# Record a PHASE//VOID music visualizer video for YouTube

Canonical guide: https://phasevoid.com/guides/youtube-visualizer/
Updated: 2026-10-28


## Choose a video workflow

### Music video

Turn a local track into a finished film.

1. Choose your MP3/WAV and its passage. Use Current Studio for one composition, or review your saved acts for a changing show.
2. Choose the frame before reviewing. Check the first picture and later passages when your show changes. Review again after edits.
3. Render, save and reopen the actual file. File render runs on this device and can take longer than the song.

[Open File render](https://phasevoid.com/studio/?flow=video&file-render=1) · [File-render steps and limits](https://phasevoid.com/guides/youtube-visualizer/#file-render)

### Lyrics & captions

Put your words at the right moments.

1. Choose your MP3/WAV in Sound. In Layers & text → Timed text / lyrics, add lines or import a plain VTT/SRT file.
2. Save start/end times, placement and size; enable Show timed text. Seek between lines with Speakers off if you want quiet work.
3. Use File render with that soundtrack. Trimmed films keep absolute file seconds. Review and inspect the saved film; the words become pixels.

[Open timed-text instructions](https://phasevoid.com/studio/?help=timed-text) · [Timed-text steps and supported formats](https://phasevoid.com/guides/youtube-visualizer/#timed-text)

### Performed take

Record the changes you perform.

1. Connect your intended file or live audio yourself. Prepare worlds, layers and cues, then choose the frame and a supported recording format.
2. Record video captures the picture and sound in real time while you perform. Keep the Studio visible; test a short passage first.
3. Stop, save and reopen the take. Check changing pictures and audio before publishing. Choose File render for a preplanned local-file film.

[Open real-time capture](https://phasevoid.com/studio/?flow=video&realtime=1&from=guide-youtube) · [Live-recording steps and browser limits](https://phasevoid.com/guides/youtube-visualizer/#record)

## Actual timed-text example

[Four-second encoded film](https://phasevoid.com/examples/timed-text/performance.webm), [first decoded frame](https://phasevoid.com/examples/timed-text/poster.webp), [second decoded frame](https://phasevoid.com/examples/timed-text/second-line.webp), [plain VTT](https://phasevoid.com/examples/timed-text/words.vtt), [original WAV](https://phasevoid.com/media/examples/techno-test-loop-160bpm.wav). The film uses source seconds 4–8 and decodes 120 pictures at 30 fps; frames at film 0.5 and 2.5 seconds show the respective lines. No physical/device-wide synchronization claim. HTML chapters expand on their old fragment links; this Markdown keeps all details readable.

## Detailed reference

## Render your track and complete show as a file

Workflow owner: https://phasevoid.com/guides/youtube-visualizer/#file-render

Make a music video from a local track and your complete composition, or give saved acts a reviewed timeline. File render advances the film on its own timeline rather than capturing the screen in real time. Processing can take longer than the finished video; the track is decoded silently, without speaker monitoring.

[Open File render in Studio](https://phasevoid.com/studio/?flow=video&file-render=1). [Prepare saved acts](https://phasevoid.com/guides/live-visuals/#sequence) or [recover local clips](https://phasevoid.com/guides/live-visuals/#media-library) first when needed. Opening Studio does not start sound, render a film or request live-input permissions.

[Try the complete four-act example](https://phasevoid.com/studio/?file-render=1&render-example=1). Its setup opens first; loading and rendering remain separate deliberate actions.

1. **Choose the composition.** Open File render with your selected world, output frame, size and motion. Use Current composition for your own look, or press Load four-act example for the original complete fixture. Loading the example checks the known session, two loops and WAV before changing worlds, mapping, sequence, Clip Deck, title and timing. It keeps your selected output frame/size/motion. Undo example setup restores prior settings and loaded local clip resources paused. Opening the link alone does not load anything, play audio or render a film. Stop running performances, mixes, recording and loading first; turn off Freeze/Blackout before reviewing a moving film.
2. **Choose the sound and segment.** A local MP3/WAV already selected in this Studio tab is handed to File render silently without choosing it again. Otherwise press Choose MP3 / WAV for a finite mono/stereo file up to 200 MB. Choosing your own track keeps the current composition; it does not load the example or reset your world/frame. The example loader supplies its original WAV silently; Use original 12-second test WAV also supplies that soundtrack without replacing the composition. Set Start in seconds and End in seconds inside the chosen file and review again after replacement. Set the Studio audio/record level before Review: zero produces silent output; Speakers off only disables monitoring. This path does not request live-input permissions or start monitoring.
3. **Review the picture and timeline.** Choose Current composition or Show sequence · once, then hold last act. Keep Current Studio frame or choose delivery formats inside File render, then choose 24, 30 or 60 fps, 6, 12 or 24 Mbps and a supported format. For a queue, choose explicit Manual holds, fit film-only holds or set Soundtrack entry times, then press Review film timeline. Review each act, incoming transition and scheduled boundary; acts beyond the selected segment are listed but do not appear in the film. The first act begins with a cut; the last reached act continues to the segment end. The film does not reconstruct pad clicks, MIDI gestures, Tap, Resync or a live performance.
4. **Resolve the preflight.** Check the sound segment, total film length, clip dependencies and encoder support before rendering. A missing file, incomplete hold or unsupported configuration must be resolved first. The result is limited to 30 minutes and 750 MB; memory and device limits may require a shorter segment or smaller output.
5. **Render, with control.** Press Render reviewed film deliberately and follow its progress. Studio pauses the show during the job; the exported composition does not replace Program or recalibrate your map. Keep the page open and the device awake. Cancel render discards the unfinished job; cancellation or an error is not a completed film. Music playback stays paused after completion or cancellation: resume it explicitly.
6. **Inspect, save and reopen.** When the result is available, choose Check rendered video. Use Play start, Play middle and Play end, watching clip trims, transitions, surface crops and the first and last intended sounds. Use Download rendered video, then Open a saved video to check to reopen the actual MP4/WebM. Download timing report keeps the local inspection results. Check the saved file in your intended player or editor before choosing I saved this rendered video. Completed films remain available until explicitly discarded; an unsaved result requires its own confirmation.

