What is a good Opus Clip and Riverside alternative, and why would you build your own?
by Alessandro · updated 2 November 2026
I built Underway with Rich after Riverside and Opus Clip stopped being good enough for Risk Takers. Riverside records separate tracks well but just stitches them into a plain video. Opus Clip's clips had no real start or end. Underway is the tool I built instead: it understands what is being said and adds data cards, charts and definitions on screen, then cuts that into clips.
Key numbers
I don't care if the clip's 15 seconds or 15 minutes, to be honest I actually don't care, what I care about is that the clip has a start and an end.
In short
- Riverside records clean separate tracks but does nothing to make the finished video understand the conversation.
- Opus Clip's automatic clips jumped around with no proper start, middle or end, so I stopped using it.
- Underway is the engine I built with Rich: it reads the conversation and brings up the right chart or data card for what is actually being said.
- I don't care if a clip runs 15 seconds or 15 minutes, only that it has a start, a middle and an end that makes the point.
What Risk Takers Is
Risk Takers is a podcast I'm building where I get people who are genuinely good in their niche on for real conversations. Along the way I built an engine that takes that raw conversation and turns it into something properly produced: pop-ups on screen, data cards, definitions, the works, front to back.
I'd tell anyone curious about the finished product to go and look at the comments on the channel rather than take my word for it. What I talk about on the show is relatively complex. A lot of the terminology is hard to get your head around the first time you hear it, even though people study this stuff at university for years before they start talking about it casually.

Explaining Complex Terms Without Killing the Conversation
I felt this myself, coming at markets from a slightly different angle. Someone would say the yield curve, or the base rate, or a basis point, and I'd have to hear the term a few times before I had even a conceptual idea of what it meant, then go away and research it properly.
What I want my show to do is help explain what's going on inside the content itself, without ever interrupting the conversation to ask a guest to define something. That kills the conversation and it's irrelevant anyway. I don't want to get an expert on and then make them explain a basic concept. Basic concepts can be Googled, or you can just ask Chat. The problem is you don't know what question to go and ask until you've heard an interesting conversation first.
That's what Risk Takers started as, and it's what's let me produce episodes where people can actually understand what's being said, because we use visual representation and description alongside the talking to help them follow it.
Why Opus Clip and Riverside Weren't Good Enough
I used to use Riverside for recording the show, and I used Opus for clips, until I started building Underway instead. Riverside captures audio and video cleanly, each person on their own camera, stitched together properly, and honestly I think a remote conversation often looks better on YouTube than an in-person one, because in person it's hard to get a camera straight on someone's face. But Riverside just takes those two tracks and turns them into a video. It makes a nice video. Nothing about it makes the video exceptional.
Opus was worse. The clips it produced were horrendous: no proper start, no middle, no end, they didn't make the point, they just jumped around. The engine making them was so generic it did nothing to make any individual clip good. So I decided to try building a clipping tool of my own instead that clips my own show and does it properly.
Why I Built My Own Tool
I spent a long time getting irritated, and spending a lot of tokens, making sure Claude understood what actually makes a start to a clip. I don't care about a flashy hook, it just needs to open on a normal bit of conversation. For the middle, does the person round out their point. I don't care whether the final clip is 15 seconds or 15 minutes. What I care about is that it has a start and an end and makes the point, because clips that don't make me annoyed.
Beyond the cut, the engine has to understand what's actually being discussed. If a guest is talking about mortgage rates, the right chart isn't the Fed funds rate, it's the US 10 year, because that's what mortgage rates are actually anchored to. That takes real understanding of the subject, not a generic template, before the chart and the explanation can bring the guest's point to life on screen.

How Underway Is Meant to Work, Step by Step
Underway is software, not a service I run for people. The plan, as I've laid it out, is that a creator signs up and gets access to the engine covering the entire production front to back: Underway handling outreach by email to find good guests in a creator's niche, booking them into the calendar, recording each person's own track in HD or 4K on their own mic, then combining those tracks into a finished video.
From there the engine decides the shape of each moment, sometimes the host full screen, sometimes the guest, sometimes split screen, sometimes bringing up an article, a data card or a chart relevant to what's being said, with descriptions and headlines at the bottom. It then cuts that into horizontal and vertical short clips and can connect to YouTube and X to post with captions.
What a creator still has to do themselves is the research and the conversation. We even produce guest research, but I wouldn't recommend relying on that alone. Knowing the guest, or at least being able to let them talk and make their point without interrupting, is still what makes a podcast grow. That part doesn't get automated.

What It Actually Takes to Build This Yourself
Building your own version of this is not a weekend project. The hard part wasn't recording tracks separately, Riverside already does that well. The hard part was teaching the engine what a clip actually needs: a real start, a real end, and the judgement to pull in the right chart for the right claim, which for me meant a lot of back and forth getting the model to understand concepts like why mortgage rates track the US 10 year and not the Fed funds rate.
If you're not prepared to spend that time getting the underlying understanding right, you'll end up with something closer to Opus: technically functioning, visually generic, and not actually making anyone's point better.
Who This Suits and Who It Doesn't
This suits a show where the subject matter is genuinely complex and the audience benefits from seeing a definition, a chart or a data point appear exactly when it's being discussed, rather than being told to go and Google it mid-conversation. It also suits someone recording remotely, since I think the production genuinely looks cleaner than an in-person interview with a side camera.
It doesn't suit someone who just wants a quick clipping tool for simple content, where a generic highlight reel is good enough. It's software I'm building with Rich, and I'm now posting about that build most days on Underway so people can see why and how.
Questions people ask
Is Underway the same as Risk Takers?
No. Risk Takers is the podcast itself, the interviews with guests in their niche. Underway is the separate production engine, built with Rich, that turns those conversations into finished, produced episodes and clips.
Do I need to be technical to build something like this myself?
You need patience with a model like Claude more than traditional coding skill. The real work described here was back and forth getting the engine to understand what makes a clip's start and end work, and which chart actually belongs with which claim, not writing code from scratch.
Why not just keep using Riverside for recording?
Riverside still does a solid job capturing separate, clean tracks from each person. The gap is what happens after that: it turns those tracks into a plain video and nothing more, with no understanding of what's actually being said.
What was wrong with Opus Clip specifically?
The automatic clips had no real structure, no clear start, middle or end, and jumped around without making the guest's point. The underlying engine was generic enough that it didn't do anything to lift any individual clip above average.
Does the clip length matter?
Not by design. The stated rule is that a clip can run 15 seconds or 15 minutes, the length isn't the measure of whether it works. What matters is whether it opens on a normal bit of conversation and closes once the guest has made their point.