Every effect available in Duplyfi, what it does, which part of the file it changes, and the recommended range drawn from the built-in presets. All effects randomize their parameters independently per output file, so every copy comes out uniquely different from the source and from every other copy.
Not sure which effects to enable?
See which effects to use per platform →Adjusts how vivid the colors in the video appear. A value above 1.0 boosts color intensity; below 1.0 reduces it. Each output file gets a randomly selected value within the configured range.
Shifting saturation changes the relative weight of each color channel across every frame. Because the value is randomized per copy, each output file has a distinct color signature, making it visually unique from the source and from the other copies.
Adjusts the difference between the darkest and lightest areas of the frame. Higher contrast makes shadows darker and highlights brighter.
Like saturation, contrast changes the brightness distribution across every frame. A randomized per-copy adjustment gives each output file a different tonal balance, adding to its visual uniqueness.
Adds or subtracts a fixed offset from every pixel's luminance value. Positive values make the frame brighter; negative values darken it.
Shifts the overall brightness level of every frame. A small randomized offset per copy changes the tonal profile of each output file, contributing to its visual uniqueness.
Darkens the frame toward the edges, creating a radial gradient from the center outward. Adds a visual region that was not present in the source.
Adds a new darkening gradient in the frame border region that did not exist in the source. Because this changes the corners and edges of every frame, it gives each output file a distinct visual signature from the original.
Applies a non-linear tonal adjustment to the luma channel. Higher values brighten midtones; lower values darken them. Unlike brightness, which is linear, gamma creates a curve in the tonal response.
Non-linear tonal changes reshape midtones differently from a flat brightness shift, so a randomized gamma value gives each copy a distinct tonal curve and adds to its visual uniqueness.
Applies a custom tone curve to the luma channel of every frame. The curve is randomized per file: the midtone control point shifts by ±4 values (in the 124–132 range), and two additional control points in the shadows and highlights shift independently. The visual difference is less than 2% change in midtone response.
A tone curve reshapes the brightness of every frame. Even a 4-point shift in the midtone curve changes the tonal profile of the whole video, so each copy is measurably different from the source with no visible quality change.
Changes the playback speed of the video. A value of 1.03 means the video plays 3% faster than the source. The audio is pitch-compensated automatically to prevent the chipmunk effect.
A speed change alters the duration, the total frame count, and the timing of every frame relative to the source. A randomized per-copy speed value makes each output file run for a slightly different length with a different frame timing, adding to its uniqueness.
Crops a center portion of the frame and scales it back up to the original dimensions. A zoom value of 1.05 crops the outer 5% on each side, then upscales the remaining 90% of the frame area. The effective field of view narrows.
Zoom changes what content sits in every pixel position. The value is randomized per file within the configured range, so each copy has a slightly different field of view, a real geometric difference from the source while keeping the subject centered.
Rotates the entire frame by a fixed number of degrees. Small rotation values (1–2 degrees) create a slight tilt. The area outside the rotated frame is filled with black.
Changes the geometric layout of every pixel in the frame. Even a 1-degree rotation moves every pixel position, making the output measurably different from the source.
Mirrors the video horizontally, reversing left and right.
Flipping is a structural geometric change that reverses the left-right layout of every frame. On its own it is a mild change, since a mirrored frame can still look similar to the original, it works best as part of a multi-effect combination.
Scales the output video to a specified width and height, regardless of the original dimensions. This is a fixed resize; the aspect ratio is not preserved if width and height have a different ratio than the source.
Changing output dimensions affects both the file's byte-level data and the scale of every visual feature in the frame, contributing to each copy's uniqueness.
Scales the video to a randomly chosen width from a set of standard 9:16 bases (1072, 1076, 1080, 1084, or 1088 pixels wide), then calculates the matching height. Each output file gets a different resolution from this set.
Varying output dimensions produces different compression artifacts and a different pixel grid for each file. Two files at 1080×1920 and 1084×1928 differ at the pixel level, so each copy comes out unique.
Trims a percentage of the video from the beginning. A value of 0.05 removes the first 5% of the clip's duration. For a 60-second clip, that's 3 seconds removed from the start.
Trimming the start changes which frame opens the video and shifts the timing of every frame that follows, giving each copy a different starting point and overall duration.