### Keep your chosen passage

In the catalogue, Propose this passage in Studio is off by default. Enable it to send an optional private range with your File. In File render, choose Use proposed fragment or Keep current range; receipt alone does not apply the proposal. Use the waveform, seek and Start/End controls to check the selection. Preview selected soundtrack plays decoded PCM for up to 60 seconds; this preview limit does not shorten a longer film. Hear soundtrack preview defaults off. Use whole track resets the selection; changing it requires Review again. Preview and export share the decoded soundtrack timeline. Recognized valid MP3 gapless metadata removes encoder delay/padding; otherwise timing remains explicitly encoded-time. Check the displayed timing basis and intended first/last sounds: this is not a universal sample-accurate or audio/video-sync guarantee.

### A dated soundtrack cut check

Native desktop Chrome checks compared the start, middle and end of seven authored WAV/MP3 fixtures across 12–48 kHz. Two actual saved films retained their complete selected soundtrack interval with a measured content offset of zero samples at 48 kHz; decoded codec tails and container lengths were measured separately. This bounded software check does not certify every encoder or physical audio/video synchronization. [Read the test method, profiles and results](https://phasevoid.com/evidence/soundtrack-trim-2026-10-14.json).

### Start with the original four-act fixture

[Open the complete example setup](https://phasevoid.com/studio/?file-render=1&render-example=1), then press **Load four-act example**. The explicit loader fetches, validates and decodes the known session, both original loops and separate synthetic WAV before applying the setup. It replaces worlds, mapping, sequence, Clip Deck, title and timing while retaining your current output frame, size and motion. **Undo example setup** restores prior settings and loaded local clips paused. The link itself loads no session, starts no sound and renders nothing.

Choose the show sequence, review its timeline, then render deliberately. Use **Choose MP3 / WAV** to replace the soundtrack with your own file while retaining the composition; review the new segment and holds. This does not automatically fit every act to a song or detect its BPM. For the published reference, select 960 × 540 and 30 fps; another frame/motion choice produces a different composition.

Manual download/recovery fallback: save your own session first. Download the [synthetic session](https://phasevoid.com/examples/file-render/session.json) and use **Import session**, which replaces settings including its saved frame. Download [original loop A](https://phasevoid.com/examples/clip-show/loop-a.webm) and [original loop B](https://phasevoid.com/examples/clip-show/loop-b.webm), then reattach them in Assets paused. Download the separate [original WAV](https://phasevoid.com/examples/clip-show/sound.wav) and select it in File render. This fallback does not have the loader's frame-preserving transaction or Undo. Session JSON contains settings/descriptors rather than media; the fixture is synthetic, not a customer's show.


### Keep the composition, review the timing

The render includes the procedural Program, both assigned local video layers and their opacity, blend, fit, flip, trim, rate, loop and order, plus the Program master rack. Current composition starts clip playback from trim In, not from the live decoder position. Queue acts use their captured playback intent: a playing layer starts at its In point when its act begins; a paused layer stays paused. Captured surface looks and full-composition transitions use the existing calibrated layout. Program clips and master FX are applied before mapping; independent procedural surfaces use their own FX. Keep the original source files available throughout the job.

Review the calculated timeline, not just the sum of hold fields. A hold starts after its incoming transition completes. For Bars, add the requested four-beat bars to that point, then round the end forward to the next global four-beat boundary: bars multiplied by seconds per bar alone is not the full schedule. Manual holds need explicit durations for this export. Both Track and Manual grid-pulse modes use the chosen Studio BPM with phase reset to film time zero, not the captured live or source-track phase; the chosen Visual timing offset still applies. There is no automatic BPM detection or recreation of live cue launches. Choose the intended soundtrack segment and ending deliberately.

### Check the actual saved film

The inspector accepts local MP4/WebM up to 800 MB. Video packets, packets per second, track starts/ends and largest picture gap are encoded metadata, not decoded-frame counts or a synchronization verdict. Matching endpoints do not verify content sync. If the final packet duration is missing, the tail/end difference stays unknown and packet rate is approximate; codec priming can also affect decoded audio boundaries. Play the final seconds. An unreadable, oversized, unsupported or timed-out inspection means unknown, not passed. The timing report retains syncVerified: false.

