Renee's Object Obliterator

Remove small imperfections from your clips using a classic technique trusted by compositors for over a decade.

This is a port of the widely acclaimed Autodesk Flame setup of the same name, created by Renee Tymn. The technique pulls colors from the area surrounding your mask to fill in and hide unwanted elements.

Overview

Renee's Object Obliterator provides fast, controllable object removal:

  • Classic Technique - Proven blur/divide method used in hundreds of commercials and videos
  • Directional Control - Add directional blur to match motion or pull colors from a specific direction
  • Retexture - Steal texture from another part of the frame for seamless blends (originally added by Ted Stanley in his matchbox shader implementation).
  • Grain Matching - Add grain to help the patched area blend with surrounding footage
  • Real-Time Performance - Fast enough for interactive tuning on the timeline
Not AI Inpainting: This tool uses optical blur techniques rather than AI. It won't intelligently reconstruct complex objects, but it's fast, predictable, and works beautifully for removing small imperfections like blemishes, dust, or minor distractions.

Quick Start

  1. Create a mask around the object you want to remove
  2. Track the mask to follow the object through the shot, if needed.
  3. Apply the effect to your clip
  4. Connect your mask to the "Mask Shape" input. In Resolve's color page that is the node's second input (green tab); in Fusion, the Mask Shape port; in After Effects or Premiere, pick the mask layer in the Mask Shape dropdown at the top of the Effect Controls.
  5. Increase the Obliteration-> Blur Amount until the object disappears
  6. Fine-tune with directional blur, retexture, or grain as needed
Best Results: This technique works best on small imperfections surrounded by similar colors and textures - blemishes on skin, dust spots, small distractions in soft backgrounds. For larger or more complex removals, consider AI-based inpainting tools.

Inputs

Input Required Description
Source Yes The image to process
Mask Shape Yes The mask defining the area to obliterate. White = remove, black = keep. Your mask doesn't need to be a perfect digital matte - levels are normalized internally, so slightly-off whites from color-managed or compressed sources work fine.

Parameters

Source Offset

Parameter Type Default Range Description
Source Offset vec2 0.0, 0.0 -2.0 to 2.0 Shifts where colors are pulled from while the mask stays in place

Use Source Offset when the colors immediately surrounding your mask aren't ideal. This lets you pull replacement colors from a different area of the frame - useful when the object you're removing is near an edge or color boundary.

Obliterate

The core controls for the blur/divide operation.

Parameter Type Default Range Description
Blur Amount float 50.0 0.0 - 1000.0 Primary blur amount in pixels. Increase until the masked area fills in.
Directional Amount float 0.0 0.0 - 400.0 Additional directional blur in pixels
Directional Angle float 0.0 -180.0 to 180.0 Angle of directional blur in degrees
Directional Bias float 0.0 -1.0 to 1.0 Shifts where the colors are pulled from. If 0, pulled evenly from both sides; from the left when -1.0, from the right at +1.0 - useful when the object you're removing has a clean area on one side but not the other.
Highlight Unfilled bool false - Tints any not-yet-filled area red so you can see exactly where to add more blur. Off by default so a stray red patch can never sneak into a final render - turn it on while dialing in, off when done.

Blur Amount

This is the main control. If this is at 0, you'll see a black hole where your matte is. As you increase this amount, you'll see the edges of the matte being "sucked in" to the center, removing the object, and replacing it with the pixels surrounding it. Start with the default and increase until the masked area is completely filled in. The right amount depends on the size of your mask - larger masks need more blur, and you might need to animate this value across a shot, as the mask changes shape or size.

Directional Blur

Add directional blur when you want to pull colors from a specific direction. This is particularly useful if there are strong edges in the area you want to preserve - for instance, if the target object is sitting on a shelf. In such a case, using the directional blur, and adjusting the direction to be parallel to the shelf will effectively preserve the shelf, but remove the object. The bias control shifts where the blur samples from along the direction axis.

Post Blur

Optional smoothing applied after the main obliterate operation.

Parameter Type Default Range Description
Enable Post Blur bool false - Enable additional smoothing
Post Blur Amount float 10.0 0.0 - 100.0 Smoothing amount in pixels
Post Directional Amount float 0.0 0.0 - 200.0 Additional directional blur for post smoothing
Post Directional Angle float 0.0 -180.0 to 180.0 Angle of post directional blur
Post Directional Bias float 0.0 -1.0 to 1.0 Shifts post blur center along direction

Post blur can help smooth out any artifacts or "puckering" from the main operation.

Retexture

Transfer texture and fine detail from another part of the frame to make the obliterated area match its surroundings.

