An effect is one entry in a clip's effects list: a type, and only the settings you want to change. Each page below shows one on footage chosen for it, next to the same request without it, with the settings, their defaults, and the SDK call that rendered it.
Most are shown with nothing but their type. The defaults are the look.
Grade, soften, sharpen and finish the picture.
Color 
color A warm cinematic grade: the light pulled towards gold, contrast and color lifted a little, skin left believable. 
LUT lut A film grade from a LUT file: shadows toward teal, skin and highlights toward orange, the look a colourist hands over as a .cube. 
Looks look The whole archive look in one word: sepia, grain, dust, scratches, projector flicker and a vignette together. 
Blur blur Knock the footage back so a title over it stays readable. 
Sharpen sharpen Crisp up footage that is slightly out of focus. Look at the rocks and the life jackets. 
Denoise denoise Clean the grain a small sensor adds in the dark. Punched in so the grain is actually visible at this size, because at full width it is too fine to see. 
Glow glow Make the lights bloom on a city at night. 
Film grain grain A light film stock, so digital footage feels less clinical. 
Vignette vignette Darken the corners to pull the eye to the middle of the frame.Deliberate looks for a beat, a cut or a whole short.
Camera shake shake Handheld energy on a steady shot. The frame's edges never show. 
Glitch glitch Digital break-up for a beat or a scene change, kept light enough that the shot underneath still reads. 
Pixelate pixelate Large square pixels over the whole picture. 
Posterize posterize Flatten the picture into a few steps of each color, like a screen print. 
RGB shift rgbShift Red and blue pulled apart sideways for an analog edge, kept to a fringe so the shot underneath still reads. 
Scan lines scanLines The dark lines of a CRT monitor. 
Sketch sketch The picture as a pencil drawing. 
Trails trails Traffic leaves streaks behind it. The camera has to be locked off, or everything smears.Keys, outlines and shadows for anything placed over something else.
Mask mask The footage seen through the wordmark: a logo reveal, with the logo's own transparency as the matte. 
Chroma key chromaKey Remove the green screen so the product sits on a new background. 
Region blur blurRegion Hide one part of the picture and leave the rest sharp: here what is on a laptop screen. 
Outline outline The sticker look: a keyed product with a thick white edge and a shadow under it. 
Drop shadow dropShadow Keep a white logo readable over bright water.Clean, level and color a voice or a music bed.
eq More low end and more sparkle on a music bed. Compressor compressor Even out a voice so quiet words and loud words sit at one level. Limiter limiter Make music louder without letting the peaks clip. Noise gate gate Close the microphone when nobody is speaking, so the river drops away before and after the voice. High pass highPass Cut the rumble under a voice. Wind, traffic and water live there. Low pass lowPass Music from the next room. Loudness loudness Bring a voice recorded far too quietly up to a streaming loudness. Pitch pitch Lower a voice without changing its speed. Echo echo A quick repeat behind the voice. Reverb reverb Place a dry voice in a hall. Voice noise removal voiceDenoise Remove the river from under the voice, including while it speaks. 
Visualizer visualizer The audiogram: the narration drawn as a moving line over the picture. On the timeline it draws the final mix.Move from one clip to the next.
Crossfade crossfade One picture dissolves into the next. The everyday soft join, and the only decorative one worth using often. 
Fade fade Both clips pass through the background colour, so the cut reads as a chapter break rather than a join. 
Slide slide The new clip slides in over the old one, which stays where it is. Compare it with push, where both move together. 
Zoom zoom The frame scales into the next clip. Energetic, and worth keeping to section boundaries. 
Wipe wipe Reveal the next clip in place through a growing circle. Nothing moves: the new clip is uncovered. 
Push push The next clip pushes the last one out of frame, both moving together.Animate any numeric value across a clip, with 33 easings.
How a picture sits in its box when their shapes differ.
Position 
position Where the clip sits in the frame. x and y are the clip's centre by default, as a percent of the frame or in pixels. 
Scale scale How big the clip is drawn, as a multiple of its natural size. 0.34 makes it a picture in picture. 
Rotate rotate Turn the clip, in degrees, about its anchor. A few degrees reads as a photograph laid on a desk. 
Anchor anchor What `position` points at. With topLeft the clip's corner lands on the point instead of its middle. 