Follow the [recording and saved-file check](https://phasevoid.com/guides/youtube-visualizer/#check).

### What this mode needs

File render is local processing, not an offline-installed application. The sound file must be a mono/stereo MP3 or WAV up to 200 MB; the film is limited to 30 minutes and 750 MB. Choose 24/30/60 fps and 6/12/24 Mbps. WebM uses VP9/Opus; MP4 uses H.264/AAC when that exact configuration is supported. Full-resolution source pictures have a separate 64-million-pixel job budget. Codec, graphics or memory limits can reject the configuration: no surface is silently dropped and there is no silent quality fallback. Choose a smaller delivery target or Studio output, another supported format, lower bitrate or a shorter segment explicitly. Rendering is not guaranteed to be faster than playback or to continue while the device sleeps. There is no cloud copy or automatic YouTube upload. This workflow does not establish zero drops, universal audio/video synchronization, physical-projector timing or long-set reliability.

For live input or a manual performance, use the existing [real-time workflow](https://phasevoid.com/guides/live-visuals/).

Sound files, clip media, session settings and generated output are processed locally; this workflow does not upload them. A chosen soundtrack stays in memory, including an explicit same-origin tab transfer; optional proposed ranges, waveform data and decoded previews remain local. Public URLs, MCP and saved sessions exclude the File and proposed range; reload still requires the local file. The explicit example loader fetches only its known public session, loops and WAV, never a private media URL. A local session can contain private names, clip descriptors and calibrated geometry: review it before deliberately sharing. Public setup links and the read-only MCP do not carry these private files or mappings. Four fixed first-party milestone labels count opening, starting, result-ready and the user saved-confirmation once per page load, without private filenames, media, timelines, settings or mappings. They are independent aggregate counts, not a joined conversion funnel; saved-confirmation is not evidence that a browser download completed. Ordinary website requests still reach the hosting provider. See [media and privacy](https://phasevoid.com/privacy/).

### A twelve-second film, rendered while processing took longer

This first-party **WAV → WebM** run used four acts, both original muted loops and two surfaces: **Composed wall** plus an independent procedural side. Program master FX, surface FX and Cut, Dissolve, Wipe and Dip were active. The film timeline is **12.000 seconds**; the saved container is about **12.02 seconds**, including a small Opus tail. Intentional **6× CPU throttling** made processing take about **26 seconds** in one desktop Chrome setup.

[Watch or download the actual file-render WebM](https://phasevoid.com/examples/file-render/performance.webm). It was saved from the visible local encoder Blob output and reopened through the file chooser; native-browser download completion was not certified. This is the output film, not a recording of the controls or a physical projector.

![Actual decoded Drop frame from the four-act file render](https://phasevoid.com/examples/file-render/output.png)

The image is a decoded Drop frame from that WebM, not promotional artwork.

![Actual native File render preflight with the original WAV, four reviewed acts and both recovered clips](https://phasevoid.com/examples/file-render/workspace.jpg)

Actual native preflight before starting another render of the original WAV from 0 to 12 seconds: 360 planned frames at 30 fps, 960 × 540, 160 BPM, four acts, two mapped surfaces and both files recovered. This 1265 × 949 JPEG is a still of the reviewed setup, not a capture of the clicks or a result from another format.

| Independent check | Observed scope |
| --- | --- |
| Pictures | 360 submitted / 360 decoded / 360 distinct at 960 × 540, 30 fps. |
| Video packet timeline | Maximum absolute PTS error ≤ 0.334 ms; largest video packet interval 34 ms. |
| Soundtrack | Stereo Opus, 48 kHz. Source WAV versus decoded PCM in windows at 0.5, 5.5 and 10.5 seconds: lag 0 ms, correlation > 0.998; declared tolerance one frame, 33.334 ms. |

The native reopened inspector reported **360 packets** and a **34 ms** largest picture gap; the audio tail remains unknown from packet metadata. Decoded-picture counts and PCM comparisons are separate checks. Waveform alignment in three windows does not establish sound-to-visual reaction timing, physical synchronization or support for every source/codec. A separate 20× throttled job kept Program PNG bytes identical; Cancel retained the earlier result's Blob URL. These are bounded native checks, not an endurance or universal device claim.

[Independent decode and waveform report](https://phasevoid.com/examples/file-render/decode-check.json) · [starting session](https://phasevoid.com/examples/file-render/session.json) · [original loop A](https://phasevoid.com/examples/clip-show/loop-a.webm) · [original loop B](https://phasevoid.com/examples/clip-show/loop-b.webm) · [original synthetic WAV](https://phasevoid.com/examples/clip-show/sound.wav). MP3, MP4 and portrait native acceptance are not demonstrated by this example.