Parameter Type Default Range Description
Enable Retexture bool false - Enable texture transfer
Retexture Offset vec2 0.0, 0.0 -2.0 to 2.0 Where to sample texture from
Retexture Blur float 5.0 0.0 - 50.0 Controls which detail frequencies transfer (higher = coarser detail)
Retexture Strength float 100.0 0.0 - 200.0 Intensity of texture transfer as percentage
Beauty Work: Retexture excels at restoring skin texture after removing blemishes. Point the Retexture Offset to a nearby area of clean skin, and the effect will clone the subtle skin texture onto the obliterated area for a more natural result.

Retexture Blur

This controls the frequency of detail that gets transferred: - Lower values (1-5): Fine detail like skin pores and film grain - Higher values (10-20): Coarser texture and patterns

Grain

Add film grain to help the patched area blend with the surrounding footage.

Parameter Type Default Range Description
Enable Grain bool false - Add film grain to the obliterated area
Grain Amount float 0.5 0.0 - 2.0 Intensity of grain
Grain Size float 1.0 0.01 - 5.0 Size/coarseness of grain particles
Grain Chromatic float 0.0 0.0 - 1.0 Color variation (0 = monochrome, 1 = full RGB variation)

Match the grain characteristics of your footage to make the repair invisible.

Edge Handling

Parameter Type Default Range Description
Reflect Edges bool false - Off = clamp/stretch edges, On = reflect edges

Controls how the blur handles samples near the edge of frame. Reflect mode can produce more natural results when working near frame boundaries.

Tips & Tricks

Beauty Work Workflow

For removing blemishes while preserving skin texture:

  1. Create a soft-edged mask around the blemish
  2. Track if necessary
  3. Increase Obliterate->Blur Amount until the blemish disappears
  4. Enable Retexture and point Retexture Offset to nearby clean skin
  5. Adjust Retexture Blur to match the scale of skin texture (typically 3-8)
  6. Fine-tune Retexture Strength (start at 100%, adjust to taste)

Directional Blur

When removing objects from situations where there are strong edges that you don't want to remove:

  1. Use Directional Amount to blur directionally
  2. Tune the Directional Angle to match the angle of the edges you want to keep - like shelves, window frames, etc.
  3. Adjust Directional Bias to pull colors from the leading or trailing edge

When Results Look Too Smooth

If the obliterated area looks too clean compared to surrounding footage:

  1. Enable Grain and match your footage's grain characteristics
  2. Enable Retexture to add back fine detail
  3. Try Post Blur with a small amount to blend edges

Troubleshooting

My mask stopped working after upgrading from version 1.0

  • Version 1.1 renamed the mask input internally (it is now "Mask Shape"). In Resolve and Fusion, a project saved with version 1.0 silently loses the connection to the second (mask) input - the effect then behaves as if the mask were solid white, and the whole frame gets obliterated. Reconnect your mask to the Mask Shape input and it works as before. New projects are not affected. (Version 1.0 had no After Effects or Premiere plugin, so there are no Adobe projects to update.)

Credits

Concept & Original Flame Setup: Renee Tymn - The technique and original Autodesk Flame setup that this effect is based on.

Retexture Feature: Ted Stanley - The idea to incorporate texture transfer within the object obliterator workflow, as seen in his Matchbox shader implementation.

Limitations

This technique works best for: - Small imperfections (blemishes, removing small objects like distant storefront signs, tracking marks, minor distractions) - Areas surrounded by similar colors and textures - Situations where speed and predictability matter

Consider other tools for: - Large or complex object removal - Objects crossing varied backgrounds - Reconstructing detail that doesn't exist elsewhere in frame

Extra Inputs, Mattes & Alpha

Applies to AVFX effects updated August 2026 or later - you'll know your build has it if you see the "Alpha/Matte" parameter group.

This section is shared by the documentation of every AVFX effect - the information here is general, not specific to this effect.

The short version

The AVFX effects framework supports any number of extra inputs in addition to the clip the effect is applied to - for example: a blur map, a custom bokeh shape, an external matte, and so on. In most hosts, each one simply gets its own connection: After Effects and Premiere Pro give every input its own layer selector in the Effect Controls, and in Fusion (and other node-based hosts) each node gets a "port" for each input, as you'd expect.

Alpha is the fourth channel of the image itself - the "A" in RGBA - used for compositing downstream of the effect. It rides along with the RGB through the processing, and embedded alpha behaves just as you'd expect: if your source clip already carries an alpha channel, nothing special is needed.

The one ability AVFX adds everywhere: your input alpha can come from the source's embedded alpha, or from an external matte injected in its place. A matte is any grayscale or RGB clip - its brightness becomes the alpha (the matte clip's own alpha channel is ignored, so it doesn't need one). Unlike other extra inputs, which steer how the effect processes the image, the Matte input feeds this alpha. From there it rides through the processing along with the source's RGB, just like embedded alpha would, and what emerges is the output's alpha - consistent with the processed image, because it went through the same processing.

