DEV Community: Aaron Colbry

I built a scheduler so AI assistants can post for you (safely): SocialRoster + an OAuth 2.1 MCP server

Aaron Colbry — Sun, 18 Oct 2026 13:19:02 +0000

Showdev time. I'm Aaron, and I just shipped SocialRoster (https://socialroster.dev) — a calendar scheduler for TikTok, Instagram, YouTube and Facebook, built for people who batch-produce short-form video. Import up to 50 clips, caption them, space them 1–24 hours apart, done.

The part I think this crowd will appreciate: instead of bolting on an "AI feature," I exposed the whole product as an MCP server. It uses OAuth 2.1 with PKCE and dynamic client registration — no API keys pasted into chat, no shared passwords. Two scopes: read (accounts, queue, plan) and write (drafts, scheduling, publishing). 15 tools total, including a bulk schedule_batch.

Hardest lesson so far: every platform's "publish" means something different. TikTok requires posts to be reviewed and published in-app, so my connector flatly excludes TikTok from instant publish and says so in the docs. Modeling each network's constraints as pre-flight checks beat four hand-rolled integrations.

It's $19.99/mo for 200 posts, $39.99 uncapped, 3-day free trial. Questions about the MCP auth flow very welcome — DCR in production has some sharp edges I'd happily compare notes on.

Manual clip assembly after Opus picks the wrong 30 seconds

Aaron Colbry — Wed, 30 Sep 2026 19:36:46 +0000

Opus is useful for finding candidate moments, but it will sometimes pick the wrong 30 seconds—or pick one good moment when the story needs three.

That is where browser-based assembly and CapCut desktop solve different problems:

CapCut desktop wins when I need a full timeline, layers, effects, keyframes, and detailed audio work.
A browser workflow wins when I already have the clips and want to trim, reorder, crop to 9:16, preview, and export one clean vertical compilation without installing a desktop editor.

My practical fallback after an Opus miss is simple:

Keep the source clips, not just the suggested highlight.
Mark the usable in/out points myself.
Put the clips in a deliberate order: hook, context, payoff.
Preview the vertical crop before exporting.

This is clip assembly, not AI video generation. The editor should make the choices visible and reversible.

I’m building ReelWorkshop, so this is a maker recommendation: the free demo lets you try the browser assembly flow. It’s free to edit and preview; Starter is $19.99/month if you need the paid tier. No auto-highlight promise—just a faster path from footage you already have to a finished reel.

Why we don’t claim auto-highlights

Aaron Colbry — Wed, 30 Sep 2026 18:02:04 +0000

Auto-highlights sound like a shortcut. Sometimes they are. But selecting a clip is not the same as understanding why a moment matters.

A useful short often depends on context: the sentence before the punchline, the product detail that proves the claim, the reaction shot that makes the story land, or the pause that should be removed. A model can suggest candidates, but a creator still needs to decide what the cut is trying to say.

That is why ReelWorkshop is deliberately an assembly tool for existing footage. You bring the clips, choose the moments, arrange the story, and preview the result in a browser. It is not a text-to-video generator, and it does not pretend to discover the “right” highlight for every recording.

Our practical workflow is:

Mark candidate moments while reviewing the source.
Keep the strongest claim or visual proof.
Assemble a short 9:16 sequence.
Check the crop, captions, audio, and platform UI safe zones.
Export and watch the file on a phone before publishing.

This also makes the product easier to understand. Edit and preview are free. Starter is $19.99/month for creators who need more room in the workflow. If you want to see the browser assembly experience, there is a demo at https://reelworkshop.com/demo.

The broader lesson: automation should remove repetitive work, not hide the editorial decision. A transparent manual step is often better than a magical button that gives you a confident but irrelevant clip.

CapCut vs Opus Clip: choose by job, not by hype

Aaron Colbry — Wed, 30 Sep 2026 17:26:55 +0000

If you already have a folder of clips, the CapCut-versus-Opus Clip question is often framed backwards.

CapCut and Opus Clip are useful, but they solve different problems:

CapCut is a broad editor. It is a good fit when you want effects, captions, templates, audio work, and hands-on control.
Opus Clip is aimed at finding short moments inside longer footage. It is a good fit when the hard part is discovering candidate clips.
A third job is easy to miss: you already picked the moments and just need to assemble them into a clean vertical compilation.

That last job is what I kept rebuilding by hand, so I made a small browser tool for it.

Start with the footage, not the logo

Before choosing an editor, ask one question: do I need to find the moments, or do I already know which moments I want?

If you have a one-hour interview and need candidate highlights, an auto-clipping tool may save time. Review every suggested cut, though. Automated selection can misunderstand a pause, a sentence boundary, or the context that makes a moment worth keeping.

If you have 8–30 short clips from a shoot, trip, game, or event and already know which ones belong, a full editor can be more machinery than you need. In that case the useful sequence is usually:

Import the selected clips.
Put the strongest hook first.
Trim dead air at the head and tail.
Arrange the clips into a beginning, middle, and payoff.
Preview the whole sequence at phone size.
Export one 9:16 H.264 MP4.

