update progress, and a VidNexus window says what is happening while an update installs
  • FeatureLicense holders download the installer from their account page
  • ImprovementClear messages when the AI model or the app's engine can't be reached, instead of raw errors
  • ImprovementLicense notices for all 572 bundled packages ship with the app
v1.0.0-rc.4
2026-10-23

Release candidates 2-4: automatic updates and a much smaller install

  • FeatureAutomatic updates: VidNexus checks for a new version when it starts and every hour, downloads it in the background and installs it when you restart
  • FeatureSettings > General shows the installed version with a Check for updates button
  • FixUpdating or reinstalling over an existing copy no longer stops with "VidNexus cannot be closed"
  • FixMotion graphics render in the installed app: the renderer now finds its own packages instead of relying on a developer setup
  • PerformanceThe install is less than half the files it was (70,500 to 32,500), so installs and updates finish much faster
  • ImprovementRemotion on one version everywhere; ElevenLabs and Pikzels models are picked from a list in Settings, including Eleven v4
  • FixThe background processes stop cleanly when VidNexus closes
v1.0.0-rc.1
2026-10-20

Release candidate: installer, runtime fixes, company details

  • FeatureFirst packaged build: the installer now ships Remotion with its Chromium, a portable Node runtime, a full ffmpeg (AMD AMF, NVIDIA NVENC and CPU encoders) and the motion-graphic sources, so nothing has to be installed separately except Ollama
  • FixThe app no longer assumes the API is on port 8000; the desktop shell picks a free port and the browser side follows it
  • ImprovementOne version number across the app, API, site and installer
  • ImprovementTerms, Privacy and Refund policy are editable pages; a Support page with a contact form (spam-protected) and a support inbox in the admin
  • ImprovementLegal pages, EULA and installer metadata name MYD International LLC
v0.17.0
2026-10-19

Research tools, thumbnails, YouTube publishing and plain wording everywhere

  • FeatureTopic pulse works again (Google Trends, YouTube or Google, 16 countries) and Keywords measures competition from real search results instead of a capped total
  • FeatureThumbnails: real Pikzels edits of an existing image, a score with a YouTube-feed preview, reference images, prompt reuse
  • FeatureYouTube publishing reports what YouTube applied, schedules in your time zone and never uploads twice
  • FeatureAudit your own channel: takeaways, what to make more of, a views-per-upload heatmap and pacing advice
  • FeatureCinematic shots through Higgsfield's API, cropped to the video's shape
  • FeatureUse your own recording on any scene; brand kits can leave the speech speed to the app
  • ImprovementScript review: faster Regenerate outline, outline edits that stick, an honest hook score and a fact-check you can act on
  • ImprovementHooks, titles, tags and descriptions stay on the script's facts; chapters cover the whole video
  • ImprovementBrand defaults: the graphics blocklist covers every kind and every way a graphic is added
  • ImprovementSettings, Library, Tools and the editor panels reworded in plain language; honest error messages throughout
  • FixRender overlays every motion graphic of a scene in order and keeps graphics in step with the real voice length
  • FixSRT export matches the MP4, with no one-word tail blocks
  • FixSettings: clearing a field can no longer stop the API; keys paste cleanly; the license key is masked
v0.16.0
2026-10-12

A script writer that works like a professional

  • FeatureScript writer rewritten: a story spine, one home per fact, continuity between sections, an editor's pass that rewrites only flagged scenes, and a linter for machine-text tells
  • FeatureShorts: research before writing, a Shorts-only structure, vertical clips and narration timed in seconds
  • FeatureTimeline editing: real split, trim as a pause, undo for scene edits; b-roll edits keep footage and add per-clip zoom
  • FeatureResearch workbench: Trends to Keywords hand-off, live context, all tools in one place
  • FeatureDocs admin: every guide and tutorial on the site is editable like a blog post
  • ImprovementAI editor: one undo per request (voice restored too), readable tool cards, scenes named by number
  • ImprovementCaptions use the script's words and Whisper only for timing
  • ImprovementMusic and SFX: library folder, crossfades that work, auto-placed sounds follow the cuts
  • PerformanceBeat detection with numpy instead of librosa; film-grain renders capped to a normal file size
v0.15.0
2026-10-01

