How to Generate B-Roll That Matches Footage You Already Shot

Build the missing shot in Dreamina before adding motion

Start with Dreamina to prepare a matching B-roll frame, animate that approved frame and refine the clip in your edit. This gives you a practical way to handle an insert whose subject or framing differs from the footage you already have.

Suppose you filmed an interview with a potter but need a close-up of hands shaping a bowl. Use your original frames to guide the workshop’s light, material colors and atmosphere, then prepare the close-up as a separate image. Approve its bowl, hands, apron and background before asking them to move.

In the image-to-video stage, direct one clear action and a camera behavior that suits the surrounding sequence. Dreamina’s reference-led options vary by model and workflow; use a supported video reference when the movement itself needs guidance from a clip. A still can show the look, but it does not contain the camera’s path through time.

Dreamina’s Free entry tier lets you begin with available creation credits. Start with one useful insert and place it in your edit before building a larger set of shots. That first comparison gives you a clear target for subsequent refinements.

Choose the reference that controls the part you need

Use an image for appearance, a start frame for opening composition and a suitable video reference for camera movement. The same file can be useful in more than one role, but the selected control determines how the generator uses it.

For a new angle, prepare or generate that angle first. If you animate a wide view, a prompt for “close-up of the hands” may cause a move toward the hands rather than give you a close-up at the opening frame.

Read the look of the footage before writing prompts

Export a few representative frames and watch a short section of the original footage at normal speed. Choose frames from the lighting setup you want to match, including a face or important material, the environment and a useful detail view where available.

Look for specifics you can carry into the new shot:

• Light direction and quality: which side is brighter, how soft the shadows are and which practical lamps are visible.

• Color relationships: warm skin against a cooler background, muted foliage, a distinctive product color or neutral whites.

• Camera position: viewing height, angle, distance and how much foreground or background is visible.

• Focus and perspective: how quickly the background falls out of focus and whether nearby objects look stretched or compressed.

• Movement: static, gently handheld, pushing forward, tracking sideways or another deliberate behavior.

• Texture: crisp digital detail, diffusion, grain and motion blur.

Use camera metadata or shooting notes for technical settings you know. Describe a wide or compressed look when that is what you can observe; a frame alone does not reliably identify a focal length, sensor, shutter setting or measured light temperature.

Turn those observations into a compact matching brief

Keep one short matching brief for each lighting setup, then change the shot-specific action and framing. A compact brief is easier to reuse and refine than a growing list of camera-brand names and visual adjectives.

Here is an example for the pottery interview:

Adjust this brief when the scene changes. An interview by a window and a later exterior demonstration may belong to the same film while requiring different light and color references.

Make a shot list that serves the edit

Choose a few short cutaways that support the spoken idea and can fit between the shots you already have. Each shot should do a job in the sequence, such as showing a process, clarifying an object or giving the viewer a moment to absorb a line.

Plan the required screen time, including a little usable material around the action. If an insert needs two seconds in a 24 fps timeline, that is 48 frames in the edit; you may want a longer generated take to give yourself trimming room.

Use generated coverage in a way that fits the production. An illustrative process insert should not imply that a person performed a specific recorded action when they did not.

Prepare the exact first frame of each new shot

Create and approve the new composition before sending it to image-to-video. This is where you solve the subject, light direction, camera height, framing and major material details.

Give each reference a clear purpose. One might establish the workshop and its light; another might show the bowl or apron. If a tool supports a sketch or layout reference, use it to place the workbench and hands without asking the video stage to invent the composition later.

A still-image request for the pottery insert could be:

Create a close-up at workbench height of the potter’s hands beside the rim of a pale clay bowl. Use the workshop reference for the soft light from camera left, the warm lamp in the distant background and the muted brown-and-teal palette. Keep the bowl’s shape and the dark teal apron consistent with their references. The hands are clearly formed, the bowl rim is unobstructed and the distant shelves are softly out of focus.

Inspect the image before continuing. Correct an extra finger, a different bowl, a reversed light direction or an awkward crop now. Animating an image with a visible defect can carry that defect into several seconds of footage.

A warm-lit conservatory, from still image to moving shot

This Dreamina example shows how a still image can carry a warm interior light and a blue dusk setting into a moving shot.

The image establishes the warm doorway, cool glasshouse surroundings and the view down the path. The five-second clip below, created with Dreamina Seedance 2.0 Mini, develops that scene through a forward camera move toward the door and handle.