## Fit a complete visual show to your music

1. **Keep your show and track together.** In the homepage live-show builder, choose My MP3 / WAV, select a file and continue with your show and track. Receive selected track loads it paused; Load planned show separately applies the reviewed recipe. In its rehearsal panel, Plan film with this show & track stops the rehearsal and opens File render with Show sequence and the same local File. No playback or render starts. Without a planned show, open File render and choose Show sequence from your existing queue.

2. **Choose the soundtrack passage.** Set Start and End in seconds inside your MP3/WAV, or explicitly use the whole track. The film starts at zero; a storyboard time of 0:02 with Start at 0:05 corresponds to soundtrack 0:07. Waveform preview has its own monitoring, off by default, and does not change the selected range.

3. **Fit every act, or keep your live timing.** Keep live cue timings preserves seconds and bar holds; Manual acts need explicit seconds. Fit all acts to selected passage switches to Film-only hold seconds, reserves the existing incoming transitions and divides remaining time using the live hold weights with a 0.25-second minimum per act. A blank Manual hold uses a one-second weight. You can edit every film hold afterwards. The live queue, calibration and soundtrack stay unchanged. These seconds are not snapped to bars, musical phrases or detected beats. Audio is not stretched. If the passage cannot hold all transitions and minimum holds, choose a longer passage or deliberately revise the live transitions first.

4. **Read coverage before reviewing.** The film lists each act’s entrance, incoming transition, hold and planned end. If it omits an act, cuts a hold or transition, or extends the last act, read the coverage warnings. Fit or edit the holds, or explicitly accept the partial or extended show. Changing the range or settings requires another approval and Review; the tool never silently refits your edits.

5. **Review actual later pictures.** Choose one output or a sequential 16:9 / 9:16 / 1:1 delivery set, then Review film timeline. Each selected size gets a real first-frame framing check. For a sequence, the first selected output is then replayed from frame zero at the chosen film FPS, using causal audio analysis and the same compositor as the encoder. The storyboard samples holds, incoming transitions and the final frame, with film time, soundtrack time and frame index. Its PNGs are scaled to at most 640 pixels for display. Other formats get first frames, not their own later-act storyboard. Processing can take as long as rendering one film; Cancel review releases that job without publishing a partial storyboard.

6. **Render, save and reopen each format.** Render reviewed film uses the same frozen timeline, soundtrack passage, media references and overlay for each output, one at a time. Completed films survive Cancel; Continue renders the missing outputs. Save each actual video, reopen it in the existing saved-video inspector and check moving geometry, every transition, start/middle/end and the intended soundtrack. A storyboard proves sampled pictures only. It is not a motion, decoded-video, audio-alignment or projector-sync verdict. Save and explicitly discard retained results before reviewing a different delivery set.

The loaded web app is required; this is not an offline installation. Audio and local media are not uploaded or included in shared recipes, saved settings or read-only MCP responses. The existing limits are 30 minutes and 750 MB across the retained delivery set; codec, resolution, GPU and memory support vary by device. Live microphone/tab audio and performed MIDI or pad changes use real-time Capture instead. A physical projector still needs on-site calibration and checking.

[Plan a soundtrack film](https://phasevoid.com/studio/?flow=video&file-render=1). [Complete film storyboard instructions](https://phasevoid.com/guides/youtube-visualizer/#film-storyboard).

### One passage. Four acts. Three checked files.

A local Chromium check on 15 October 2026 fitted a four-act, three-surface show to seconds 2–8 of the original test WAV. All three saved WebM files were reopened: each has 180 decoded, distinct pictures at 30 fps, stereo Opus audio and a 6.020-second container including the codec tail. Eight storyboard samples were compared against the corresponding decoded landscape frames; their mean RGB error was at most 4.25 / 255 after scaling and lossy encoding. Four normalized waveform windows per output matched the selected source passage at zero sample offset, with correlation above 0.996. This bounded check does not certify another device, long render, codec or physical synchronization.

![Intro · film frame 16](https://phasevoid.com/examples/film-storyboard/intro.webp)

![Wipe into Drop · frame 90](https://phasevoid.com/examples/film-storyboard/transition.webp)

![Drop · frame 114](https://phasevoid.com/examples/film-storyboard/drop.webp)

![Same soundtrack and frame 114 in all three saved formats](https://phasevoid.com/examples/film-storyboard/formats.webp)

[Measurements and limits (JSON)](https://phasevoid.com/evidence/film-storyboard-2026-10-15.json). These are decoded stills, not a motion preview.

## Place visual changes on the soundtrack

1. **Choose the show and passage.** Load your own queue or the eight-act Journal exercise, choose a local MP3/WAV, then choose Start and End inside the file. Use Show sequence as the render source. No song or clip is uploaded.

2. **Choose Soundtrack entry times.** In Timing mode, choose Soundtrack entry times, then Edit entry times to move to the editor. The first draft fits all queued acts into the passage. This is a starting distribution, not an analysis of the song. Every input uses absolute soundtrack seconds; if Start is 5, the first act is fixed at soundtrack 5 and film zero.

3. **Place the incoming changes.** Edit each entry, or Seek preview to that entry and move the existing Preview position slider. Preview selected soundtrack starts only after your click; Hear soundtrack preview is separate and off by default. Place here reads the current audio-clock position while the preview runs or is paused, then invalidates old film Review without stopping this preview. The ruler and each row show the resulting film frame. Entries round to the nearest chosen film frame, not to beats or bars.