New machine, 26 new motion graphics, the first full test pass

  • Feature26 motion graphics and themes derived from HyperFrames: number wheel, split flap, bar-chart race, line graph, arc gauge, world and US maps, callouts, hand-drawn arrows, headline slam, proof cards and more
  • FeatureScene transitions: whip pan and light leak; a film look with grain and vignette; a camera punch-in zoom profile
  • FeatureBundled fonts (Inter, Bebas Neue, Playfair Display, JetBrains Mono, Caveat, Permanent Marker, Courier Prime) so graphics render the same on every machine
  • ImprovementRuns on AMD graphics: hardware encoding picks AMF or NVENC automatically, captions fall back to CPU, the model recommender understands laptop iGPUs
  • ImprovementMotion graphics follow the pacing budget and are spread across the video instead of piling up
  • ImprovementOnboarding: low-disk warning with Install disabled, re-tests the model when Ollama comes up, never saves an uninstalled model
  • FixSix scene transitions that broke the render; 4K stock clips in motion-graphic backgrounds; per-clip zoom that only panned
  • FixScript fact-check never ran; timestamps shown in the wrong time zone; em dashes in narration
  • FixBuilds interrupted by a crashed API are marked failed on restart instead of spinning forever
v0.14.5
2026-09-24

Summer release: licensing, onboarding, AI editor, research tools, YouTube publishing

  • FeatureLifetime licensing: activate on up to two machines, release a seat from your account page, offline grace period, LemonSqueezy checkout and key delivery by email
  • FeatureSetup wizard: detects your hardware, recommends and pulls a local model, checks it answers, collects your API keys and creates the first channel
  • FeatureAI editor: rewrite a scene, swap, add or remove a motion graphic, set a visual, regenerate thumbnails from a chat box, with undo and snapshots; Ollama or a hosted model
  • FeatureResearch workbench: keyword research with saved lists, Google Trends topic pulse, title scoring, tags, SEO descriptions, competitor analysis and benchmarks, brief synthesizer, entity discovery, Humanize and the Lab
  • FeatureTools: channel audit, hooks, pacing dashboard, script length, thumbnails
  • FeatureYouTube: connect a channel, publish with title, description, tags, thumbnail, playlist and schedule, update metadata, publish history
  • FeatureBrand kits: logo, LUT, intro and outro with position, opacity, scale and strength; per-channel pacing profiles; channel history sync
  • FeatureFootage: Higgsfield cinematic shots, YouTube Creative Commons as a source, per-scene clip offset and zoom, one-scene voice regeneration, voice preview
  • FeatureHooks with voiced previews, SFX and music libraries with auto-placement and beat detection, caption style presets
  • FeatureExports: MP4 download, SRT, FCPXML for Final Cut and Resolve; asset library with favourites
  • FeatureSettings: API keys editor, storage paths, GPU runtime install, health checks, Ollama status and model pulls, cost chip and cost breakdown, cost estimate before a build, usage page
  • FeatureOpt-in telemetry, crash reporter and Report an issue with diagnostics attached
  • ImprovementDesktop shell with auto-update, per-user install, user data kept outside the app folder so updates never touch projects
v0.14.0
2026-06-18

Cyberpunk website + comparison pages

  • FeatureFull cyberpunk design system: scanlines, HUD-frame panels, neon glow accents, terminal blocks, glitch text
  • Feature6 head-to-head competitor comparison pages (VidRush, InVideo, Pictory, HeyGen, Synthesia, Lumen5) with per-tier pricing transparency
  • FeaturePublic changelog + public roadmap with vote counts
  • FeatureSecurity page + sub-processor list + verification commands
  • FeatureDocumentation site with Getting Started walkthrough
  • FeatureShowcase, integrations, community, enterprise, affiliates, education, press, blog pages
  • FeatureAdmin panel (auth + DB-backed roadmap + changelog + content blocks + license dashboard)
  • ImprovementMulti-tier dropdown navigation (19 routes organized into 4 dropdown menus)
v0.13.0
2026-05-24

Brand finalized as VidNexus

  • FeatureFinal brand identity locked: VidNexus (after NexusVid → VidForge → VidKing iterations)
  • FeatureCustom hexagonal NexusMark SVG logo with satellite-node geometry
  • FeatureMarketing site v1 launched on Next.js 15 + Turbopack
  • FeaturePricing set: $999 Early Bird, $1,499 standard lifetime
  • FeatureLemonSqueezy chosen as merchant of record for license sales
v0.12.0
2026-04-20

