# Marcus Tran, Senior Product Engineer at Helix Health (38 employees) — read of FFmpeg WebCLI, June 5 2026

> 8 years building health-adjacent SaaS, currently the guy who gets handed "just add video" tickets that turn into 3-week infrastructure spirals.

## How I got here

I was googling "ffmpeg wasm react" for the third time this month because we keep hitting the same wall: patients send us intake videos and we need to strip metadata and transcode before storage, and I do not want to spin up another Lambda pipeline to handle it. Search result was a Stack Overflow thread from 2023 that had a comment linking here. No ad, no LinkedIn. Just someone in a thread who said "this exists now."

## What I clicked first

The hero subhead: "Edit, transcode, and process video locally with zero cloud dependency." That's the whole thing. I don't care about editing. I care about processing. The fact they said "transcode" in the first line told me they're not just talking about trim-and-export.

## Where I paused

"100% compatible with FFmpeg filters and codecs. Use the same commands you know from the CLI."

I stopped here because I know this is not true and they probably know it too. WASM FFmpeg has real constraints: no hardware acceleration, some codec licensing issues, container format gaps. Saying "100% compatible" in a headline on a product page is either a lie or they're using "compatible" to mean something narrow. I'd want to know exactly which filters are missing and what the perf delta is on a 1080p file.

## What I distrusted

"Real-time preview. See your edits instantly." I've used ffmpeg.wasm before. On my M2 MacBook Pro, a 10-second 1080p clip with a simple filter takes noticeable seconds to preview. "Instantly" is doing a lot of work in that sentence. I don't know what their preview system actually is but the word "instantly" makes me think they haven't tested on a Windows laptop from 2021, which is what half our users are on.

Also: "Built for creators and developers" with six feature bullets, but zero screenshots of the actual UI in the page text. The "video editing concept image" alt text tells me what that image actually is: a stock photo. That's fine but it means I have no read on whether the UI is a textarea with a run button or something genuinely usable.

"The middle ground doesn't exist" is lazy. Remotion exists. Shotstack has a local mode. Creatomate has client-side options. I'm not saying those solve my problem but the framing that nothing else is in this space is the kind of thing founders say when they haven't done competitive research past the top five.

## What would convince me

One benchmark: "We ran these 10 common ffmpeg commands on a 2-minute 1080p file in Chrome on a mid-range Windows machine. Here are the times." Not vibes, not "near-native," just a table. That would tell me more than the entire features section.

A real answer on codec support. Not "100% compatible" but a page that says "here's what works, here's what doesn't, here's why." If they have that and I just didn't find it in the stripped text, that's a navigation problem.

One sentence on what "hosted version" means. Is that a UI they host? An API? A self-hostable thing? I genuinely could not tell.

## What I'd ask in an email reply

1. What's your actual test setup for the "real-time preview" claim? I need to know if this holds on a 2019 Windows machine before I show it to anyone.
2. Is there a headless or SDK mode, or is this browser-UI only? My use case is server-side processing in a sandboxed iframe, not a user-facing editor.
3. "Commercial support" -- what does that mean in practice, is that a paid tier with SLAs or just "email us and we'll help"?

## Verdict: on-the-fence

The core idea is technically sound and the use case they're describing is real. But the page makes three claims I don't believe ("instantly," "100%," "middle ground doesn't exist") and tells me nothing about the actual UI or what commercial support looks like. I'd read the docs before I'd reply to an email.

---
*Memo by skeptic persona, generated 2026-06-05. Studio breaks own self-grading loop.*