4. **Inspect later musical passages.** Preview window starts at soundtrack second lets you decode a later part of the selected passage. Each window contains at most 60 seconds of PCM. Seek preview prepares a window around an act and seeks there without automatically playing. Moving this window or playhead does not trim the film, move another entry, start Studio speakers or change your live show.

5. **Resolve timing conflicts.** An entry means the beginning of the incoming transition, not the end of its fade. Every act needs its existing incoming transition and at least 0.25 seconds of hold before the following entry or film end. Out-of-order, duplicate-frame, out-of-range or overlapping choices block Review; no act is silently dropped. A different file, passage, FPS or show makes the old draft stale. Reset entries to fit all acts deliberately rebuilds it for the new context.

6. **Review and check the saved films.** Review film timeline freezes the same entry schedule for the existing causal storyboard and sequential delivery set. Render, save and reopen each intended 16:9, 9:16 or 1:1 film. Inspect moving transitions, the start, later acts, end and soundtrack. Entries are an in-tab film draft; they do not rewrite Manual live holds, stretch audio, detect beats or prove timing on a physical projector.

### Check the change at the selected frame

A native Chromium check placed eight uneven entries over seconds 1–11 of the original WAV. The actual saved 16:9, 9:16 and 1:1 files each decoded to 300 different frames at 30 FPS; all were reopened muted. Fourteen landscape storyboard samples matched their decoded frames within 2.72 / 255 mean RGB error after lossy encoding and scaling. Four sampled audio windows per output aligned with the selected WAV at zero samples. A separate controlled VBR MP3 sequence produced 360 distinct frames and decoded audio identical to the earlier controlled render; its quiet sampled window has lower lossy correlation, as recorded in the report. This is a bounded digital-output check, not a physical synchronization verdict.

![Decoded frames around the selected Cut entry and Impact wipe](https://phasevoid.com/examples/soundtrack-cues/boundaries.webp)

Actual decoded landscape frames: Cut enters at frame 81, soundtrack 3.700s; Impact starts at frame 102 and completes at frame 117. Labels added. These are still samples, not a motion preview. [Measurements and limits](https://phasevoid.com/evidence/soundtrack-cues-2026-10-15.json).

[Start with the eight-act set](https://phasevoid.com/studio/?journal-exercise=techno-vj-set). [Read the visual direction](https://phasevoid.com/journal/techno-vj-set/).


## One soundtrack. Three delivery frames.

Workflow owner: https://phasevoid.com/guides/youtube-visualizer/#file-render-deliveries

Make landscape, portrait and square films from the same selected local soundtrack passage and reviewed composition. Choose delivery sizes inside File render without changing the main Studio frame. Each target has its own first-picture check and output; the films render one after another.

[Prepare my delivery frames](https://phasevoid.com/studio/?flow=video&file-render=1) · [Choose a frame and pixel size](https://phasevoid.com/guides/aspect-ratios/)

1. **Keep one soundtrack and composition.** Choose MP3 / WAV once, or keep the local file already selected in this tab. Set Start in seconds and End in seconds. Choose Current composition or Show sequence · once, then hold last act; complete any Manual holds. The same selected passage, act timeline, soundtrack level and composition supply every target.
2. **Choose the delivery frames.** Keep Current Studio frame for one output at its existing dimensions, or select Choose delivery formats. Pick 16:9 · Landscape, 9:16 · Portrait and/or 1:1 · Square, then choose a Delivery short edge of 540, 720 or 1080 pixels. Each selected target displays its actual dimensions. The delivery choices do not resize Program.
3. **Review every target.** Set FPS, quality and file format, then press Review film timeline. Review checks encoder support for every selected size before a render starts. Each card shows Rendered first frame · not a motion preview: a still rendered in that target’s own frame, including the reviewed title, artwork, clips and mapping. Check text edges, clip Fit/Fill and surface crops. This is not a finished film or proof of its later motion.
4. **Render the reviewed set.** Press Render reviewed set for several targets, or Render reviewed film for one. Only one target renders at a time. Waiting, Rendering and Ready states identify progress. Keep the page open and device awake. Processing time can differ from film duration; audio is decoded silently and Studio playback remains paused.
5. **Cancel without losing completed films.** Cancel render stops the unfinished target; completed results remain available. After cancellation or an error, Continue missing targets processes only unfinished targets in the same reviewed set. Earlier completed films are not rendered again. Resolve the displayed failure before continuing; a Cancelled or Error state is not a completed film. Closing the panel ends Review and continuation; completed films stay in this tab.
6. **Check, save and manage each result.** Use Check rendered video for each Ready result, then Download rendered video. Reopen the actual download with Open a saved video to check and examine its start, middle and end. Mark I saved this rendered video only after checking the file. Discard this result releases its retained output; an unsaved result requires the explicit Discard this unsaved result confirmation. A requested download is not confirmation that the file reached disk.