Pikzels thumbnails + tools sidebar

  • FeaturePikzels.com integration: 4 AI thumbnail variants per project
  • Feature/tools sidebar section with utility menu
  • FeatureProject overview dashboard with status cards
  • FeatureScript review gate on project creation
  • ImprovementHook scorer rates retention strength of the first 8 seconds
  • FixLong-render memory leak fixed
v0.11.0
2026-03-15

Multi-pass long-script generation

  • FeatureDeep mode: 24-call multi-pass script generation for 20+ minute documentaries
  • FeatureTone presets: documentary, listicle, story, educational
  • FeatureFact-check pass flags claims that need citations
  • ImprovementLLM keep_alive: -1 to prevent model unloading between scenes
v0.10.0
2026-02-08

Per-clip motion graphics + editor refactor

  • FeaturePer-clip motion graphics (was per-scene only)
  • FeatureSnap-to-clip resize for motion graphics
  • Feature+ Add Transition button between clips with transition settings modal
  • ImprovementRemoved right Animation Library sidebar; moved to left-side menu icon
  • FixEditor 404 fix when navigating directly to project URL
v0.9.0
2026-01-12

NVENC + AMF GPU acceleration

  • PerformanceDirect NVENC encoding on NVIDIA GPUs: 5-min videos render in ~3 min
  • PerformanceAMF support for AMD GPUs (RDNA 2+)
  • PerformanceRender stage timestamps for debugging
  • FixNVENC probe size raised from 128×72 → 320×180
v0.8.0
2025-12-10

31 motion graphic types + auto-fill

  • FeatureExpanded motion graphic library from 9 → 31 types
  • FeatureAuto-fill: the LLM suggests the appropriate motion graphic per scene
  • FeatureBrand-aware styling on every motion graphic type
v0.7.0
2025-11-05

Parallelization + SQLite tuning

  • PerformanceParallel 4× voice synthesis: voiceover time cut ~70%
  • PerformanceParallel 3× motion graphics rendering
  • PerformanceSQLite tuned: cache_size, temp_store=MEMORY, mmap_size
  • PerformanceCLIP scoring on GPU (batch 32 on CUDA, 8 on CPU)
v0.6.0
2025-10-08

Coverr + Pixabay integration

  • FeaturePixabay API as second stock footage source
  • FeatureCoverr API as third stock footage source
  • FeatureB-roll cycling via StockCandidate.clip_order
  • PerformanceParallel multi-source clip search (3× faster matching)
v0.5.0
2025-09-12

Channel library + brand profiles

  • FeaturePer-channel brand profiles (voice, color palette, motion style, fonts)
  • FeatureChannel library view with project grouping
  • FeatureMulti-channel workflow: manage 5+ faceless channels from one install
v0.4.0
2025-08-14

Whisper captions + ASS subtitle styling

  • FeatureWhisper transcription for burned-in captions
  • FeatureASS subtitle styling per brand profile
  • FeaturePer-scene voiceover MP3s preserved as separate files
v0.3.0
2025-07-18

Remotion motion graphics: the first 9 types

  • FeatureRemotion-based motion graphics system: 9 initial types
  • FeaturePer-scene motion graphic assignment in editor
  • FeatureVisual timeline editor for motion graphic timing
v0.2.0
2025-06-22

End-to-end first render

  • FeaturePexels API integration: the first stock footage source
  • FeatureElevenLabs voiceover integration
  • Featureffmpeg-based scene assembly + final render
  • FeatureFirst end-to-end render: prompt → script → footage → voice → MP4
v0.1.0
2025-05-15

Script generation via Ollama

  • FeatureOllama integration for local LLM script generation (qwen2.5:14b default)
  • FeatureOpenAI-compatible endpoint configuration
  • FeatureHook + body + outro structured script output
  • FeatureBasic project creation flow in Next.js editor
v0.0.1
2025-04-10

First commit

First commit
  • FeatureProject scaffolding: FastAPI backend + Next.js 14 frontend + Remotion compositions
  • FeatureSQLite database for projects + channels + scenes + clips
  • FeatureInitial pipeline architecture defined
  • ImprovementTech stack locked: Python/FastAPI/SQLite + Next.js 14 + Remotion + ffmpeg

Want to know what's coming next? See the roadmap