Neonify

Neonify turns images, videos, audio and 3D meshes into neon art. It runs locally on your machine, no accounts, no uploads.

Images: edges become glowing tubes, PNG transparency follows the source art, and the original can sit under the effect three ways — wiped to black, kept with neon only on the edges, or kept under the full global glow. Videos: the same treatment per frame, with an option to run the audio through it too. Audio: seven profiles (fire, ice, robotic, ghost, void, echo, slash) that each react to the Glow setting, plus an intensity level — normal, high (the chain runs twice), extreme (four times). 3D: OBJ / PLY / STL rendered as neon wireframes, still or as a 360° turntable video, plus image-to-relief remeshing with OBJ export.

On first run the binary probes the machine once — graphics device, opencl, cuda through the nvidia driver, a real cpu/gpu benchmark, the video encoders ffmpeg can actually drive here, and a real hardware-decode pass — and writes neonify.ini next to the binary. Every value in that file carries two comment lines (what it is, what ranges are valid) and a checker repairs corrupt or extra entries back to defaults without touching the rest. Delete the file and the next run detects everything fresh.

Download

Get a package from the releases, extract, run:

• neonify.exe — opens the GUI (Windows)
• cli.bat — interactive CLI (Windows)
• ./neonify — GUI (Linux), ./neonify cli — interactive CLI

ffmpeg is needed for video and audio only. Install it once: winget install FFmpeg on Windows, sudo apt install ffmpeg on Linux.

The Python source under src/python still runs from source if you have the dependencies, but it is frozen — the native build is the product and the only thing the releases ship.

The GUI

Left: input files, each with Preview and Delete. Right: results, each with Preview and Compare. Center: the preview — zoomable image viewer, video player, audio waveform player, and a 10-second 360° orbit render for 3D inputs. Bottom: settings for the files you queued, on 0–100 scales. Compare shows the result next to its original — shared zoom and pan for images, synced seek and speed for videos, position sync for audio with waveforms.

CLI

neonify image photo.png --palette ice
neonify video clip.mp4 --neon-audio --profile ghost
neonify audio track.mp3 --profile fire
neonify mesh model.obj --turntable 240
neonify batch ./photos
neonify profiles

Without a command, neonify opens the GUI. neonify cli opens the interactive CLI.

Options

Option	Default	Description
--palette	electric	electric, synthwave, toxic, ice, fire, ghost, spectrum (rainbow edges)
--glow	1.0	glow intensity 0.1–3.0
--threshold	0.12	edge sensitivity 0.02–0.5
--env	1.0	ambient detail 0–2
-o, --output	auto	output path
--next-to-input	off	save next to the input file instead of results/
--keep-inside	off	keep the original, neon only on the edges (images/videos)
--global-glow	off	keep the original under the full global glow field
--profile	slash	audio profile (audio mode, or video with --neon-audio)
--intensity <level>	normal	audio profile strength: normal, high (x2), extreme (x4) — non-default levels show in the output name (echo_x4)
--neon-audio	off	neonify the audio with the video
--no-spatial	off	disable spatial glow (stereo pan)
--advanced-audio	—	JSON overrides for profile parameters
--turntable <n>	0	3d turntable frames (0 = still)
--azimuth / --elevation	30 / 20	3d view angles
--depth	0.85	relief depth 0.1–3.0
--export-mesh	off	relief runs also write the remeshed OBJ
--json-progress	off	progress as JSON lines (for the GUI)

Showcase

Real runs of the release binary — every result below came out of the same package the releases ship, nothing mocked. Sources and outputs live in showcase/.

Image — Forza Horizon 6 in-game frame through electric: original (left), neonified (right)

	file	run
video	video-original.mp4 → video-neonify-electric-spatial-robotic.mp4	20s of Forza Horizon 6 gameplay (stereo engine audio, no voiceover) through electric with spatial glow, the audio through robotic
audio	audio-original.mp3 → audio-neonify-echo-x4.wav	a vocals-carrying NoCopyrightSounds track through echo at the new extreme intensity (the chain runs four times — the name carries the level)
3d	mesh-original-orb.obj → mesh-neonify-electric-turntable.mp4 · mesh-neonify-fire-still.png	a generated orb — geodesic core, two rings, twelve icosahedral spikes, six pods — as a 48-frame electric 360° turntable and a fire still

Outputs

Default outputs land in results/, named name_neonify_effect_timestamp — photo_neonify_ice_261003152708.png, track_neonify_fire_261003153228.wav, clip_neonify_synthwave_261003154510.mp4. The interactive CLI also offers a custom output path as a third location. Nothing ever overwrites.

Requirements

Any recent desktop CPU. A GPU runs the neon math when the first-run probe proves it is actually faster here (an nvidia card with cuda — the gt 1030 included — reaches it through opencl, and its nvdec decoder can take ffmpeg's decode work even where no hardware encoder exists), and ffmpeg can get hardware encoding when the machine has it — the ini records what was really detected. ffmpeg for video and audio. Windows 10+ or Linux.

License

MIT — see LICENSE.