Producing UTV Ride Videos

For the last 15 years or so, we’ve been going on offroad UTV rides with friends and family. Over the past few years of that, I’ve experimented with recording and editing those rides into action short films for all of us. The end result is usually a video somewhere around 90-120 seconds long per hour of riding (longer is usually easily possible, but I find that to be a good balance of interesting and unique clips vs very similar segments). The version I’ll present here is about half the usual length for brevity and to keep it a bit more anonymous.

While there are plenty of videos showing the finished product and random tutorials of individual pieces, there is surprisingly little available on the complete workflow behind producing one. Most discussions focus on cameras, other hardware, or editing software in isolation (often monetized heavily with recommendations for promotional products they got for free), but the finished video is really the result of combining three separate processes: capturing footage, generating supporting assets, and assembling everything into a final edit.

This post covers the workflow I currently use, some general notes, and a few dead ends that turned out not to be worth the effort. My goal isn’t to produce a Hollywood-quality film. It’s to capture enough of the experience that everyone who was on the ride can relive it later, while still making something enjoyable for people who weren’t there. Everything in this workflow is ultimately in service of that goal.

Excluding experimentation and research, this particular ride’s video required roughly three to four hours of editing work, much of it waiting on the software itself. Being already familiar with the footage, the second condensed cut I made for this post took around an hour. Those figures include importing footage, assembling the timeline, generating and synchronizing music, creating map animations, rendering, and making final adjustments.

Recording the Footage

I have five cameras available: two drones, two GoPros, and a Canon mirrorless SLR. They were obtained over a period of years and most are far from current models, but they still work great for my needs.

Each camera has a different purpose, doing what it’s best suited for.

DJI Neo 2: follow footage or automated flight patterns (e.g., a vertical spiral)
DJI Mini 3: general high aerial, hovering, and stills
GoPro Hero 5: gimbal-stabilized footage using a GoPro Karma gimbal (great in particular for stop footage like trail clearing with a chainsaw)
GoPro Hero 8: cage-mounted footage
Canon EOS M6: longer focal lengths and stills

Ultimately, the actual camera models don’t really matter. I use what I use because they were the best choices available to me at the time and still work great. Many newer options may be somewhat better in some areas (e.g., low light), but what I have is completely sufficient for what I want to do. It’s worth noting, however, that dedicated cameras are preferable — camera phone footage has typically been mostly unusable, though it does allow including content on occasion from other riders when it’s high enough quality.

Further, not every ride uses the same cameras or even all of them. The majority of the footage for the original cut of the video discussed here was recorded with a GoPro Hero 5 mounted on a Karma gimbal, in large part because this was my first attempt at incorporating the DJI Neo 2 and in trying to use a dampener on the GoPro Hero 8.

Make sure to synchronize their clocks. It makes the final stage way, way easier. Additionally, own enough batteries to cover the day. Video is fairly battery-intensive, and each camera may require multiple batteries to last more than a couple hours. I also have multiple USB charging ports on the machine to keep things topped off when possible, and it’s not at all cumbersome to do.

Think Like an Editor

A thing I didn’t recognize initially is that the editing process really begins during the ride itself. It’s very easy to spend the entire day driving with the camera recording and expect to sort it out later, but that usually leads to a video that’s visually repetitive.

Instead, I intentionally look for opportunities to capture footage that effectively makes the editing decisions now rather than in post-production. I’ll stop for a drone shot before or after an interesting obstacle, let other vehicles drive past while I’m stationary, or record a section of trail from outside the vehicle before continuing on. Those short interruptions only take a few minutes during the ride, but they provide natural transitions later and make the finished video feel much more varied.

Many of the other riders have also developed a sense for which trails lead to better footage and will set up for it.

Stabilization

Raw footage is rarely usable as-is. No one enjoys watching a shaky, disorienting video.

Drones stabilize themselves, but everything else needs a little help. I spent a fair amount of time trying to mechanically eliminate vibration. Shock mounts, dampeners, and similar products all seem like obvious solutions, but in practice many simply replace high-frequency vibration with lower-frequency oscillation that is even more noticeable because they can’t absorb enough of the impact.

Unless you’re prepared to go down the road of thousands of dollars of rigging, my experience has been that for fixed cameras like a cage-mounted GoPro, a rigid mount combined with good electronic stabilization generally produces better results than most any practical option.

Another surprisingly effective option is to use your passenger. A person holding a gimbal stick, or even wearing a helmet- or chest-mounted camera, uses their body’s natural reflexes to dampen the big drops. Most trail driving also feels slower on video than it does from the driver’s seat, so a modest 20–30% speed increase often restores some of the sense of motion without looking unnatural.

DJI Neo 2

This was a last-minute addition to the line up. I’d wanted to try incorporating a follow drone for some time, and I had heard from a friend that there was actually a drone with obstacle avoidance and tracking at a price point I could tolerate. It arrived two days before the scheduled ride, but it turned out there were a couple important pieces not mentioned that I needed for my purposes.

