DEV Community: Alberto Loddo

Skip video normalization. Your AI pipeline will thank you.

Alberto Loddo — Sun, 17 May 2026 02:24:29 +0000

We had an ffmpeg normalization step at the start of our video processing pipeline. Every uploaded file got re-encoded to a standard format (H.264, 1080p, 30fps) before any AI analysis.

It seemed obviously correct. Standardize inputs, simplify downstream code, guarantee consistent behavior.

It was the most expensive mistake in our architecture.

The numbers

A user uploads 47 clips from a GoPro Hero 12. Total raw size: 12GB.

After normalization: 36GB. Three times larger. The re-encode expanded the files because our target bitrate was higher than GoPro's efficient HEVC encoding.

Processing time for normalization alone: 8 minutes. Before any actual AI work started.

Storage cost: we kept both raw and normalized copies "just in case." R2 bills tripled.

Why we removed it

We were building an AI video analysis pipeline using Gemini 2.5 Flash. The assumption was that Gemini needed normalized inputs.

We tested sending raw files directly. MP4, MOV, AVI, WebM, HEVC, H.264, different resolutions, different framerates.

Gemini handled all of them. Every format. Every resolution. No errors. No quality difference in the analysis output.

The normalization step existed because we assumed the AI needed clean inputs. It did not.

What changed

Before:

upload -> normalize (8 min, 3x storage) -> analyze -> segment -> render

After:

upload -> analyze -> segment -> render

Results:

Storage: 36GB down to 12GB per batch (70% reduction)
Processing: 8 minutes saved per import
Complexity: one fewer stage to maintain, debug, and monitor
R2 costs: dropped immediately

The lesson

Every pipeline has a step someone added "because it seemed right" that nobody questioned. For us it was normalization. For you it might be:

Resizing images before sending to a vision model (most handle arbitrary sizes)
Converting audio to WAV before analysis (most speech models accept MP3 natively)
Transcoding video before thumbnailing (ffmpeg can extract frames from any container)

The fix is always the same: test what happens when you remove the step. If downstream still works, the step was waste.

Technical details

Our ingest stage now does metadata extraction only: hash, probe (resolution, fps, codec, duration), orientation detection, and EXIF parsing. No file transformation.

# Before: normalize then analyze
normalized = await ffmpeg_normalize(raw_file, target="h264_1080p_30fps")
result = await gemini_analyze(normalized)

# After: skip straight to analysis  
metadata = await ffprobe_extract(raw_file)
result = await gemini_analyze(raw_file)  # Gemini handles any format

The raw file goes directly to Gemini. Segments get extracted on-demand from the source using seek + duration, not from pre-cut clips.

If you are building a video AI pipeline and your first step is re-encoding, try removing it. You might be surprised.

I built an AI that edits GoPro footage automatically. Here is how it works

Alberto Loddo — Fri, 15 May 2026 12:24:12 +0000

Every action camera owner knows the feeling. You come back from a surf trip, a mountain bike ride, or a ski weekend with 200 clips on your SD card. You tell yourself you'll edit them this weekend. You never do.

I've been sitting on GoPro footage from trips I took years ago. The editing process is just too painful: scrub through hours of shaky, boring footage to find the 30 seconds of gold, then figure out transitions, music timing, pacing. Most people give up and the footage sits on a hard drive forever.

So I built FirstCut Studio to fix this. You upload your raw clips, pick a vibe, and the AI does the rest. No timeline. No editing skills required. Just a highlight reel that actually looks good.

Here's how it works under the hood.

The Pipeline

FirstCut runs two separate pipelines: one for understanding your footage, one for creating the edit.