Review captures the composition, title and artwork for the chosen target dimensions. The same setup is rendered separately for each shape; portrait and square are not resized copies of a landscape film. Relevant edits require a fresh Review. Preparing first frames pauses Studio playback and holds Program; rendering also holds Program. Rendering, cancelling and continuing preserve the main Program settings and frame; they do not promise an uninterrupted live performance.

### Keep the set within device limits

Choose at most three delivery targets. Completed outputs share a 750 MB retained-result budget; each film still has the 30-minute limit and the existing source-picture pixel budget. Save and explicitly discard retained results when space is needed, or deliberately choose a shorter passage, smaller frame or lower bitrate. Codec support is checked for the actual target dimensions, FPS and format. An unsupported configuration is not silently replaced or reduced in quality. MP4 uses H.264/AAC when supported; WebM uses VP9/Opus. A supported encoder is not a guarantee of successful rendering on every device.

The first-frame cards only check one picture. Inspect the saved film for moving geometry, transitions, clip trims, title placement and the intended first/last sounds in each shape. The existing timing inspector reports encoded metadata, not decoded-frame counts, content synchronization or physical-projector timing. Device codec and memory limits still apply; the set is not an endurance or universal synchronization guarantee.

A dated desktop Chrome check saved and reopened three twelve-second WAV/WebM films, each with 288 independently decoded, distinct pictures at 24 fps. The same four acts, two video sources, two mapped surfaces, title and artwork were composed at each size. Cancelling the unfinished third target retained the first two; Continue kept their file hashes unchanged. Separate two-second MP3/MP4 outputs tested all three shapes. The AAC onset changes its first 20 ms in that fixture; it is not a lossless waveform copy. These are bounded software checks, not universal sync or physical-projector certification.

[Read the profiles, hashes, methods and limits](https://phasevoid.com/evidence/delivery-set-2026-10-14.json).

![Actual decoded first frame of the landscape delivery film; a still, not a motion preview](https://phasevoid.com/examples/delivery-set/landscape.webp)

![Actual decoded first frame of the portrait delivery film; a still, not a motion preview](https://phasevoid.com/examples/delivery-set/portrait.webp)

![Actual decoded first frame of the square delivery film; a still, not a motion preview](https://phasevoid.com/examples/delivery-set/square.webp)

The local File, selected range, target previews, reviewed composition and generated films stay in this tab’s local workflow; they are not uploaded or placed in public setup links. Reload still requires local media again. Nothing renders, plays sound or requests live-input permission merely from opening the link. Read-only MCP can explain the public workflow; it cannot inspect private media or start, cancel, continue or save a delivery. This is browser processing, not an offline application, cloud backup or automatic YouTube upload.


The following capture/preview-draw instructions describe the separate real-time path for live input or performed gestures. A completed File render can be inspected directly through **Check rendered video** without making another take.

### Check a take with changing clip cues

A clip cue can change both local video layers and the Program rack, as well as the procedural look. Attach the files and press **Prepare show** before audio, recording or going live. It checks cue decoders off air and warms the show's procedural worlds while keeping the stage, settings, map and transport. A missing-file error names what to reattach. Rehearse, then Record. Follow the [original clip-show kit and capture workflow](https://phasevoid.com/guides/live-visuals/#clip-deck).

Record a short take, download it and reopen the actual saved file. Check the intended clips, trims, effects and transitions, audible music, framing and unexpected motion gaps. Stage, projector and recorder share the final image. A short successful take does not establish long-set or physical-projector performance.

### Check the file you will actually use

1. Choose the sound, frame and complete composition. For a cue show, attach the files and press **Prepare show** before audio. Keep Studio in front and uncovered.
2. Press **Check this setup** at your recording target. It measures preview draws; a good result does not certify encoded cadence or synchronized sound.
3. Under **Capture method**, leave **Timestamped · where supported** selected or choose **Browser capture · compatibility**. The default uses audio-graph timestamps where supported and falls back for unsupported encoding configurations. Choose **10-second test**, an offered format and **Start recording**. Enable **Restart track from the beginning** for a take from the start; disable it for the current position/live input. Wait for finalization.
4. The finished take is inspected locally. Read **Video packets**/packets per second, codecs, dimensions, sound channels/sample rate, track timelines, **Largest picture gap** and **Track start difference / Track end difference**. These are encoded packet metadata, not decoded frames or a sync verdict. Use **Play start**, **Play middle** and **Play end**, checking motion, cue changes and the first/last intended sounds.
5. Download, then use **Open a saved video to check** for the actual MP4/WebM, up to **800 MB**. Reopen it in your intended player/editor too. **Download timing report** saves local results with synchronization unverified. Only after checking the saved video choose **I have saved this video**. Preserve an earlier useful take before replacing it.

If final packet duration is absent, the timeline says **last packet, tail unknown** and **Track end difference** is **Unknown · play the final seconds**. Packet rate is approximate with an unknown tail. Codec priming can also make decoded audio boundaries differ from container timestamps. Use **Play end**: the last packet timestamp is not its exact picture/sound end.