Look at the relationship between the amber light near the doorway and the cooler blue beyond it as the framing changes. That relationship is a useful part of a matching brief.

For an insert that should begin directly on the handle, prepare a handle close-up first and animate that image. The example is a created scene from a still reference; choosing a matching reference frame is the next step when applying this workflow to your own footage.

Prompt the motion once the appearance is right

Let the approved image carry the look, and use a short motion prompt to direct the action and camera. You do not need to describe every visible object again when the first frame already establishes it.

For the approved pottery close-up, begin with:

The hands turn the bowl slowly and smooth its rim. The camera remains locked off. One continuous shot.

If the action works and you want a little more movement, change one part:

The hands turn the bowl slowly and smooth its rim as the camera makes a short, gentle push-in. One continuous shot.

Consider five kinds of motion information: the subject’s action, movement in the environment, camera movement, timing and the direction or speed of movement. Include the parts that matter to the shot. Start simply, then add a detail to solve a specific problem.

For a multimodal workflow, name the role of each supplied asset. A motion clip may guide the camera, while the approved still provides the scene. Keep the instructions complementary: a locked tripod view and a drifting handheld camera describe different intentions.

Keep the clip to one readable action

Give a short B-roll shot one clear action and enough time to begin and finish. A hand lifting a brush is easier to fit into a brief insert than lifting it, painting the bowl, putting it down and walking out of the room.

You can describe a small sequence in natural order: “The hand reaches toward the rack, lifts one brush and holds it briefly.” Approximate timing can help organize the request, but leave enough time for the movement to be readable and inspect the actual result before choosing the cut.

Ask for a continuous shot when that is what the edit needs. Avoid wording that introduces another scene, a sudden angle change or several incompatible camera moves.

Generate a little more than the exact duration you intend to use where the available controls allow it. The useful beginning or end may need trimming, and a small handle can make the difference between an awkward insert and a clean cut.

Choose a generator by the control you need

Start with Dreamina for an integrated frame-to-video workflow, then compare alternatives by their reference controls.

Choose the actual model and mode that provide the needed control. A platform’s model menu can contain several products whose upload, reference and editing behavior differ. For a broader selection, see our guide to AI video models compared.

Runway: create the new frame, then animate it

Runway is useful when you want to develop new compositions with saved image references before generating motion.

In its Gen-4 Image References workflow, open Tool > Image > References and use up to three active references. Assign each a purpose, such as the workshop, bowl or person. Tag references you want to keep across sessions; temporary uploads do not automatically appear in a new session.

Develop the composition as an image, choosing the result with the right props and lighting. For a complex change, refine the subject and environment separately before bringing the approved elements together. Keep an earlier approved reference available so repeated revisions do not quietly redesign the setting.

Then load the accepted image into the video workflow. The image defines the opening view, while the prompt directs the motion. Runway’s current image-to-video guidance emphasizes starting with a clear motion request and adding detail as needed.

Kling: assign a clear role to each reference

Kling is useful when appearance and camera movement need guidance from different reference assets.

In Kling 4.0’s Omni Reference workflow, tell the model what each asset contributes. For example, one image can define the workshop’s appearance while a video guides the camera’s path and pace. Use the uploaded asset identifiers so the roles stay clear.

Keyframes provide a separate route when particular visual moments should be defined by images. Choose that route for a planned opening or ending view; the Omni Reference and Keyframes entry points do not combine all their controls within one generation task.

This makes mode selection part of the creative decision. Use a reference clip when its movement is the useful ingredient, and check the resulting direction, speed and framing against the shot you need.

Firefly: guide camera movement with a reference clip

Firefly is useful when you have a short clip whose camera movement you want to use as a guide.

In Generate video, select Firefly Video, then upload the clip under Camera > Motion reference. Prepare a 5–10-second file under 200 MB, with the useful movement in the first five seconds; that is the portion used by this control.

Add a starting image through Frames when the opening composition matters too. The still establishes that view, while the motion reference guides the camera behavior. Style presets may be unavailable when a frame reference is supplied.

Use these instructions for the native Firefly Video workflow. Partner models available inside Firefly can have different controls. Preview the generated clip against the original movement before treating it as a usable match.

Google Flow with Veo: separate frames from ingredients

Google Flow is useful when you want to choose between a defined starting frame and reference ingredients.