Import Pipeline (what happens when you upload clips):

Ingest - We extract metadata, compute file hashes, probe video properties, and detect orientation. One important decision: we do zero video normalization at this stage. Raw files go straight through.
Gemini Analyze - This is the core. We send each clip to Gemini 2.5 Flash with a structured prompt asking it to grade quality, identify scene boundaries, detect key moments (big air, crashes, scenic views, celebrations), and tag the emotional tone. Gemini returns JSON with timestamps and confidence scores.
Music Analysis - We run librosa for beat tracking on the audio, then pass it through Gemini for semantic understanding (is this a buildup? a drop? a chill section?).
Segment - Scene detection and clip extraction using the boundaries Gemini identified.

Render Pipeline (what happens when you hit "Create Edit"):

Narrative Planner - An LLM-driven composition engine that selects which clips to include, in what order, with what effects.
Music Timing - Beat-aligns every cut so transitions land on the beat. This is what makes auto-edits feel professional instead of random.
Render - A 3-pass memory-efficient FFmpeg render. We process segments sequentially with garbage collection between operations, keeping peak memory around 1.5GB instead of 8GB.
QC - Automated quality check validating EDL integrity and beat alignment accuracy.

The Interesting Engineering Challenge

The CreativeExecutionEngine bridges the gap between "make it feel energetic" and actual FFmpeg filter parameters.

The LLM outputs creative intent like "speed ramp into the jump, hold the apex, snap cut to landing." The engine maps that to concrete VFX: a 2x speed ramp with ease-in curve, a 0.5x slow-motion hold, and a 3-frame hard cut. We enforce hard caps (max 2 split screens, 3 speed ramps, 5 text overlays per edit) to prevent the AI from going overboard.

The stack: Next.js frontend, Python/FastAPI backend, Gemini 2.5 Flash for all video understanding, FFmpeg for rendering, and Cloudflare R2 for storage.

One thing I learned: skip video normalization. Early versions re-encoded every upload to a standard format before analysis. This tripled storage costs and added minutes to processing time. When I tested sending raw files directly to Gemini, it handled them perfectly. Removing normalization was the single biggest infrastructure win.

What Actually Happens

A user recently uploaded around 200 clips from a multi-day trip. The import pipeline processed all of them, Gemini graded each one, and the system identified the strongest moments across the entire collection. When they hit render, the narrative planner pulled the best footage, beat-matched everything, and delivered the final edit in minutes.

Try It

FirstCut is live at firstcutstudio.xyz with a free tier. Upload your forgotten GoPro footage and see what comes out. If you're a developer curious about the video AI space, I'd love to hear what you think.

Building in public, so reach out with questions about the architecture or suggestions for what to build next.

Insta360 Video Editing Workflow: From Raw 360 Footage to Final Edit

Alberto Loddo — Sun, 03 May 2026 06:27:42 +0000

Insta360 cameras capture everything. That is their superpower and their curse. You press record once and get a full spherical view. The footage looks incredible in the app. Then you try to actually edit it, and the workflow gets complicated fast.

The Insta360 Editing Problem

Insta360 footage is not like normal video. A single .insv file contains two fisheye streams that need stitching. The Insta360 app handles this automatically, but it locks you into their ecosystem for the critical reframing step.

Reframing is where you choose which direction the virtual camera points within the 360 sphere. This is genuinely powerful, but also time-consuming. Every clip needs individual attention.

Result: Insta360 footage takes 3 to 5 times longer to process than standard flat video.

The Workflow

1. Export Flat Video First

Export everything from the Insta360 app as flat MP4 files before editing in another tool. Set keyframes, export at target resolution, standard codec.

2. Organize Before You Edit

After export, you have a folder of flat MP4s with unhelpful filenames, mixed with footage from other cameras. Tools like FirstCut Studio can ingest all your clips and automatically segment, grade, and organize them by quality.

3. Choose Your Editing Path

Quick social clips: Stay in the Insta360 app. Its templates and direct sharing are fastest.

Longer multi-camera edits: Import exported flat files into your preferred editor alongside GoPro, drone, and phone footage.

4. Color Match Across Cameras

Insta360 skews cooler with more contrast than GoPro or DJI. Add warmth (+5 to +10) and reduce contrast slightly on Insta360 clips to match.