No tool can decide whether the story works for your audience. A thumbnail and a quality score are not a substitute for watching the sequence all the way through.

The small browser workflow I wanted

I built ReelWorkshop around that assemble-first workflow. It imports clips you already have, lets you arrange and trim them, previews the full vertical sequence, and exports a 9:16 video for TikTok, Reels, or Shorts.

It is intentionally not an Opus Clip replacement for automatic highlight discovery. It does not claim to generate a finished video from a prompt, and it does not pretend that a caption helper is an AI editor. Optional AI help is limited to writing titles, hooks, captions, or hashtags; the footage and the edit remain yours.

The point is a fast, bounded step between “I picked these clips” and “I have one vertical file ready to publish.”

An honest freemium boundary

The demo lets you edit and preview free. Finished exports are on the Starter plan at $19.99/month. That boundary is deliberate: you can test the workflow with your own clips before paying, but “free” does not mean unlimited exports.

That is also why I would not describe ReelWorkshop as a universal replacement for CapCut or Opus Clip:

Choose CapCut when effects, templates, audio mixing, or detailed manual editing are the job.
Choose Opus Clip when automatic moment discovery from long-form footage is the job.
Try ReelWorkshop when the clips are already selected and the bottleneck is assembling, previewing, and exporting a vertical compilation in the browser.

You can start editing free and see whether that narrower workflow fits. If it does not, the comparison still gives you a useful decision rule: pay for the part of the process that is actually slow for you.

A quick pre-export checklist

Before exporting, I check:

Is the first second clear without an explanation?
Does the subject stay inside a safe vertical crop?
Did I remove pauses without cutting a sentence in half?
Does the audio jump noticeably between clips?
Did I watch the full sequence once at normal speed?
Is the file named clearly enough that I will not schedule the wrong version?

The last check matters more than it sounds. A faster editor does not help if the wrong export gets uploaded.

Disclosure: I built ReelWorkshop. CapCut and Opus Clip remain legitimate tools, and this is a workflow distinction rather than a claim that one product wins every edit. ReelWorkshop is for assembling clips you already selected; it is not an auto-highlight cutter.

A 9:16 crop workflow that keeps browser reel assembly honest

Aaron Colbry — Wed, 30 Sep 2026 17:20:21 +0000

It does not generate a video from a prompt, auto-pick highlights, add captions or music, or promise reach.

Editing and preview are free. Finished MP4 export is on the Starter plan at $19.99/month. That boundary is intentional: use the free tools to validate the framing and sequence before deciding whether the export workflow fits.

The general lesson: make the crop decision visible, keep the subject in the safe area, and let the editor handle the mechanical assembly—not the creative judgment.

I render vertical reel compilations in the browser (honest freemium)

Aaron Colbry — Wed, 30 Sep 2026 15:14:17 +0000

Built ReelWorkshop so clip dumps become TikTok/Reels/Shorts compilations without juggling CapCut + Drive.

Import your own clips
Arrange / trim / preview in the browser
Export 9:16 H.264 MP4

Editing/preview are free; MP4 export on Starter ($19.99/mo). No auto-highlight or scheduling claims.

Demo: https://reelworkshop.com/demo

Happy to answer questions about the client-side pipeline.

ReelWorkshop — turn raw clips into Reels & TikToks in minutes

Aaron Colbry — Tue, 29 Sep 2026 20:55:50 +0000

Hey dev.to! I'm launching my project ReelWorkshop on Product Hunt tomorrow and wanted to share it with fellow builders first.

The problem: Batching short-form content was eating my week — importing clips, cutting, captioning, reformatting for TikTok/Reels/Shorts, on repeat.

What I built: ReelWorkshop, a web studio that collapses that workflow. Import raw clips, plan your sequence, render a publishable MP4 in minutes. No installs, works right in the browser.

Pricing: Free to start — editing and preview are free, and the Starter plan ($19.99/mo) only kicks in when you export the MP4.

Would love feedback from builders: what's the most painful part of your short-form workflow right now?

Try it: https://reelworkshop.com
Product Hunt launch: Wednesday, September 30

What I learned shipping a browser-based video editor: trimming, captions, and vertical export

Aaron Colbry — Fri, 25 Sep 2026 19:22:09 +0000

Short-form video has a brutal production bottleneck: you shoot ten minutes of footage and need thirty usable seconds — captioned, framed vertically, ready to post — before the trend moves on. Editing that on a phone is fiddly; editing it in a desktop suite means an install, an import, and a render queue.

The browser turns out to be a legitimate middle ground. <video>, <canvas>, MediaRecorder, and WebCodecs give you nearly the whole pipeline without leaving the tab. After spending a long stretch building and refining a browser-based video editor, here are the three areas where the real lessons live: trimming, captions, and vertical export.

1. Trimming is a UX problem, not a decoding problem

Decoding video in the browser is a solved problem — drop a file into a <video> element and it plays. The hard part is helping someone find the exact moment they want.

The first lesson: thumbnails beat timestamps. Scrubbing blind is miserable. Pre-generating a thumbnail strip by seeking a muted <video> element to intervals and painting frames to an offscreen canvas makes trimming feel instant, and it only needs to happen once per imported clip. On slower devices, generate fewer, larger thumbnails instead of a dense strip — perceived responsiveness matters more than precision at this stage.

