Elysia Tools
Navigation
Workflow Playbook
Prepare one still image and a music excerpt, tune an audio-reactive particle animation, inspect the exported video and deliver MP4, poster, settings and review evidence.
Hubs
This workflow is for music creators and visual designers turning one authorized still and one music excerpt into a reviewed particle clip. It does not edit a multi-shot timeline or prepare an entire social image kit. Agree the reviewer, intended player, canvas and excerpt before changing effects. Keep originals and a manual review record; the Hub is a playbook, not an automatically connected execution pipeline.
Inspect the image's dimensions, orientation and alpha. Use a still PNG, JPEG or WebP no larger than 20 MiB and 25 million pixels. Correct composition with a crop only when necessary, then resize proportionally if the input needs reducing. Do not crop simply because the canvas differs: the visualizer fits the image proportionally. Inspect the actual prepared image, especially camera orientation and small lettering. Retain PNG or WebP for alpha; JPEG cannot preserve it. Transparent pixels are omitted, dark opaque pixels remain eligible and a fully transparent input is rejected. Fine text may remain unreadable in a sampled particle field.
Supported audio is mono or stereo WAV, MP3, FLAC, Ogg, M4A or AAC up to 100 MiB. The chosen duration is 1–30 seconds and start plus duration must not overrun the file. For a simple excerpt, use the visualizer's start and duration without an extra cut. If the ending needs a fade, cut the desired excerpt first and use WAV or FLAC as an intermediate where appropriate; avoid unnecessary lossy generations.
For example, take seconds 42–54 from an authorized track to prepare a 12-second excerpt. Apply a one-second fade-out only to that excerpt if approved, listen to the resulting file, then select start 0 and duration 12 in the visualizer. The fade must be shorter than the excerpt. Applying it to the entire original track would fade its end, not necessarily the selected passage. Keep the mapping back to 42–54 and verify the actual prepared file is long enough. A fade also changes motion-driving energy, so judge picture and sound together.
The canvases are landscape 1280 × 720, portrait 720 × 1280 and square 720 × 720. Fix one source pair, seed and shape, then begin with a short candidate. Choose breathing, drifting or bursting motion; compare one parameter at a time. Particle counts are 1,000, 3,000 or 5,000 requested points, but visible eligible pixels can yield fewer. More points do not restore missing image detail.
Lower motion or trail if the subject smears; raise contrast against the chosen background if dark particles disappear. Smoothing and sensitivity shape visual response, not soundtrack gain. Quiet passages can have little audio response while a preset retains some baseline motion. Neither a moving scene nor a motionless instant proves reliable beat detection. This is image sampling with approximate energy/transient response, not a 3D model or exact musical beat map.
Preview stays in the browser. Export uploads both sources to the site and uses its media queue; preparation and review tools are not promised browser-only either. Obtain permission before sending private artwork or unreleased music. Render the chosen candidate rather than submitting many variants at once.
The output is a 30 fps H.264/AAC MP4 with an opaque background, a PNG made from the first rendered frame and a settings JSON. Audio is not automatically normalized. Download and play the actual MP4 from beginning to end in the target player. Check excerpt identity, duration, framing, sound, boundaries and whether strong motion hides the subject. Browser and server antialiasing may differ, so a preview is not final acceptance.
Extract representative frames at clip-relative timestamps, such as 0, 6 and 11.5 seconds for the 12-second example. This is a limited visual sample, not an exhaustive frame audit or audio test. The generated first-frame poster is not an automatically selected musical highlight. Review it independently and label any alternative extracted cover separately. If a defect occurs, record the time and setting: wrong excerpt means revisit the cut; smeared detail means revisit motion; poor soundtrack level means separate audio work. Do not try to repair audio clipping with visual sensitivity.
Keep the accepted MP4, its first-frame PNG and JSON together, plus the prepared image and audio, original-to-clip mapping, review frames and a manually written approval record. Record seed, preset, dimensions, color/background, particle request, actual count when reported and all changed controls. JSON records settings and analysis data; it does not replace the input files or a review decision and is not a promise of one-click project restoration.
Download actual files rather than sharing only temporary processing URLs. Calculate SHA-256 for the final local artifacts and record reviewer, date, accepted limitations and remaining issues. A seed helps repeat layout, not guarantee a byte-identical MP4; hashes identify the accepted copies. For a static platform image kit, use the related single-photo workflow. For additional timeline editing or format conversion, continue with the video-editing or video-conversion workflow. Platform publication and rights clearance remain separate approvals.
Workflow playbook
Read dimensions, orientation and alpha, then crop or resize only if the composition or limits require it. Check the prepared still visually and preserve the original.
Choose an in-range excerpt. For a controlled ending, cut a separate WAV or FLAC and optionally fade its tail, listen, then use start zero with the prepared file.
Fix shape, seed and source pair. Preview a short candidate, adjust one visual control at a time and export the selected 1–30-second version with its poster and settings.
Download and play the exported MP4 with sound, inspect representative frames and record defects at clip-relative times. Return to the failing preparation or render stage rather than stacking unrelated effects.
The visualizer fits the image without stretching. Crop only to remove unwanted content or improve the subject; resize only when necessary for input limits. Preserve alpha with PNG or WebP when transparency matters.
Use the original and a start/duration for a simple selection. Cut first when a fade-out is needed; fade the resulting excerpt, then start at zero in that new file. Do not apply the original offset twice.
Choose a motion preset and change one parameter at a time. Lower motion, sensitivity or trail if details disappear; increasing visual sensitivity does not repair clipping or normalize the soundtrack.
Save the accepted MP4, poster, settings, prepared sources and review record locally. Name variants, document source timing and approvals and calculate SHA-256 for the actual final files.