5. Use 360 View for B-Roll Discovery

Go back to original 360 files for hidden angles. One 30-second 360 clip can yield three or four distinct B-roll shots you didn't plan.

6. Manage Storage

A single minute of 5.7K 360 footage is roughly 500MB. Archive raw .insv files after export, keep flat MP4s in your working folder.

Summary

Shoot with Insta360 alongside other cameras
Reframe and export as flat MP4
Combine all footage in one location
Organize and find best clips (manually or with AI assistance)
Edit in your preferred NLE
Color match across cameras
Archive raw .insv files

The key: treat Insta360 footage as a source requiring pre-processing before it enters your main workflow. Once exported, it behaves like any other footage.

Best Video Editor for Action Cameras in 2026

Alberto Loddo — Sat, 02 May 2026 16:23:40 +0000

Action camera footage is a different beast from anything else you shoot. The files are massive, the clips are short and chaotic, and 80% of what you capture is unusable.

Most video editors were built for sit-down content. Action camera users have the opposite problem: 50 to 200 short clips from a single session, and the real challenge is finding the three minutes of good footage buried in the noise.

What Action Camera Footage Needs

High frame rates. 60fps or higher, some up to 240fps for slow motion.
Wide-angle distortion. Ultra-wide lenses need correction.
Volume. A single session can produce 30 to 100 clips.
Stabilization. Even with in-camera stabilization, post-processing helps.
Quick turnaround. Most creators want a shareable video within hours.

The Best Editors

GoPro Quik

GoPro's own app handles basics: imports from GoPro cameras, applies HyperSmooth, generates quick edits. But struggles with non-GoPro footage, minimal timeline control, desktop app discontinued.

DaVinci Resolve

Genuinely excellent and genuinely free. Industry-leading color grading. But the learning curve is steep and it's overkill for quick highlight reels.

Adobe Premiere Pro

Handles action footage competently with built-in Warp Stabilizer. But $22.99/month is hard to justify for hobbyists.

CapCut

Free and surprisingly capable. Decent stabilization, good templates. But no help finding the best moments in your footage, and unavailable in the US.

iMovie

Free with every Apple device. Simple and intuitive. But maxes out at two video tracks, no rating system, no way to quickly identify best footage.

FirstCut Studio

FirstCut Studio was built specifically for action camera footage. Upload raw clips from any camera (GoPro, DJI, Insta360, phone, drone) and the AI analyzes every clip, grades them S through C tier, and identifies the best moments.

The clip curation is the core value. Instead of scrubbing through 100 clips, you get an organized library sorted by quality. Then compose a highlight reel with automatic music beat matching.

How to Choose

Quick Instagram clip from today's ride: GoPro Quik or CapCut
Professional results, willing to learn: DaVinci Resolve
Hundreds of clips, need help finding the good stuff: FirstCut Studio
Already pay for Creative Cloud: Premiere Pro
Mac, dead simple: iMovie

The biggest misconception: the hard part is editing. It's not. The hard part is sorting through massive footage to find the moments worth editing. That's the problem FirstCut solves.

Is CapCut Banned? What Happened and What to Use Instead

Alberto Loddo — Sat, 02 May 2026 10:23:41 +0000

Yes, CapCut has been affected by the US ban on TikTok and other ByteDance-owned apps. Millions relied on CapCut as their go-to video editor, and the sudden removal from US app stores left a lot of creators scrambling.

CapCut was genuinely one of the best free video editors available. It wasn't banned because it was a bad product. It was banned because of who owns it.

What Happened to CapCut

CapCut is developed by ByteDance, the same company that owns TikTok. When US lawmakers moved to ban TikTok over national security concerns, CapCut got swept up in the same legislation.

Both apps were removed from the App Store and Google Play Store for US users. As of early 2026, CapCut remains unavailable for new downloads in the United States.

Does CapCut Still Work?

If you already had it installed, the app may still open for basic editing. But you won't receive updates, and OS updates will likely break compatibility over time.