The second lesson: seeks are async and approximate. Setting video.currentTime doesn't land exactly where you ask, and it definitely doesn't happen synchronously. The trim handles in your UI should snap to the seeked frame (listen for the seeked event) rather than the requested time. If the UI promises frame accuracy, the plumbing underneath has to deliver it.

The third lesson: preview the cut, not the clip. When someone drags a trim handle, the most useful feedback isn't the whole clip replaying — it's a short loop of the moment just before the cut point and the moment just after it. That micro-preview answers the only question that matters: "did I cut the good part off?"

2. Captions: render them, don't overlay them

Most short-form content is watched on mute, which makes captions non-negotiable. But how you deliver them matters:

Burn captions into the video rather than relying on platform caption files. Platforms recompress, reframe, and re-upload your video constantly — burned-in captions survive every re-upload, duet, and stitch. Caption sidecar files get lost the moment someone screen-records the clip.
Style for the platform, not the spec. Classic white text with a dark stroke still wins for one reason: it stays readable over any background. But positioning matters even more — keep captions inside the vertical safe zone (roughly the middle band of the frame) because TikTok, Reels, and Shorts all paint their own UI over the top and bottom of your video.
Word-by-word highlighting is worth the effort. Viewers follow along more easily when the currently spoken word is emphasized, karaoke-style. The implementation is straightforward: render each word as its own text run on the canvas using per-word timing data, highlight the active word, and dim the rest. It's more work than a single static block, but it's the biggest single readability upgrade you can make.
Generate the words, let humans fix them. Automatic speech recognition gets you most of the way there; the remainder — names, slang, brand terms — needs a fast correction UI. A caption editor that lets you click a word and retype it, with timing untouched, will get used constantly. One that forces you to re-time anything will get abandoned.

3. Vertical export: design the canvas first

Everything about short-form is vertical, and this is where browser editors either shine or fall apart. The key insight: treat the export canvas as the source of truth, not the preview.

Draw your final 1080x1920 composition on an offscreen canvas with drawImage, applying the same crop-and-scale math used everywhere else. Then capture that canvas with MediaRecorder (or WebCodecs when you need finer control) and mix in audio from the source clips. When the export canvas is the render target:

What the creator sees while editing and what gets exported can't drift apart.
Reframing landscape footage becomes a layout problem (pan-and-scan, blurred-fill backgrounds, subject tracking) instead of an export bug.
Effects — zooms, captions, overlays — live in one compositor instead of being bolted onto the file afterward.

Two practical gotchas: first, MediaRecorder output codecs vary by browser, so validate your container and codec choices on every target browser rather than assuming H.264-in-MP4 everywhere. Second, keep preview resolution and export resolution separate — preview at whatever the device handles smoothly, and only render full resolution during export. Your users' laptop fans will thank you.

The through-line

Trimming, captions, and vertical export look like three separate features, but they're really one design principle: reduce the distance between having footage and having a finished post. Every seek that lags, every caption that needs manual re-timing, and every export that looks different from the preview adds friction exactly where a creator's motivation is thinnest.

Keep the pipeline in the browser and that distance shrinks dramatically — no installs, no uploads parked on a server, no round-trips. Raw clips go in, and a vertical, captioned MP4 comes out, minutes later.

Disclosure: I build and maintain ReelWorkshop, a browser-based studio for turning raw clips into TikToks, Reels, and Shorts — this is where I ran into everything above. The techniques themselves are portable to any web video project.

I built a browser-based studio that turns raw clips into short-form compilations — here's what I learned

Aaron Colbry — Fri, 25 Sep 2026 17:32:20 +0000

Hey DEV — first post here. I wanted to share something I built and what building it taught me, in case it's useful to anyone working on creator tools or browser-based media.

What I built: ReelWorkshop (https://reelworkshop.com) — a browser-based studio that turns raw clips into polished short-form compilations for TikTok, Reels, and Shorts. The motivation: creators sit on hours of raw footage — stream highlights, project clips, event recordings — and the tedious middle (trimming, ordering, captioning, reframing to vertical) is where videos go to die.

What building it taught me:

Creators don't want another timeline. The biggest unlock was designing around "pick the good moments, we handle the assembly" instead of a traditional NLE. For this audience, constraint beats flexibility.
Captions are non-negotiable. So much short-form is watched on mute that anything which doesn't make captioning effortless simply doesn't get used.
The browser is enough now. Keeping the workflow client-side meant no upload waits and no render queues — and modern browser media APIs handle the heavy lifting.
Design vertical-first, not cropped-later. Reframing, safe zones for platform UI overlays, and pacing for the first two seconds all change when 9:16 is the starting point instead of an afterthought.

My biggest takeaway for anyone building creator tooling: watch someone do the task by hand first. Every feature in ReelWorkshop maps to a step I watched a creator do slowly, manually, in a desktop editor.

Open disclosure: I'm the builder of ReelWorkshop — this is my own project, not a third-party recommendation. Happy to answer questions about the build.