If preparation/encoding fails, save any available take and retry **Browser capture · compatibility** or another offered format. For uneven motion, try 30 fps and a 720-pixel short edge, or lower Render detail without changing output size. Compare one change at a time. More bitrate cannot replace pictures never drawn. Capture is real time; memory, encoder and device limits can stop it early. Unreadable/timed-out analysis is unknown, not a successful check. Follow [recording troubleshooting](https://phasevoid.com/guides/troubleshooting/#recording).


## Realtime for a performed take

[Open real-time capture](https://phasevoid.com/studio/?flow=video&realtime=1) for microphone/shared-tab input or live MIDI/pad gestures. Choose and start the source deliberately; opening this link does not start sound or request live-input permission. A local file can also be played for this separate performed-take workflow.

1. Rehearse the composition, frame, title/artwork and intended gestures with the audio playing. Prepare saved acts and recover missing clips first.
2. Press **Check this setup** in Capture with Studio visible and uncovered. The short check measures preview draws, not encoded-video FPS or sound; see [output troubleshooting](https://phasevoid.com/guides/troubleshooting/#visual).
3. Make a **10-second test** that stops automatically before a longer **Full recording**. Select an available capture format and press **Start recording**. Keep Studio visible: this records playback in real time, so a five-minute performed song takes roughly five minutes to capture.
4. Play the result with sound, inspect its beginning/middle/end, download and reopen the actual saved file before replacing the in-page copy.

A segment stops at a non-looping track's end, 30 minutes or 750 MB; browser/device limits can stop it earlier. Requested resolution and 30/60 fps are targets, not performance guarantees. There is no automatic YouTube upload or cloud backup. This playback capture is separate from the silent [File render workflow](https://phasevoid.com/guides/youtube-visualizer/#file-render).


The [HTML guide](https://phasevoid.com/guides/youtube-visualizer/#example) contains a playable 12-second first-party WebM with stereo audio, recorded from an original 160 BPM test WAV in desktop Chrome. It proves one short setup, not universal performance. The [aspect-ratio guide](https://phasevoid.com/guides/aspect-ratios/#recorded-ratios) also shows portrait and square recorded examples.


## Timed text / lyrics

Timed text / lyrics in Layers & text edits up to 500 non-overlapping plain-text cues of at most 320 characters, with file-second start/end times, a ±600-second display offset, three positions and three sizes. Plain VTT/SRT timing/text imports replace the track atomically; styling, overlapping cues, word highlighting and automatic transcription are not supported. Session JSON and local Studio settings keep the text. File render captures an immutable track and uses absolute soundtrack seconds, including trimmed starts. Program surfaces include it; independent worlds exclude Program overlays. No file clock is inferred from live input, and no text is sent to MCP or analytics.

1. Open Layers & text → Timed text / lyrics after choosing your MP3/WAV. Add a line, set Start/End in seconds or MM:SS.mmm, type text and Save line.
2. Or import a plain VTT/SRT file up to 256 KB. Import replaces the track only after all cues validate; Undo restores the previous track. No markup, overlapping captions or auto-transcription.
3. Set position/size, turn on Show timed text, then Seek to start (keeps Play/Pause). The static selected-line preview is separate from actual timing. Positive Offset delays display.
4. Review File render again after text changes. Trimmed films use absolute soundtrack seconds. Inspect the finished file; text is burned into pixels, not a selectable subtitle track.

[Open Studio instructions](https://phasevoid.com/studio/?help=timed-text). [WebVTT source specification](https://www.w3.org/TR/webvtt1/); plain timing/text subset only. Use music and words you have permission to include.


## Beat-loop effect modulation

Beat loop is a separate driver in Effects or the selected mapping surface’s Audio links & beat loops editor. Each of four targets—Kaleidoscope, Pixelation, RGB split and FX amount—can use Manual, one existing audio band, or its own beat loop. Loops have Low/High endpoints, a quarter-beat to sixteen-bar period, Cycle offset and Sine, Triangle, Ramp, Alternating or deterministic Step random shape. Reversed endpoints reverse the response. Track sync follows the loaded file’s BPM/aligned beat-one grid through seeks and trimmed renders; other sources use the existing manual BPM grid, starting at film zero for file rendering. Visual timing offset also shifts loop phase. Gentler motion makes loops four times slower. Loops do not detect beats, start audio or alter playback speed. Program/surface loops share the same clock and persist in their settings, maps, sessions and captured acts; they do not reset on each cue or surface selection. Freeze holds final output while the clock can continue. Bypass retains routing, Manual removes one route and Clear links removes every route in that editor. Public MCP describes the feature without reading or operating private Studio settings. [Complete workflow](https://phasevoid.com/guides/live-visuals/#beat-loops).


