How to use EgoVisual 5.8.5

Launch price on release: €29
Future regular price: €49 (planned)
No increase date set.

Coming later. This product is not available for purchase or new downloads yet.

EgoVisual live visual workspace with effects and background and foreground controls
EgoVisual live visual workspace with effects and background and foreground controls.
Guide contents

Style favourites: select a style in the preview browser and press STAR STYLE (or UNSTAR STYLE). FAVOURITES filters the list to your saved choices; APPLY still selects the actual style. Favourites are saved with the session. The preview continues to use its demonstration beat.

This guide covers the Windows 10/11 x64 Standalone and VST3.

Source Studio covers per-source playback, subject preparation, cutout, reframe, compositing, hover previews and timeline editing.

Install and start

Use EgoVisual-5.8.5-Windows-x64.msi for a per-machine installation.

With BG/FG SYNC enabled, the tempo dials stay disabled and fixed at their saved manual values, including under Auto Director. Playback follows the active BPM.

Foreground media keeps its aspect ratio and can shrink to reveal the background. Song energy and build/drop analysis shape its zoom. Water uses a steady rhythm: half the BPM takes twice as long when that layer's SYNC is on. With SYNC off, the layer's manual tempo controls its speed.

Auto Director selects and clears the non-strobing FX automatically. Their buttons light up to show the current selection and are read-only while Auto is on. Switch Auto off to edit FX and restore your saved manual choices. Stroboscope remains editable and requires your explicit selection.

The new creator, brand, queue/recovery, portable-session and local performance workflows are described in Creator workflows.

Windows installers are provided as MSI only. Close EgoVisual and DAWs using it before upgrading, then restart the DAW and rescan VST3 plug-ins. The default machine paths are:

These are development packages. The adjacent .sha256 files verify package integrity. Source/build details and validation limits are in the README.

Make a visual from a song

  1. Open the Standalone and load audio on LIVE. Optional local video, background/foreground media and subtitles can be added there too.
  2. Choose a Style and Composition. Run ANALYZE AUDIO, then MAKE IT VISUAL to build the music-aware scene schedule.
  3. Press PLAY. PAUSE holds position; STOP rewinds. The audio progress control follows the loaded audio's elapsed and total time.
  4. Adjust the independent BG and FG rows. Opacity sets visibility; Color, Brightness, Saturation, Sharpen and De-Mist control appearance. FX sets the extra treatment. Choose NONE to remove a role entirely. Foreground Opacity 0% hides its FX completely. Raising foreground FX from zero keeps a clean background video intact and adds treatment within the foreground; small FX or opacity values do not enable full-frame distortion.
  5. Save the session so its source references, score, controls and subtitles can be recalled. Keep the referenced source files available when reopening it.

Infinite Scroll, Infinite Zoom and Auto Director

Background images, video and generated FX always stretch to the entire visual canvas, including LIVE, detached/fullscreen output, recording and export. Square and portrait sources stretch to the selected output shape without FIT bars or a cover crop. FIT/FILL applies only to foreground video.

Choose EgoB Flow, select INFINITE SCROLL or INFINITE ZOOM for the background, and use MORPH 3D for the foreground. Start with the packaged backgrounds to establish a working scene.

BG Motion sets the requested travel amount. BG Tempo sets its pace; enable that row's Sync to follow the host/standalone beat position. The master Motion control and active audio transport still gate movement. Raising BG Motion while BG Tempo is zero does not advance the BG travel clock.

Auto Director retains this explicit travel request. The late plate and recovery pass move together, and zoom cycles crossfade without a visible reset. You can leave BG/FG De-Mist at 0%: zero adds no De-Mist correction, and the final recovery no longer adds the extra gamma lift that caused haze. Increase De-Mist when the actual source needs additional contrast correction.

If the background looks still, check PLAY/audio activity, master Motion, BG Motion, BG Tempo/Sync, the selected background mode, and its opacity. If it looks absent, check that a usable StylePack or local source is selected.

Direct a longer arrangement

In GENERATE, build the VISUAL SCORE, then edit section motifs, palettes and transitions. Lock decisions before REGENERATE UNLOCKED. Artist DNA saves a reusable visual language as .egodna; it does not embed the original audio, video or complete song arrangement.

Auto Director shapes the show from the music while the explicit source choices and control limits remain in force. STROBOSCOPE is the explicit flash control; it is not required for ordinary automated transitions.