The Neo 2 straight out of the box can only track human subjects, which is not exactly viable when the target is going to be seated in a UTV. Research led to the need for a controller, and I was able to overnight a DJI RC-N3. Upon pairing this, or at least attempting to, I then discovered the need for a transceiver (now installed) on the drone. I could not get this in time, so the follow drone was going to be of very limited use this ride.

The appeal wasn’t just having another camera in the air. Every other camera I use is attached to the vehicle in some way. A follow drone can visually show the vehicle itself moving through the landscape rather than only showing the driver’s perspective.

GoPro Hero 8

I also experimented with a GoPro Hero 8 mounted directly to the upper-left corner of the roll cage above the windshield. The original plan was to have a camera controllable by a remote on the steering wheel and to stabilize the footage through mechanical isolation using a dampener intended for microphones (similar weight to the GoPro).

The result was the opposite of what I expected.

Instead of eliminating movement, the dampener introduced additional oscillation after larger bumps and suspension hits. The footage developed a floating, spring-like motion that was often more distracting than the original vibration, and no change in orientation seemed to work any better.

During the original cut, I discarded most of the Hero 8 footage as unusable. It wasn’t until later that I discovered Gyroflow and revisited some of the clips. The results were impressive — footage that looked far too shaky in its raw form became surprisingly smooth after processing. It achieves the stabilization by reading the embedded gyroscope data within the clip and then rotating/cropping to maximize stability and the field of view. You ultimately lose some content around the edges as a result, but at least the clip is usable. The cut here is after rendering much of the footage usable with Gyroflow, footage I was unable to use in the original.

That one discovery completely changed how I look at action camera stabilization. Instead of trying to solve every vibration problem mechanically, I’m now much more willing to let the camera record imperfect footage and recover it in post-processing. It’s far, far cheaper in the end.

Creating Supporting Assets

The ride footage itself only tells part of the story. Maps, music, titles, and transitions help provide context and keep the video engaging. All of these have become much, much easier to create in the AI era.

Music

For years I relied on royalty-free music libraries, including the collection that GoPro previously made available. Over time I simply ran out of tracks that fit the style I wanted, but Suno launched and filled the void.

Rather than searching through hundreds of existing tracks, I can generate music that matches the pacing and tone I want for the video. For longer projects, I generate multiple related tracks as I found it much easier to generate songs for the introduction, action sequences, and outro separately than to try convincing the model to produce a single 13-minute piece.

It can often take several generations before I get something that fits right, but that’s still considerably faster than spending an hour digging through royalty-free music libraries hoping to find a track.

The first song in the video discussed here is this prompt, generated by using an AI agent to analyze songs I had previously used:

High-energy off-road instrumental, turbo-charged ATV pulse with punchy four-on-the-floor kick accents, chopped engine rev samples, syncopated toms, and rising snare rolls, Verse rides sparse with grit and tire-spit percussion; pre-chorus strips to rumbling bass and filtered revs; chorus hits with stacked brass-like synth stabs and a climbing lead riff, Finale opens wide with layered percussion and chain-rattle textures, Aggressive close-mic mix, bright and punchy with dusty trail-edge grit, bright, light

The overall audio structure of the video consists of:

An introductory track
One or more action tracks
An outro track

The intro and outro tracks are deliberately chosen to ramp up into the fast action and then slow down from it. The action tracks in the middle carry the video along and share a similar style and energy level.

As each track is added to Final Cut Pro, I’ll listen once and (M)ark the major transition points in the audio to help with aligning the transition points in the clips. Anything that changes overall tempo, strong beats, or other notable features gets marked for potential use.

Route Maps

I record the route the day-of using Gaia GPS and later export a topographic map image of the ride area from CalTopo. Once home, the route is exported from Gaia as a KML file (KML is a widely supported format for geographic paths, making it easy to convert into other formats), which I then feed through an AI-created tool to convert it into a Apple Motion bezier curve.

A web-based version of that converter may be found here: KML/KMZ to Apple Motion Bezier Curve. This tool produces a file that looks like this:

The topographic map from CalTopo is placed on a layer below the route curve and then animation is added to both. The curve draws itself on top of the map (this is the Write On Behavior), and the overall framing applies zoom, pan, rotation, and shear effects to create the impression of a moving camera over the terrain (for this I just added a keyframe at the start and the end, then resized, panned, rotated, and sheared to where I wanted it to be). The result is not truly three-dimensional, but it provides enough depth and movement to keep the sequence visually interesting.

A static map shown for ten seconds quickly becomes something viewers ignore. Small amounts of motion from zooming, rotating, and slowly revealing the route keep it visually interesting while also helping bridge longer sections of driving footage.

