Field Test · Part One
AI That Finds the Music—Without Making It: Jeff Perkins on Soundstripe
Before ECG puts Soundstripe into a month of real production work, CEO Jeff Perkins walks us through its AI-assisted discovery tools, human-made catalog, editor workflow, and licensing philosophy.
Music search can be the part of an edit that looks simple until it quietly eats an afternoon.
The picture is working, the pacing is close, and the team knows what the piece should feel like. But translating that feeling into a useful search term is another job entirely. That is the problem Soundstripe is trying to shrink.
Soundstripe CEO Jeff Perkins joined Jason Sirotin, Jason Marraccini, and Trey Gregory for a working session on the company’s two AI-assisted discovery experiences: Supe, its conversational search assistant, and Soundstripe LIVE, a new companion that recommends music while an editor works.
This is a first look, not a final verdict.
ECG received access to evaluate the platform. We are documenting what Jeff showed us, what stood out in the first session, and what we will test before he returns for a follow-up podcast.
The Core Idea
Use AI to find the music. Keep the music human.
The most useful idea in the conversation was not a feature. It was a boundary. Soundstripe is using AI to interpret scenes, translate creative language, and reduce search friction. It is not using AI to generate the music in its catalog.
“AI is great for workflow efficiency improvements. But when it comes to music, we like human music. We think it’s better, more authentic, and resonates with people more.”
Soundstripe’s own policy says its catalog music is composed, performed, and produced by human artists. For Perkins, AI belongs in the connective tissue around that work: discovery, matching, refinement, and workflow. Soundstripe’s human-made music policy
Two Ways In
Supe searches with you. LIVE watches the scene with you.
Both products are trying to solve the same creative bottleneck, but they enter the workflow at different points.
Inside the web catalog
Supe
- Describe the mood, scene, instrument, or creative brief in normal language.
- Upload an image or video so the search begins with the actual visual material.
- Paste a Spotify, Apple Music, YouTube, or other supported reference-track link.
- Refine results conversationally: slower, more synth, no vocals, or more like this.
- Open similar-song results once one option gets close.
Alongside the edit
Soundstripe LIVE
- Select the application or window containing the picture you want LIVE to analyze.
- Let the Mac app interpret changing scenes and surface tracks that fit the moment.
- Preview music without leaving Premiere Pro, Final Cut Pro, or DaVinci Resolve.
- Use feedback controls to improve recommendations during the session.
- Download and drag a selected track directly into the editing timeline.
At the time of ECG’s July 2026 session, LIVE was available for Mac. Trey Gregory, who edits primarily on a custom PC workstation, would use Supe in the browser while ECG evaluates the difference between the two workflows.
The First Demo
We gave Supe an actual ECG promo.
Rather than searching from a hypothetical brief, ECG uploaded a finished TextNow promo. Supe analyzed the file, described what it saw, and began building a set of music recommendations.
Perkins showed how a broad first result can become useful through follow-up direction. Ask for something more playful. Request electric guitar or synth. Remove vocals. When one track gets close, use the similar-song function to search around that choice instead of starting over.
Picture-in-picture kept the promo visible while the team auditioned tracks. That sounds like a small interface choice, but it addresses the real problem: music cannot be judged separately from the timing, edit rhythm, performance, and reveal it is supposed to support.
“The catalog is good. It’s unique. It doesn’t sound like the typical generic stuff that’s out there.”
The first impression was promising, especially the playback fidelity and the ability to move between visual analysis, conversational direction, and reference-track matching. But a guided demo is the easy part. The real test begins when the deadline is real and the first recommendation is wrong.
Quality After Search
A match still has to survive the mix.
Search speed is not much of a victory if the chosen file creates a new post-production problem. Perkins told ECG that Soundstripe’s music is delivered at 48 kHz and that stems are prepared by audio engineers rather than generated by an automated stem-separation tool.
That matters when dialogue needs room, a client asks for a six-second cutdown, a percussion layer becomes too busy, or the music has to work across broadcast, social, and internal versions. ECG will test the stems as production assets—not just listen to full mixes in a browser.
Rights After Download
A good song is not enough. The coverage has to fit.
The least glamorous music decision can become the most expensive one: who is licensed, for which client, in what media, across which territory, and for how long? A production company needs those answers before the campaign travels farther than expected.
Soundstripe says plan and license levels determine whether a project is covered for personal, commercial, client, business, or broadcast use. It also says projects properly created during an active subscription remain licensed. The right choice still depends on the actual project, so ECG will evaluate the paperwork alongside the search experience.
The AI Question
A tool can be useful without becoming the author.
The conversation did not turn into a simple human-good, AI-bad argument. Perkins acknowledged that generative tools can help someone sketch an idea, express something they could not otherwise produce, or build a demo before involving musicians. ECG described a similar hybrid approach in visual effects: use AI to extend or enhance a shot when it is the right tool, then integrate it into a larger human-led production pipeline.
The distinction becomes sharper when the work is commercial. A playful personal experiment and a national client campaign do not carry the same rights exposure, audience expectations, or approval process. For Soundstripe, that makes human-made catalog music plus AI-assisted discovery a deliberate product position rather than a nostalgic one.
The question is not whether AI touched the workflow. The question is what it did, what it replaced, and whether the people publishing the work can explain the rights and creative choices behind the result.
The ECG Test Plan
What we will measure before Part Two.
A month of real use should make the next conversation more useful than another feature tour. We will bring Jeff evidence: where the tools saved time, where they got in the way, and what happened after a track entered the timeline.
- 1Time from opening the tool to a genuinely usable shortlist
- 2Recommendation quality from a creative brief, uploaded video, and reference track
- 3How well conversational refinements change the results
- 4Workflow friction for Mac editors using LIVE and PC editors using Supe
- 5The usefulness and quality of stems, alternate versions, and cutdowns in a real mix
- 6How easily producers, editors, and clients can review the same options
- 7License selection, proof-of-license documentation, and client handoff
- 8Whether the final track still feels distinctive after revisions and repeated review
Coming Next
The follow-up podcast starts after the honeymoon period.
ECG will invite Jeff back after the team has used Soundstripe in real edits. We will compare Supe with LIVE, look at the tracks that made it into timelines, discuss the recommendations that missed, and ask what Soundstripe learned from the way professional editors actually used the tools.