Video, subtitles and output

Standalone video supports IN/OUT, loop, scrub, frame nudge, project delay and foreground FIT/FILL. Background video always stretches to the full canvas. USE VIDEO AUDIO starts off; enable it deliberately when that soundtrack should drive playback and analysis. Import SRT/VTT/WebVTT subtitles from LIVE or by dragging the file into the editor. Open SOURCE STUDIO beside the source timing controls to edit each role's IN/OUT, offset, loop and playback mode; use APPLY to save the change.

Detach/fullscreen the output or select Spout for a receiving application. F11 toggles fullscreen and Escape exits. Use LIVE recording for a performed show; recording can be armed before PLAY and is finalized by STOP or the record control. Use SOCIAL for a selected output aspect or EXPORT HQ MASTER for the standalone 4K/60 path. Wait for finalization before opening the output. Resolution and GPU capability affect export speed.

Use the VST3 in a DAW

Insert EgoVisual on the master bus. Audio passes through unchanged and the DAW owns playback and BPM. The plug-in's transport controls mirror host state; they do not start the DAW. END UNKNOWN is expected when the host supplies no total project duration. An analyzed file's duration is labelled FILE.

Use the Standalone for decoded local photo/video workflows: those source modes remain disabled in the current VST3 host policy. Test the intended recording/output configuration before relying on it in a performance.

React and Graph controls

Choose ADVANCED or EXPERT in the interface selector to open REACT. Its seven dials are Master, Kick, Snare, Hi-Hats, Bass, Vocals and Instruments. The dials sit above the cue timeline and remain visible at the supported minimum window size (1000 × 760). Cue controls remain accessible by scrolling on shorter windows.

Bass controls low-end warp/pulse response. Instruments controls harmonic flow and morphing, using midrange/harmony analysis. These are musical estimates from the mix, rather than separated audio stems. The new Instruments setting is saved in sessions and exported takes; Analyze can recommend its starting value, and you can adjust it afterward.

In EXPERT → GRAPH, AUTO / AUDIO GRAPH follows song-section analysis and live musical features. Auto Director also drives the graph unless you have taken manual control. The graph shows its current automatic layer levels and routes. All graph controls remain editable: changing one keeps the displayed graph as your starting point and switches Graph Auto off. Auto Director can continue directing scenes. Re-enable the Graph Auto checkbox to resume automatic routing. Manual graph ownership survives saving and reopening a session.

In manual mode, keep Master above zero and raise only the musical input(s) you want to follow. A musical dial at 0% excludes that input from camera, flow, particles and other audio-driven effects, including export. All six musical dials at zero, or Master at zero, disables audio response. Authored video playback and timed scene changes retain their transport. Master scales the enabled inputs; it does not add a separate full-mix response.

Enabled lanes combine at their individual sensitivities. Analyze can recommend those sensitivities; manual adjustments remain effective. Whole-mix cues are limited by the lowest lane setting, so they cannot bypass a muted input.

Independent sources and subtitles

Choose COMPOSITION | NONE to select foreground and background sources independently. Loading a video or photo keeps None selected. Choosing None also keeps already selected sources; it does not apply a new composition preset.

Use SUBTITLES to import SRT or WebVTT. A sidecar with the video's exact basename loads automatically (for example, concert.mp4 and concert.srt). The SHOW checkbox hides or restores cues in preview, detached output, recording and export. This setting is saved with the session.

Subtitles associated with a video follow that video's source position. Selecting an unrelated video clears the previous cues and loads only its matching sidecar, if present. Reopening the same video keeps its imported cues and SHOW setting. Selecting a subtitle file in a chooser opened for an earlier source does not attach it to a newly loaded video. Embedded video subtitle tracks are not imported by the SRT/VTT sidecar workflow.

Interface appearance

Use the Theme selector near the top of the window to choose Original, Graphite (charcoal panels and teal accents), or Copper (warm brown panels and amber accents). The choice is remembered for this app on this computer. It changes only the interface; audio, saved projects, and rendered visual colours keep their existing settings.

Source Studio

Open SOURCE STUDIO beside the source timing controls on Live. This is an optional editing surface; choosing a style and using Auto Director remains the main workflow.

Source playback

Choose Background or Foreground. Each role can save its own IN, OUT, signed offset, loop, rate and playback mode. OUT = 0 uses the source's natural end. Enter times as M:SS.mmm; positive offset delays a clip, negative offset advances it. Click APPLY to save. UNDO restores one prior edit.