Trims a percentage of the video from the end. Works identically to Cut Start but removes frames from the tail. The two values are randomized independently per file.
Changes the closing boundary of the clip. Combined with Cut Start, each output file has both a different opening and closing point. No two copies trim by exactly the same amount.
Adds a randomized grain pattern to every frame. The grain is generated with a unique random seed per output file, meaning every copy has a genuinely different noise pattern. Three noise modes are applied simultaneously: temporal (varies frame to frame), uniform distribution, and pseudo-random mode.
Unique per-file noise seeds ensure that two copies of the same source have different grain patterns in every frame. Because the grain also varies frame-to-frame within each file, every copy ends up with a genuinely different pixel-level pattern from the source and from the others.
Applies a chroma shift: the blue-difference (Cb) and red-difference (Cr) channels are offset by 1 pixel horizontally and vertically in opposite directions. The luma channel is unaffected.
The slight lateral offset of the color channels creates a subtle chromatic aberration that changes every pixel's color values without affecting the perceived sharpness of the image. This gives each output a different color signature.
Delays the audio track by a fixed millisecond value (the configured value × 10ms). Both audio channels receive the same delay, keeping them in sync with each other.
Changes the time position of the audio track relative to the video. On its own this is a mild change, since it only shifts the whole track in time. It is most useful as a supplemental effect alongside Spectral Tilt, which alters the audio profile more substantially.
Creates a composite frame using a blurred, zoomed version of the video as the background and the original video (slightly shrunk) as the foreground. The border area of the output frame contains a blurred version of the video content rather than black bars or empty space.
Adds new visual content in the border region that was not present in the source. Because the perimeter of every frame now contains a distinct blurred background, each output looks meaningfully different from the original across the whole frame.
Applies a subtle projective transform to the entire frame. Each of the four corners of the frame is shifted by a random small percentage of the frame dimension (0–1.5% for corner positions). At these levels, the effect is equivalent to watching the video on a very slightly non-rectangular screen, invisible to human viewers.
Projective transforms change the geometric layout of every pixel in the frame. Even a 1% corner shift means the pixel at position (100, 100) in the source lands at a slightly different position in the output. This is a genuine geometric change that color and brightness adjustments cannot replicate, so it adds a strong, independent layer of per-copy uniqueness.
Adjusts the overall audio level of the output. A value of 1.3 increases the volume to 130% of the original. Randomized per file within the configured range.
Adjusting volume changes the overall amplitude of the audio signal. A randomized per-copy level gives each output file a different audio profile, contributing to its uniqueness.
Shifts the audio pitch up or down by the configured number of semitones. The time-stretching compensation chain keeps playback speed unchanged. A value of +1 raises pitch by one semitone; -1 lowers it by one semitone.
Pitch shifting changes the frequency of every audio component, giving each copy a subtly different audio profile. On its own it is a mild change, so it is most effective combined with Spectral Tilt and Reverb.
Applies two parametric equalizer bands per file, each with randomized gain values. The first boosts high-frequency energy around 8kHz by +1.5 to +4.5 dB. The second cuts low-frequency energy around 150Hz by -0.5 to -3.0 dB. Both values are randomized independently per file.
Redistributing energy across the spectrum (boosting high bands, cutting low bands) reshapes the frequency balance of the audio. With randomized gain values per file, each copy ends up with a measurably different audio profile from the source and from the others.
Applies an echo effect with a randomized delay time and moderate decay. Implemented as a single-tap echo with input gain 0.8 and mixing gain 0.88. Delay values: 60–100ms. Decay values: 0.4–0.6 based on delay.
Reverb adds a short decaying reflection after each audio event, changing the time-domain shape of the audio. Combined with Spectral Tilt, it differentiates the audio profile across both frequency and time, adding to each copy's uniqueness.
Sets the output audio encoding bitrate. A value of 192 kbps produces higher-quality audio than 128 kbps but a larger file. Randomized within the configured range per file.
Different bitrates produce different compression artifacts in the audio stream, changing the byte-level data of the audio. Not a primary uniquification effect, but it contributes to file-level differentiation.
Sets the output framerate. A value of 60 outputs at 60 frames per second regardless of the source framerate.