## Local Stage workspace

Stage view enlarges the operator preview by folding the tools, world library and show-sequence editor. Hide controls also folds the inspector and mapping edit handles; Show controls or an explicit Map/inspector navigation reveals it. The transport, Freeze, Blackout, Record video and Go live remain available. Back to Studio or Escape returns to the full workspace; Escape respects open dialogs, text/select editing and browser fullscreen. A draggable, keyboard-accessible divider adjusts the inspector to a bounded 260–520 px while retaining preview space. Left widens controls, Right widens preview, Shift uses larger steps, Home/End reach the bounds and Enter resets width. On narrow screens controls stack below the picture. View, visibility and preferred width persist locally under phasevoid-workspace-v1, independently of portable sessions. Reset layout restores Studio and 352 px. Layout changes do not start sound, change output pixels, render detail, normalized mapping corners or export dimensions; Stage view is not projector routing or browser fullscreen. [Human workflow, screenshots and verification](https://phasevoid.com/guides/live-visuals/#stage-workspace).


## Independent audio-to-effect links

Audio links independently map Bass, Mid, High, Energy or the detected audio Transient to Kaleidoscope, Pixelation, RGB split and FX amount. Each link has Low/High endpoints, gain, a noise floor and Linear/Smooth/Punch response. Descending endpoints reverse the response. Program links sit in Effects; Mapping has separate links for the selected surface. Links follow the existing shared audio analysis without starting audio or creating another analyser. Sessions, saved compositions/cues and maps retain their respective settings; causal file rendering consumes the current decoded frame. Manual restores the rack value, Clear links removes the routing, Bypass FX disables the rack, and Reset rack also clears links. For repeating BPM-driven movement, choose the separate Beat loop driver. Audio-band mapping is relative signal mapping, not source separation, a new adjustable frequency crossover, independent attack/release timing, a calibrated meter or physical latency correction. [Full human workflow and actual output](https://phasevoid.com/guides/live-visuals/#audio-links).


## Audio spectrum composition

Soundwave opens a free Program layer with frequency Spectrum bars, a radial Ring or an instantaneous Waveform. Choose color, position, size, response, opacity and background darkening without replacing or starting audio. The layer uses the existing MP3/WAV or supported live-input signal, sits above Program video/worlds and below titles/artwork, and reaches the composed projector and realtime recording. File-render first-frame review, causal storyboard and encoded films consume the current decoded signal with frozen layer settings at each actual output size. Local/settings-only sessions retain the layer; legacy sessions default off. Freeze/Blackout cover the complete output. Independent mapped world sources and shared procedural-look links exclude this global layer. This is not a whole-track waveform, calibrated meter, lyric generator, source separation or physical latency correction. [Complete workflow](https://phasevoid.com/guides/audio-spectrum-visualizer/).


## Align the file beat grid

Sound → Timing & beat sync → Align file beat grid decodes a 2–32 second waveform window locally and silently. The operator places beat one, enters 30–300 BPM, inspects later windows for drift and deliberately applies BPM, file alignment and Track sync. Guarded Undo restores the earlier BPM, exact file phase and pulse source. Track sync and file-render reviews/films use the same pulse envelope at absolute soundtrack seconds, including trimmed starts and pulse delay. A different soundtrack cannot borrow the loaded file’s grid. Audio, playback, cue-grid position, mapping and delay stay separate. Alignment lasts in this tab for this file; replacement, session load and reload reset it. No changing tempo map, automatic downbeat detection, audio stretching or physical latency correction. [Full workflow](https://phasevoid.com/guides/music-visualizer/#file-beat-grid).


Already choosing a look? [Preview your local MP3/WAV on the homepage](https://phasevoid.com/guides/music-visualizer/#homepage-local-track), then explicitly pass the same file to a new Studio tab. It arrives paused, with speaker monitoring off. Review the final segment, dimensions and timeline before rendering; the homepage preview is not the exported film.

## Make a film in three clear steps

Studio → Output → Render music video opens a three-step film workflow: Sound, Picture, Review & render. Sound reuses the selected local track or explicitly loads an MP3/WAV or original test WAV; a valid passage of up to 30 minutes is required to continue. Picture keeps the current composition or show sequence, offers Studio frame or selected 16:9/9:16/1:1 outputs, and keeps file format beside the picture choices and places FPS/quality controls and the replacing four-act example in separate optional disclosures. Review prepares actual first frames and, for a sequence, later-act storyboard stills. Rendering remains a separate deliberate action. Back/navigation preserve entries and results; editing still invalidates the prior review. Sound preview monitoring starts off; no step auto-connects, plays, applies an example or starts rendering. Completed files stay available on reopening the dialog. This runs in the loaded web app, with browser codec and resource limits; no offline application or source download is offered.

1. Sound: select a real local soundtrack and a valid passage.
2. Picture: keep a composition or sequence and choose output shapes; disclose encoding/example options only when needed.
3. Review & render: prepare actual frames, correct blocking errors, then explicitly render and inspect the finished file.

Canonical instructions: https://phasevoid.com/guides/youtube-visualizer/#film-steps

## Choose the sound and file you will deliver

Before a long export, read [which sound reaches the saved video](https://phasevoid.com/journal/music-visualizer-video-soundtrack/) and [MP4/WebM delivery choices](https://phasevoid.com/journal/mp4-vs-webm-music-visualizer/). Source, monitoring and output level are distinct; containers, codecs, local export support and destination processing need their own checks. These articles complement the existing Sound → Picture → Review & render path and start no audio, export or upload.