The Alpha Source selector (in the Alpha/Matte group) chooses where the input alpha comes from: Auto (the effect picks sensibly), Source's Embedded Alpha, Matte Input's Luminance (called Matte Layer's Luminance in After Effects and Premiere), or Opaque (Alpha = 1) - the escape hatch when you just want a solid frame.

Premultiply Output depends on what the effect hands you, so there is no single right setting:

  • Turn it ON when the output is a processed picture carrying a matte in its alpha - the usual case when you feed a Matte input. Without it, Resolve's timeline composite leaves bright fringes or sparkle along the matte edges.
  • Leave it OFF when the output is an element on black, such as Optics Reborn's "/ Alpha" glow modes. Those already have their coverage baked into the color, so switching it on multiplies it in a second time and gives you a dark edge instead.

Rule of thumb: ON if the frame still looks like your shot, OFF if it looks like a glow floating on black.

One host is different: Resolve's color page handles inputs and alpha in its own way, and is described in the next section. If you don't use the color page, skip straight to your host's section below.

Resolve color page

A color-page node accepts at most two connections. The first is always the main image being processed - so every extra input the effect can accept has to share the second connection, and that second input can only serve one function at a time. This is a limitation of Resolve's color page.

That's what the "Resolve Color Page: Input 2 Routing" group is for: choosing which of the effect's extra inputs the second connection serves - a map that steers the processing, or the Matte that feeds the alpha. Take Advanced Lens Blur: there are three possible uses for its second input - the Kernel (an image used as the bokeh shape), the Blur Map (brightness controls blur strength per pixel), or the Matte Input (external alpha for compositing). You can only use one of these at a time, and the "Use Input 2 As" selector is how you choose: connect your clip to the node's second green input, then set the selector to the input it should feed. (The Matte Input is always one of the choices - the framework provides it on every effect. If an effect has no other extra inputs, there's no selector at all: the second input simply is the Matte.)

If you need two of them at once - say, an external matte and a custom kernel - you'll have to use the Fusion page (or another host), where all of the inputs are available simultaneously.

Don't rely on embedded alpha on the color page - its results there are unpredictable (the color page does some things under the hood that are out of reach of the OFX plugin). Feed a matte through Input 2 instead (Use Input 2 As → Matte Input).

Compositing: in Resolve, the color page and the timeline share the job: the actual comp happens on the timeline - your clip sits on a video track above whatever it composites over - while the color page is where the alpha is produced. The timeline only sees that alpha if you explicitly export it from the node graph: right-click in the color page's node area, choose Add Alpha Output, and connect your node's blue alpha output to it. The effect's node already carries the alpha - "OFX Alpha" is enabled by default - but you should also right-click the node and deselect "Use OFX Alpha for Mixing": that setting turns the alpha into a mix mask between the node's input and output instead of compositing coverage, which is almost never what you want.

Resolve/Color Page Troubleshooting:

  • Bright fringes or "sparkle" at matte edges during playback - turn ON Premultiply Output. If you followed the Compositing steps above, "Use OFX Alpha for Mixing" is already off, so that is not the switch that helps here - Premultiply is. (Stills can look clean while motion sparkles: playback renders at reduced resolution, which creates the fractional alpha that exposes the mismatch.)
  • The composite looks like a blend of the clean and processed image at the edges - same cause: "Use OFX Alpha for Mixing" is ON. Turn it off. (If the whole frame looks blended with the source, that's a different knob - check that Global Blend is set correctly.)

After Effects & Premiere Pro

Each of the effect's inputs has its own layer selector in the Effect Controls - pick any layer in your comp (or Premiere sequence) for each one. It's best to precompose vector layers (shapes, text) before using them as inputs.

Compositing: for glow or element outputs, set the layer blend mode to Luminescent Premul - the Normal mode attenuates premultiplied elements twice and leaves a subtle dark edge. Keep Blend at 1 when compositing with the effect's alpha (Blend mixes the untouched source back in, alpha included) - fade with layer opacity downstream instead.

AE Troubleshooting:

  • Dark edge on glows - the layer blend mode is Normal; switch to Luminescent Premul. If it's darker still, Premultiply Output is ON; turn it off.

Fusion & everywhere else

Fusion (and Resolve's Fusion page) gives the node a connection per input, and compositing needs no special setup: the effect's output carries its alpha embedded, just as you'd expect, so wire it into an ordinary Merge node and leave its settings at their defaults.

General Troubleshooting:

  • If the same comp looks different in different hosts - the math is identical in every host, but the color transforms might not be. Make sure you're evaluating everything through the same color pipeline. (In Resolve, that means pinning the project's Timeline and Output color space to your delivery standard before comparing.)