The web version has been intermittently accessible from US IPs. Not reliable.

The desktop app works offline for local editing, but cloud features, templates, and AI tools depend on servers that may not respond to US requests.

Bottom line: existing installs are on borrowed time.

Best Alternatives After the Ban

FirstCut Studio

If you used CapCut for highlight reels, travel videos, or montages, FirstCut Studio is the closest replacement. Upload your clips, AI analyzes footage, builds a polished highlight reel with music and pacing. No timeline editing required. Free to try.

DaVinci Resolve

The free option professionals actually use. Full color grading, audio mixing, multi-track editing. Steep learning curve but nothing you can't do with it.

iMovie

Apple's free editor. Limited features but reliable, easy, and completely free for Mac/iPhone users.

Clipchamp

Microsoft's browser-based editor in Windows 11. Decent templates and basic auto-captioning. Closest to CapCut's simplicity on Windows.

Adobe Express

Adobe's free tier with templates and basic AI features. Projects can move into Premiere Pro if you outgrow it.

Comparison

Feature	FirstCut	Resolve	iMovie	Clipchamp	Adobe Express
Price	Free	Free	Free	Free	Free tier
AI editing	Auto reels	Limited	No	Basic	Basic
Learning curve	Minimal	Steep	Easy	Easy	Easy
Best for	Highlight reels	Full productions	Simple edits	Social media	Social media

The Bottom Line

CapCut was a great editor. The ban was about geopolitics, not product quality. But the 2026 landscape has enough options that you don't have to settle.

If you want effortless highlight reels, FirstCut Studio picks up where CapCut left off. If you want professional control, DaVinci Resolve is free. If you need something simple, iMovie and Clipchamp are already on your device.

Most important: if you still have CapCut installed, export your projects before you lose access.

GoPro Quik Alternatives 2026: What to Use Now

Alberto Loddo — Sat, 02 May 2026 09:54:27 +0000

GoPro Quik was once the go-to app for quickly editing action footage into shareable clips. But Quik has changed. GoPro has shifted its focus toward subscriptions and cloud storage.

This guide compares the best GoPro Quik alternatives for 2026.

Why People Are Leaving GoPro Quik

Subscription push. Many features now require a GoPro subscription ($49.99/year).
GoPro-centric. Phone, drone, or DSLR footage gets a clunky experience.
Stale editing tools. The automatic highlight reel feature has not evolved.
Performance issues. Lag, crashes, and sync issues with newer models.

1. FirstCut Studio: Best for AI-Powered Highlight Reels

FirstCut Studio takes your raw footage and automatically creates a highlight reel. Upload clips from any camera, the AI analyzes every clip (scene detection, quality grading S through C tier), selects the best moments, and syncs cuts to music beats. Free to try.

2. CapCut: Best Free Mobile Editor

Completely free, no watermark. Great for social media clips. But no automatic highlight reel generation.

3. iMovie: Best for Apple Users

Free on Apple devices, intuitive and reliable. But Apple only, limited effects, no AI features.

4. Adobe Premiere Rush: Best for Adobe Ecosystem

Cross-platform with Premiere Pro integration. Requires subscription.

5. DaVinci Resolve: Best for Quality-Focused Editors

Most powerful free editor available. Industry-leading color grading. But steep learning curve and overkill for quick edits.

Quick Comparison

Feature	FirstCut	CapCut	iMovie	Rush	Resolve
Price	Free	Free	Free (Apple)	Sub	Free
Auto highlight reel	Yes	No	No	No	No
AI clip selection	Yes	No	No	No	No
Beat-synced editing	Auto	Manual	No	No	Manual
Learning curve	None	Low	Low	Med	High

Which Should You Choose?

Quick highlight reel without editing: FirstCut Studio
Short social clips for free: CapCut
Apple user, basic editing: iMovie
Adobe ecosystem: Premiere Rush
Professional editing: DaVinci Resolve

If speed and simplicity are what drew you to Quik, FirstCut Studio picks up where Quik left off.