Size size The clip's own box, in percent or pixels. Give it a shape the source does not have and `fit: "cover"` crops the picture into it, which is how a wide shot becomes a tall panel. 
Corner pin corners Pin the clip's four corners to four points, so footage sits on a surface in the shot. Here the river plays on the laptop screen. 
Rounded corners crop-radius Round the corners of what the crop keeps, 0 square to 1 fully round. The corners ride the transform, so a scaled-down clip becomes a rounded card. 
Crop crop Trim the edges of the source before anything else happens, to punch in on the subject. 
Flip flip Mirror the clip left to right, or top to bottom, so two shots can face each other. 
Safe area safe area The same two lines, once pushed to the edges where a platform's own buttons cover them, and once inside the safe area: x 8 to 92 percent, y 13.5 to 83. 
Fit fit A wide photograph in a phone-shaped box. contain shows all of it and leaves bars, cover fills the box and cuts what overflows.How a clip sits over what is under it.
Kinds of clip the engine draws for you, styled like text.
Text stroke 
text-stroke An outline round the glyphs. White text over bright water is unreadable without one, and a 4 to 6 pixel black stroke fixes it. 
Text shadow text-shadow A shadow behind the text. Softer than a stroke, and enough on footage that is not fighting the text. 
Text background text-background A solid box behind the text. The most readable of the three, and the one that reads as a caption rather than a title. 
Text align text-align Which way the lines line up, and which edge of the text sits on the clip's position. left starts there, right ends there. 
Text entrance text-animate An entrance. The title slides up into place while it fades in, instead of simply being there. 
Shape gradient shape-gradient Two colours instead of one makes the fill a gradient, running the way `gradient` names. 
Shape outline shape-outline Leave the fill out and only the stroke draws, so what it surrounds is genuinely transparent and the footage shows through. 
Shape radius shape-radius Rounds a rect's corners, 0 square through to 1 fully round, which is a pill on a wide box. 
Shape softness shape-softness Feathers the edge instead of cutting it, which is how a soft spotlight or a vignette is made rather than a hard-edged block. 
Shape shadow shape-shadow A shadow behind the shape, so a card or a bar lifts off the footage instead of sitting flat on it. 
SVG path path SVG path data straight out of a design tool, drawn as a clip. The viewBox is the space the path was written in. 
Shape kinds shape The kinds a shape can be. A circle takes one side of its box so it stays round on any frame, and a triangle points down until you turn it. 
Background background The colour behind every clip, wherever nothing else is drawn. Here a wide shot in a tall box leaves bars, and the background decides what colour they are. 
Scenes scenes Sequential segments instead of tracks. Each scene holds its own clips and an optional transition into it, which is the shape most edits actually have. 
Overlays overlays Clips drawn over the WHOLE video however it is cut, so a watermark spans every scene without being repeated in each one. 
Composition composition A whole timeline used as one clip, so a group of layers moves, scales and fades together instead of one at a time. 
Lower third lower third The everyday pattern: a rounded bar on one track and the words on another, so each can be styled and timed on its own. 
Progress bar progress bar Keyframes on the box itself. The bar's width runs from nothing to most of the frame across the clip, which is a progress bar with no new field to learn. 
Timer timer A running countdown over the clip. It counts down from the clip's own length, so a 6 second clip starts at 6. 
Fonts fonts One text clip per script. A display family for the headline, and the bundled fallbacks carrying Arabic and Devanagari without a single font file uploaded. 
Lottie animation animation A Lottie lower third that sweeps in over the footage, with the words on it set as a compose text clip. 
Subtitles subtitles Every line of a narration in one clip, timed to the words, styled once.The engine's other filters, by their own names.
Give the clip an effects list and put one entry in it per effect, each with a type and only the settings you want to change. The list runs in the order you write it, and the same list works on video, image and sound clips.
34 named effects for picture and sound, the wipe and push transitions, timer and subtitle clips, and 335 more engine filters you can call by their own names. Every one of them has a page here with a before and after.
Yes. Each request on these pages went through the same parser and compiler a compose job uses and was rendered in the same engine build, MLT 7.41.0. Nothing here is a mockup, and every request can be sent to the Rendobar API as written.
Every clip and sound is public domain or CC0, from Wikimedia Commons and LibriVox, and each page credits its sources. None of it shows an identifiable person.
New to compose? Start with the compose API overview, or send timelines from an agent through the MCP server.