Select the model and video workflow in the prompt box. Video > Frames lets you supply a start frame, end frame or both, then describe the action between them. Use an approved B-roll composition when the opening needs to be specific.

Video > Ingredients uses reference assets to guide elements within the clip. Explain which reference supplies the subject, object or visual treatment. A set of compatible references is easier to direct than several images with conflicting light and design.

Check the features available for the selected Veo model and your region. Keep those interface choices separate from model settings described for a developer API or another Google app.

Luma Dream Machine: work from an existing motion clip

Luma Dream Machine is useful when you already have a motion clip that can serve as the basis for a modified shot.

Its Modify workflow gives you a way to transform material while using an existing clip as the underlying moving composition. This can be useful when you have a suitable camera move but want to change part of the visual treatment.

In the documented Ray3 Modify controls, Adhere keeps the result more closely tied to source edges and detail, while Reimagine permits a broader transformation. Higher strength can also reinterpret camera movement, so choose a restrained setting when continuity with the source is important.

Check the whole clip after a change. A convincing first frame does not tell you whether a prop stays consistent or whether the camera begins to drift later in the shot.

Match the footage in an editor before the creative grade

Bring the clips into a color-managed timeline and correct each source before applying a shared look. Camera footage recorded in a log format and a generated clip already presented as a normal contrast image need different input treatment.

Use the known camera profile and your project’s color settings. Confirm the generated file’s interpretation rather than assigning the camera’s log transform to it simply because the two shots belong in the same scene. Likewise, a shared creative LUT does not replace basic balancing.

A useful finishing order is:

1

1. Set source and timeline color interpretation. Decide how camera log, SDR or HDR material should enter the project.

2

2. Balance exposure and black levels. Preserve the important highlight and shadow detail.

3

3. Adjust white balance and tint. Compare familiar materials and skin where present.

4

4. Match contrast and saturation. Keep the palette and highlight treatment compatible.

5

5. Apply the intended creative look. Make small shot-specific trims afterward.

6

6. Check texture and motion in playback. Finish the image as part of the moving sequence.

A grade can bring a slightly warmer shot closer to its neighbor. It cannot convincingly move a hard key light from one side of a face to the other. Repair a major lighting-direction problem in the generated composition rather than trying to hide it with color alone.

DaVinci Resolve: compare the shots while grading

DaVinci Resolve is useful for comparing an original frame with generated B-roll and making a controlled color match.

Save a useful reference still in the Color page’s Gallery and use image wipe or split-screen views to compare it with the insert. Keep the original production shot as the visual target while balancing the generated clip.

Scopes help you inspect different parts of the match: waveform for tonal levels, RGB parade for channel imbalance and vectorscope for hue and saturation. Use them alongside the image. Two different compositions should not be forced into identical overall distributions when one naturally contains more dark background or a bright object.

Revisit the comparison after the creative grade. A correction that matches neutral material may still leave the clay, apron or skin looking noticeably different across the cut.

Premiere Pro: use a reference frame in Comparison View

Premiere Pro is useful when you want to compare and match the shots inside the edit you are already cutting.

Open Comparison View from the Program Monitor or Lumetri Color panel. Choose a representative original frame, position the playhead on the target clip and use a side-by-side or split display to compare them. Apply Match provides an automatic starting adjustment that you can refine.

Choose the frame carefully: the match is based on the current frame, so a moment dominated by a bright lamp may poorly represent the rest of the shot. Face Detection can prioritize facial regions when faces matter to the comparison.

Keep a useful previous correction before trying a new reference. Applying another match can replace earlier matching changes, and some existing Lumetri settings may be reset. Finish by watching the whole insert in context.

Final Cut Pro: apply and refine Match Color

Final Cut Pro is useful for a quick color-match starting point in a Mac editing workflow.

Select the timeline clip, position the playhead on the frame to adjust and choose Modify > Match Color. Skim a suitable original clip to find the reference look, click to preview it and choose Apply Match when the result is useful.

Refine the correction in the Color inspector. Check the exposure, contrast and color relationships rather than accepting the automatic result solely because the shots now share a similar overall tint.

Toggle the correction off and on to see what it changed. Then judge the moving cut, where an exposure change or overly bright highlight may be more noticeable than in the comparison frame.

Match texture, movement and frame rate as well as color

Check motion cadence, blur, sharpness and texture after the broad color match is close. An insert can share the palette of the interview while still feeling like a different production because it moves or resolves detail differently.