Normal, Half, Double, Reverse, Freeze at IN, Ping-pong, Beat ramp and Auto music remix use the project clock. Normal presents decoded frames directly; it does not blend with a previously visited frame. Non-looping sources hold their last frame. The shared Link to host switch makes an admitted DAW's absolute transport authoritative, including paused seeks and loops. This does not remove the existing host/hardware video-admission gate.

The time display offers clock, SMPTE, source-frame number and musical beat. Arrow buttons nudge frames when local seeking is available; keyboard arrows nudge one frame, Shift arrows about one second and Ctrl arrows one beat. Local controls cannot seek a linked DAW transport. Freeze holds IN; change IN to choose a different held frame.

Subject preparation and reframe

Load a video or image into a role. Choose Auto subject or User region; for a region, set Subject X/Y and region width/height. PREPARE SUBJECT runs the bundled U²-Net model locally with progress and cancellation. It saves a cache and enables reframe. Review the result before exporting.

Reframe follows the subject with adjustable damping and zoom while preserving source proportions. It responds to the output aspect ratio, including square and portrait. For a manual crop, leave analysis unprepared and set Subject X/Y. Anchor strength attaches a selected geometry/particle/feedback/zoom/symmetry/ distortion/text destination to the subject. With reframe enabled, the subject anchor is centred in the reframed composition. Text keeps a safe margin.

On the foreground role, enable Subject cutout, then Apply, to use the prepared alpha matte on planes and object materials. Preparation samples at 4 Hz with 160 × 160 mattes, supports sources up to one hour, and needs review for fast motion, fine edges, occlusion and multiple subjects. It is a saliency model: dedicated face/head/person classification is not implemented.

Caches are saved under the user's EgoB/EgoVisual/SubjectCache directory and included when collecting a session. Missing/unreadable caches stop export with an instruction to prepare the source again. Reprepare when footage is replaced or the subject region changes.

Compositing

Each role has Normal, Add, Screen, Multiply, Overlay, Soft Light, Hard Light, Difference, Exclusion, Lighten, Darken, Color Dodge and Auto Blend. Auto Blend chooses a mode from style and musical section. Match to style adds a shared grade, with distinct Classic, Tron, psychedelic and machine treatments.

Foreground Chroma key accepts a six-digit RGB colour (for example 00FF00) and has tolerance, edge softness and spill controls. Foreground surfaces adds up to three nested portals using the same source. These treatments are off by default. A collected take and GPU export use the saved source controls.

Hover previews and timeline overrides

The Preview styles / FX tab renders short demonstration loops on hover. Loops use a private 320 × 180 renderer, are cached with a bounded memory budget, and do not alter the live session. Select a style and press Apply style to use it. They preview the generated style/FX on a demonstration beat, not the loaded source or its complete Auto Director arrangement.

On the existing Performance FX timeline, drag a phrase to reorder it, or drag an edge to trim/stretch the adjacent unlocked phrases. Edits preserve the song duration and continuous coverage. Right-click to lock/unlock, bypass, replace the effect or regenerate that phrase. Existing Visual Score controls continue to own section motifs, palettes and transitions.

Current limits

The following capabilities remain outside the current release:

Creator workflows

Load and analyze the soundtrack, approve its visual take, then open CREATOR STUDIO.

Deliverables and comparison

CONTENT PACK queues a full video, vertical 15/30/45-second clips, a teaser, square clip, eight-second loop, thumbnail, square cover and portrait poster. Short sources are clamped to their available duration. Each pack gets a separate folder, a manifest and a local HTML gallery. The manifest lists all ten required outputs, source ranges, dimensions, brand, job IDs, attempts and current status. It reports completion only when all ten jobs have published files; cancellation, retry and missing files remain visible. A reopened collected clip keeps its approved source range. Open EXPORT JOBS to inspect, cancel, retry or reveal individual outputs. Reopen the gallery after jobs finish.

PREVIEW 8 renders eight deterministic, two-second motion variants from the same musical window. The tiles show their best rendered frames and measured hook scores out of 100. Click to select; double-click to play the completed preview in the Windows video player. Keep the selected variant or export it. Campaign export queues eight variants in three aspect ratios. Each slot carries a distinct palette direction through LIVE, export and saved sessions; locked scenes retain their approved palette. Source imagery, locks and strong effects can still limit perceptual differences between alternatives.