The final result:

The Blender Detour

At one point I experimented with a true 3D terrain flyover. That led to a two-day excursion into Blender, GIS tooling, digital elevation models, coordinate systems, terrain meshes, route imports, and camera animation.

Technically, it worked.

I was able to generate terrain from USGS elevation data, apply map textures, import route information, and animate cameras through the scene.

The problem was not feasibility but rather time. After spending two days building the workflow, the final result provided only slightly more visual information than the faux 3D map but was much more cumbersome to generate for a ride even with the workflow established.

USGS elevation data was tricky to align well with the topographic map so peaks actually displayed as peaks, the route itself needed transforming to align to the map and often had enough inaccuracy to display inconsistently, and getting the render camera to move smoothly along the trail was surprisingly difficult.

For now I have returned to the simpler Motion-based workflow. It is dramatically faster and produces results that are good enough for my tastes.

Assembly and Final Render

The final stage takes place in Final Cut Pro, though most any video timeline editor would work adequately. DaVinci Resolve is another popular choice. None are without some learning curve.

Once all footage has been imported, the first step is simply identifying what is worth keeping. A full day of riding generates far more footage than will ever appear in the finished video. The goal is not to document every mile of the ride but rather to produce something people will actually watch.

Long repetitive stretches are removed. Similar shots are reduced. Clips that help tell the story of the ride are prioritized, whether that’s interesting terrain, an obstacle, a scenic overlook, a water crossing, or even a short interaction between riders. The goal isn’t to preserve every mile — it is to preserve the memorable moments.

Building Around the Music

The music is the framework for the entire edit. Those markers I mentioned putting in earlier? Major transitions in the footage are aligned with those musical breaks, tempo changes, or energy shifts. If a track ramps up, that often becomes the point where the video transitions into a faster or more visually interesting section.

Likewise, slower sections provide opportunities for aerial footage, scenery shots, or map transitions. It’s not just background audio but rather the pacing guide for the project.

I also preserve most of the original camera audio, typically lowering it around 10 dB so it sits underneath the music. Completely muting the ride makes everything feel oddly disconnected. Keeping the engine, tires, and occasional conversations quietly in the background helps preserve the feeling of actually being there.

Interstitials

Map animations are used between major sections of the ride. They serve several purposes simultaneously:

Showing progress along the route
Providing geographic context
Breaking up long stretches of action footage
Creating natural transition points

I export the map animation at a high enough resolution that I can use the same file in multiple places, just cropped in to highlight the area of interest. Combined with aerial footage, transitions, and title cards, they help maintain variety throughout the video.

Keeping the Viewer Engaged

One challenge with off-road videos is that the ride is often more enjoyable to participate in than it is to watch.

The key is variety.

Different camera angles, aerial footage, route animations, scenery shots, and music transitions all help maintain viewer interest. Even if a section of trail was enjoyable in person, that does not necessarily mean it deserves several uninterrupted minutes in the final edit.

It also usually helps to speed up the footage by 20-30%. Most trail driving feels slower on video than it does from the driver’s seat, and a slight speed increase restores some of the sense of motion without looking unnatural.

The finished video should feel shorter than it actually is.

Review and Re-edit

Your first cut is going to be rough — it shouldn’t be your first and only pass. Watch it through, stopping to make adjustments as needed whether it’s a missed scene transition or cutting out a bad clip segment.

Repeat and refine until it looks good. I guarantee you’ll see something every iteration through that needs tweaking, but you will reach the point where it’s good enough, and that’s when it’s time to export. Know when to stop.

Export it, but consider waiting to do a final watch. Waiting even a few hours makes pacing problems and weak clips much easier to notice. Play it on a big TV.

Unless there’s something major missed, it’s time to upload and share. I use YouTube for that because it’s very untechnical for viewers, but there are plenty of options out there.

Conclusions

The recording equipment is only a small part of producing an off-road video. The final result is a combination of footage capture, asset creation, music selection, map generation, and editing decisions. Editing is what transforms it from an uncut video ignored on social media to an engaging short people want to watch.

After experimenting with a variety of approaches, I have gradually converged on this workflow, which balances effort and results. Some ideas, such as Gyroflow processing, turned out to be surprisingly valuable. Others, such as the attempt to build fully three-dimensional terrain flyovers, proved difficult to justify given the additional complexity involved.

The actual ride spanned most of a day while the finished video took a few additional hours. Those few hours of editing preserve memories that would otherwise fade surprisingly quickly, and it gives everyone who was on the ride something they can come back to years later.

With that, I’ll end with a few of the most important lessons:

Record more variety than you think you’ll need.
Rigid camera mounts plus modern software stabilization produces surprisingly good footage.
Generate music in sections instead of trying to create one long track.
Animated maps are worth the effort; full 3D terrain generally isn’t.
Let the music determine the pacing.
Know when the video is “done.”