One recording, a set of vertical clips
Cut on sentence boundaries with burned-in captions, reframed so the speaker stays in frame, and loudness matched across every clip so none sounds quieter than the next when they post in a row.
I cut, caption, reframe, retouch and generate media — and I run the whole thing on my own hardware, so a shot that isn't right gets re-done instead of billed as a retry. Delivered at the platform's spec, not "close enough".
Every service below has a working demo behind it — not a promise. Pick one, or hand over the whole pipeline.
One long recording into a set of vertical clips — each cut on a complete sentence, so nothing opens mid-word or stops halfway through a thought.
Reels · TikTok · ShortsBurned-in or as SRT/VTT, with word-level timing. Checked with a second pass before delivery — a single auto-transcription pass is never sold as reviewed.
Sound-off readyOne video delivered as 9:16, 1:1 and 4:5 from a single pass, each safe-area checked so nothing important lands under a platform's UI.
One source, three cutsCut on the action, colour and brightness matched shot to shot, music bed ducked under the voice, titles and end card, master at −16 LUFS.
The whole editAnimated titles, intros, outros and lower thirds, including per-word reveals and true kinetic typography timed to the audio.
Motion graphicsScript, voiceover, captions and music built end to end, delivered in 9:16, 16:9 and 1:1. Rendered locally, so scenes get re-rolled rather than re-billed.
Full commercial rightsA person, sign, logo or cable taken out of a clip or a still with the background rebuilt and grain and motion matched — or a new element added and lit into the scene. Audio untouched.
Video + photosHigher resolution and smoother motion — 24fps to 48fps without the ghosting that frame-doubling produces. Generated at 720p, delivered crisp at 1080p.
Sharper, smootherHum and hiss removal, de-essing, silence trim, level matching and loudness targets — streaming, podcast or broadcast — with a written report of what could and couldn't be fixed.
Measured, not guessedNarration in 13 languages, cued per line, mixed and level-verified. For Serbian, Croatian and Bosnian I read check the output by ear before it ships.
13 languagesThumbnails with A/B variants and source files at platform sizes, product photography and packaging, seamless texture packs and matching cover art.
With source filesA repeatable pipeline that processes hundreds of files unattended — caption, resize, watermark, convert, publish. This is the one most editors can't offer.
The high-ticket oneA site that loads fast and gets found — WordPress when you need a CMS, hand-built HTML when you don't, with the canonical tags, sitemap and security headers done rather than promised.
Sites & shopsA persistent agent (Hermes) wired into your own systems: scheduled work that runs, checks itself and reports back — the same stack that runs this business.
Runs unattendedNine pieces below. Each one was produced here, on this hardware, and the numbers quoted are measured from the actual output.
Cut on sentence boundaries with burned-in captions, reframed so the speaker stays in frame, and loudness matched across every clip so none sounds quieter than the next when they post in a row.

Measured on the delivered file: hiss energy down 11 dB, level variance between sentences cut from 8.1 to 5.9 dB, loudness range tightened from 9.6 to 8.5 LU, and the master lifted from −27.0 to −16.5 LUFS with a −1.5 dBTP ceiling so streaming platforms stop turning it down.
The subject taken out of a moving shot and the background rebuilt behind them, grain and motion matched so the repair doesn't announce itself. Audio left untouched.
A new element placed, scaled, lit and shadowed into an existing shot, with a 2× crop of the changed area so you can judge the join rather than take my word for it.

Vertical 9:16 for Reels, TikTok and Shorts, square 1:1 and 4:5 for the feed, each safe-area checked and with hook text burned in at the top.
Built around one message rather than a feature list: script, voiceover, word-level burned-in captions so it works with the sound off, a music bed mixed to −16 LUFS, delivered in three ratios.

Cut out, relit and composited onto a studio background with a real shadow, text baked on where the packaging needs it, and the set delivered at a consistent size.

Read the target brand's typography, palette and tone off their live site, then built a carousel, a thumbnail, a Story cover and a stat card inside that system. Self-initiated concept work.
Six vertical cuts, 49–59 s each at 1080×1920, every one built on the same structure, so producing the next is a re-cut rather than a rebuild. Captions burn in and highlight word by word, so the ad still lands with the sound off. Made to promote my own services, using the same pipeline I would run for a client.
Three history films — the transcontinental railroad, the Panama Canal and the Eiffel Tower — each built from a sourced research dossier, with contested figures spoken as ranges and the archival photographs credited on screen. These are examples — new films go up regularly.
Everything shown is produced in-house — demo and spec pieces built to demonstrate technique, not client commissions. Where a figure is quoted, it is measured from the delivered file. I don't show client work I can't verify, and I don't invent testimonials.
The same approach — finished output, measured claims, honest limits — applied to systems and software.
Fifteen scheduled jobs running unattended: a content engine, analytics pulls, source monitoring, off-machine backups, and a guard that parks idle GPU containers.
Open → 02A self-hosted WordPress AI chatbot plugin with its own licensing and analytics, released across seven versions — plus a free webhook plugin.
Open → 03Vertical video built scene by scene on a narrative arc: problem, cost, fix, proof, close. Six complete sets, with the matching static creative.
Open →The platform, the format, the deadline. If the brief is thin I'll say what I'd need rather than guess and hope.
A real deliverable you can react to, not a mood board. Rendering happens locally, so revisions don't cost anyone per clip.
Delivered in the sizes, codecs and loudness the platform actually wants, plus the source files where they're part of the service.
Clear notes in, clear changes out. If something can't be fixed, I'll tell you which and why instead of quietly shipping it.