Changing the framerate changes the number of frames in the output file, the timing of each frame, and the file's byte-level data, differentiating each copy at the container and timing level.
Sets the output video encoding bitrate in kilobits per second. Higher values produce larger files with less compression artifacting. Randomized per file within the configured range.
Different bitrates produce different compression patterns in the encoded stream, differentiating each copy at the codec output level.
Injects a complete set of US device metadata into every output file. Randomly selects: a US city from a pool of 50 major cities, GPS coordinates with ~1km jitter, a device model (iPhone 13–16, Samsung Galaxy A-series, or Google Pixel 7–8), iOS or Android software version, a recent creation timestamp (random time within the past 14 days), timezone, and ISO 6709 location string.
Every file carries container metadata describing the device, location, and capture time. Randomized per-file metadata means each output file carries a distinct, self-consistent set of device details instead of repeating the source file's metadata.
The image pipeline is separate from the video pipeline and uses Sharp (libvips). Effects are applied in a fixed order: crop, zoom, rotation, flip, pixel shift, color adjustments, noise, vignette, blurred border, format conversion, file size reduction. All effects randomize per file.
Adjusts color saturation for still images using Sharp's modulate function. Values above 1.0 increase color intensity; below 1.0 reduce it.
Changes the color channel balance in the image, giving each output a distinct color signature from the source.
Adjusts image contrast using a linear multiplier applied to pixel values with an offset to keep the midpoint centered. Uses Sharp's linear function.
Changes the brightness distribution across the image, contributing to each output's visual uniqueness.
Adjusts the overall brightness of the image by modifying the luminosity multiplier in Sharp's modulate function.
Shifts the overall brightness level across all pixels, contributing to each output's visual uniqueness.
Applies a gamma correction curve to the image using Sharp's gamma function. Values above 1.0 brighten midtones; below 1.0 darken them.
Non-linear tonal adjustment reshapes midtones more than a flat brightness change, giving each output a more distinct tonal curve.
Applies a radial darkening gradient toward the image edges by compositing an SVG-generated radial mask using the multiply blend mode.
Adds a new darkening gradient in the border region of the image that did not exist in the source, changing the corners and edges and adding to the output's visual uniqueness.
Crops the center of the image to a smaller region and scales back to the original dimensions, simulating a zoom-in. Uses Sharp's extract followed by resize.
Changes the framing and effectively moves every pixel position in the output relative to the source. A strong geometric differentiator.
Rotates the image by the specified number of degrees. The area outside the rotated content has a transparent or black background depending on format.
Changes the geometric orientation of all content, making the output measurably different from the source.
Mirrors the image horizontally, vertically, or both. Uses Sharp's flop (horizontal) and flip (vertical) functions.
Structural geometric change that affects all pixel positions. Most effective as part of a multi-effect combination.
Crops a percentage of the image from all four sides simultaneously, removing the specified fraction from each edge. For example, a value of 3 removes 3% from the left, right, top, and bottom.
Removes the border region of the image entirely, changing the composition and what appears at every pixel position in the output.
Shifts the image content diagonally by offsetting the canvas and re-cropping to the original dimensions. The shift amount is the same in both X and Y directions.
Changes every pixel's position in the output, making it geometrically different from the source.
Adds randomly generated grain to every pixel in the image. The noise value is the maximum amplitude of the per-pixel random offset; each pixel can shift by up to this many luminance units. Applied per-channel directly to the raw pixel buffer.
Random noise with a unique per-run distribution produces files that look very similar but carry different pixel-level values, making each output unique.
Creates a composite image where the background is a blurred, upscaled version of the source image, and the foreground is the original image centered on that background. The border area shows a blurred version of the content.
Adds new visual content in the border region of the output image. A strong geometric differentiator because the perimeter of the image is entirely different from the source.
Controls the output quality (and therefore file size) of the encoded image. For JPEG: sets the quality parameter (0–100). For PNG: sets the compression level. For WebP: sets the quality parameter. Higher values produce larger, higher-quality files.
Different compression quality levels produce different encoding artifacts at the pixel level, differentiating each file at the byte level and adding a small amount of visual variation.
Questions about a specific effect? Ask in Telegram or email support@duplyfi.com.