Every rendered export gets a .metrics.json report and .hero.png. Hook scores combine image contrast, colour separation, central edge distribution, movement, depth state and musical energy. They are comparison heuristics, not an assessment of artistic merit. Loop exports also decode the finished AAC soundtrack and measure its boundary step and local level difference. A high score is not a promise of an inaudible or musically seamless join; audition the loop.

Short-form direction uses continuous hook, change, escalation, payoff, return and call-to-action phases. It retains the take's seed and clock. Export does not add an unrequested opening white flash.

Captions and brand

Import SRT/WebVTT cues through the media workflow. Choose fade, type-on, rise, pulse, echo, word build, depth or mask motion in Creator Studio. Pulse and depth use bounded vocal, beat and section evidence from the existing analysis. Reduced motion shows the complete cue with no kinetic movement. Cue timing follows the original source timeline, including when a short clip starts partway through the soundtrack.

Brand profiles store artist name, logo PNG, font, colours, handle, website, watermark and intro/outro choices independently in .egobrand files. New profiles embed their PNG logo, so loading a profile still works after the original logo has moved. Existing version-1 profiles remain readable. Logos are limited to 4096 ? 4096 pixels and 8 MiB; loaded embedded logos are cached locally. Use Apply to update the approved session. The shared caption/brand compositor is used by preview and rendered exports. Brand fonts and colours apply to imported captions and exported titles, captions and covers. Still images show the complete cue without an animation entrance; brand artist/handle values override legacy social text. Imported cues take precedence over beat-cycled promotional captions when both occupy the caption area. Queued exports retain a logo hash and stop if that logo changes. A missing font uses the operating system fallback; portable packages include the logo but do not redistribute fonts.

Portable work and content

Collect Session creates an .egopack containing the approved session, controls, analysis, soundtrack and referenced media. Import verifies its manifest, sizes and SHA-256 hashes before publishing a new folder. Open Collected restores audio first, then the approved take, so analysis cannot regenerate the chosen look. The VST3's Open Standalone action uses the same ordering.

Create StylePack opens an authoring form for a name, author, licence, tags and palette. It packages local Backgrounds/Foregrounds with required style.json and a preview (generated from the first background if absent). The form supplies Calm, Pulse and Flow reaction profiles; their motion/intensity settings are selectable after the pack passes validation. Custom metadata can define 3?8 distinct profiles. Browse Local searches folder names and metadata files.

The portable image-pack schema accepts PNG/JPEG assets. It requires at least six distinct backgrounds and ten distinct foregrounds, comparing small pixel signatures so renamed duplicates do not count. Limits are 256 entries, 16 MiB per image, 4096 ? 4096 pixels, and 256 MiB each for compressed image files and decoded pixels. Preview PNGs must fit within 1024 ? 1024 pixels. Video, object and shader assets are not admitted by this image-pack schema. Schema, hashes, paths, image decoding, diversity and budgets are checked before import publishes a folder; local pack selection also validates on the export worker. These checks describe bounded content admission, not measured FPS on every GPU. Metadata format. Only include content you have permission to redistribute. Package import never merges into an existing destination folder.

Export requests are persisted before rendering. After an interrupted process, reopen Export Jobs and choose Retry. Rendering restarts from frame zero using the captured take, rather than attempting to append to a damaged MP4. Missing or changed source files stop recovery. Finished outputs are published atomically. Packaging and community transfers are cancellable jobs but are not restartable after a process crash. Media references remain immutable during a queued render.

Local performance and video

Setup & Performance Advisor reads the current adapter/driver and observed preview frame timings. Refresh after playback, review its suggested presentation tier/FPS, then Apply or Revert. Measure Export queues a short 1080p/30 probe and reports actual throughput. Saved reports omit source paths and project names. The recommendation does not certify another GPU, driver, source or DAW.

Standalone video uses the native Windows decoder. Preview and offline export share frame selection: a frame remains displayed until the next presentation timestamp, including fractional and variable frame rates. Unknown codecs report decoder failure. In-host VST3 decoding requires a matching local certification profile; otherwise Open Standalone retains the selected source reference.

Optional community

Community / Share Remix connects only after an explicit action. Enter the HTTPS origin of a configured service, search approved packs, or submit a media-free recipe with a publisher token and redistribution licence. Submissions require moderation before their share link becomes public. Tokens stay in memory.

The repository includes a self-hosted service and tests in Community/. No public service or hosting destination is bundled or automatically deployed. Local HTML galleries and portable packages work without a community service.