Why pay $180 a year for a tool that does exactly one thing, when the subscription you already have can do it?

That's the question I'm testing right now with a client, a company that educate adults across Latin America, entering the German-speaking market. They have hours of Spanish-language training video, built over years and they need short clips for social to reach a new audience. The catch: they didn't want to lose the Spanish audio. That's part of their DNA, not a detail.

My first instinct wasn't to point them at Opus Clip or any other clipping tool, but I asked first what they are already paying for. They have a Claude subscription. So instead of adding another one, I'm building the tool directly with Claude Code, on top of what they already have.

Here are the results:

- One 1-hour video goes in, 8 highlight clips come out.
- German subtitles get added, Spanish audio stays exactly as it is.
- It's built and installed directly on their setup, they run it themselves, no dependency on me every time they need clips.
- It runs on the subscription they already pay for, not a new one.
- At 5 source videos a month, that's a reel a day with room to spare, no extra $174 to $180 a year for a tool that also comes bundled with features they'd never touch.

Worth saying to anyone doing this kind of localization, not just this client: most short-form video gets watched on mute. The caption is doing the work, not the audio. So keeping the original voice doesn't cost you completion rate and you keep something a dub can't give you, which is the actual person and their real delivery.

This is the part that adds up quietly: one video tool here, one caption tool there, one scheduler somewhere else. Each one small, each one billed monthly, none of them checked against whether the client needed the whole plan or just one feature inside it. Multiply that across a marketing team and you're paying full price, indefinitely, for capacity you use a part of. 🥲

Before I recommend a tool, I ask what a client already has, whether it covers the job and whether they can run it without me. With Anthropic Claude Code, all three usually check out. Building the specific thing tends to beat renting the general one.

Still mid-build on this, testing it against real footage before it touches the client's content. If you or your team is "collecting" subscriptions for one-off needs, happy to discuss.

PS: turns out keeping the ORIGINAL audio while subtitling in a different language is trickier than it sounds, German reads slower than Spanish sounds, so the timing needs its own pass. 😄