Set the timeline to your delivery requirement and examine how each source plays within it. Conforming a different frame rate can introduce repeated or interpolated frames; changing a file’s frame-rate interpretation also affects timing. Check the actual movement rather than treating a label change as a complete fix.

Match perceived sharpness at the final viewing size. Reduce an excessively crisp generated image carefully, and add texture only when it serves the original look. A clean interview does not automatically need film grain.

If you stabilize a small unwanted drift, inspect the resulting crop and edges. For different ad formats, our guide to resizing product videos for ad placements explains how to preserve the important subject while reframing.

Judge the two cuts around the new B-roll shot

Watch the original shot, the B-roll and the return shot together at normal speed. The result needs to enter the sequence naturally and return attention to the next shot without a distracting change.

Check these relationships:

• Light: the main direction, softness and bright areas feel compatible.

• Color: skin, clay, wood, packaging or another recognizable material stays believable.

• Space: screen direction and background details do not imply an unintended reversal or a different place.

• Pace: the camera and action fit the rhythm of the spoken line and surrounding edit.

• Sound: the intended dialogue, music or room tone continues appropriately, with unwanted generated audio removed or replaced.

If a location or recurring person needs to remain recognizable across several shots, the guide to consistent locations and characters develops that part of the workflow further.

Repair the mismatch at the stage that caused it

Fix composition in the starting image, motion in the video step and color balance in the edit. This keeps each revision focused on the part of the process that can solve it.

Watch for implied movement in the input. Dust behind a vehicle, motion-blurred hands or a mid-action pose can pull against a request for stillness. Preparing a quieter starting image can be more useful than making the prompt longer.

Save the references and decisions for the next shot

Save the source frames, approved starting images, prompts and grading decisions as one small project package. Keep separate versions for different lighting setups.

For each insert, retain the chosen image, the motion request, the accepted clip and a short note explaining why it fits. A name such as “Studio window — bowl rim — slow turn” is more useful later than a folder of unnamed generations.

Keep the original visual references alongside accepted outputs. Reusing only the latest generated result can gradually change a room, costume or prop across a longer sequence. Save or tag platform references when you need them in future sessions, and share them only through the workspace controls appropriate to your project.

Frequently asked questions

The reference and finishing choices below address common problems when an image looks right but the resulting shot does not yet fit the edit.

Can one reference frame match my camera movement?

A frame can guide the look, while movement matching requires motion information or a clear camera instruction. Watch the original clip and describe the movement, or supply it through a supported video-reference control.

A still also cannot reliably tell you the exact camera package or measured lens setting. Use your shooting notes for known technical details and the image for visible relationships.

Why does the generated shot begin with my original wide view?

In image-to-video, the uploaded starting image often defines the opening composition. If the image shows the whole room, requesting a detail may produce a move across the room toward that detail.

Prepare the desired close-up first when the insert should open close. Keep the wider image when a gradual approach is the shot you actually want.

Why does a locked-off shot still drift?

A locked-off instruction can reduce camera movement without guaranteeing a perfectly stationary result. Check for implied motion in the input and describe the modest subject movement you want within the frame.

For a small residual drift, stabilization may help if the crop remains acceptable. Runway’s Animate Frames route also allows the same image as both inputs as an additional way to constrain a shot; inspect the result rather than assuming those inputs guarantee stillness.

Can I extend a generated shot from its final frame?

You can use the last frame as a new starting image when the tool supports that workflow. In Runway, move to the end of the generated clip, then choose Use > Use current frame to load that image for the next generation.

Combine the clips in an editor, remove the duplicated boundary frame where appropriate and check the join for changes in motion, light and subject detail. Short B-roll inserts often need less extension than a longer continuous scene.

Will the same LUT make real and generated footage match?

The same LUT does not automatically match sources with different color and exposure characteristics. Interpret each source correctly, balance the individual shots and then apply the intended common look.

Use the original production as the target. If a clip still has the wrong light direction or a different moving composition, return to generation rather than expecting a color adjustment to solve it.

Build one matching shot before making the whole sequence

Start with Dreamina, approve one useful B-roll composition and test that short shot between your existing clips. Keep the appearance references, motion instructions and finishing decisions that help it fit.

Once that insert works, reuse the same brief for the next shot while changing its subject, framing and action deliberately. You will build a sequence with a consistent visual language and enough variation to support the story.