Ahmet Yıldırım

Video producer · Istanbul

“I have been making video for fifteen years. Agencies bring me the work, I deliver it. This year I started building the systems that do the work too.”

Hello, I am Ahmet. I studied Film and Television at İstanbul Bilgi University (2016). My first step onto professional sets was acting, in 2011 (PİS YEDİLİ · Show TV · Tükenmez Kalem · 2011–2014 · as Peker Canaydın). After that I worked in the camera department on a range of productions. In 2019 I sat down at a desk and I have not got up since. I have been making video for fifteen years. Agencies bring me the work, I deliver it. This year I started building the systems that do the work too.

Stills
FLUX
Video · lip-sync
InfiniteTalk
Voice
ElevenLabs
Produced locally in ComfyUI · 480×832 · 48 seconds

Work

video · design
7+years
70+brands
3000+videos delivered

Ongoing monthly social content

Storytel Türkiye: 63 consecutive months, 1,344 videos. Bitaksi: 11 months, 482 videos. Every month, on time.

Monthly content programStorytel Türkiye · AFF Reklam · 2020–2025
Monthly content programBitaksi · AFF Reklam · 2024–2025

Commercials and brand films

Films made for television and digital.

Opmar Optik — StingAFF Reklam · 2025 · 67 sec
TV spotKuzeyden · AFF Reklam · 2024

Motion graphics

This is the craft underneath most of the work on this page: social content, venue screens, and the animation inside commercials and corporate films. Almost everything I have delivered since 2019 passed through it.

Insurance explainer seriesZurich · BREX.CO · 2026
Bumper adEnterogermina · AFF Reklam · 2026

Corporate film

History films, internal communication, sales-meeting content, award submissions. I have worked for dozens of local and global companies, among them Fairy, Ariel, Unibaby, Duracell and Amazon.

Corporate history filmNestlé · BREX.CO · 2025
Campaign case videoMustela · BREX.CO · 2025

Logo animation

The moment a brand name first appears on screen.

Efes — bottle animationAFF Reklam · 2025 · 30 sec
RozyLabs — logo animation2026 · 7 sec

Event and venue screens

Kristal Elma award show: 280 animations, zero errors, in under ten days.

Award show screen graphicsKristal Elma · Res Publica · 2025
Bar signage animationEfes · AFF Reklam · 2025

Live-action shoots

From a one-person shoot to a multi-camera production, scale is not the issue. I build and run whatever crew the job needs. Cast, kit and schedule follow the work; I take the brief and deliver the film.

Product filmRozyLabs · 2026
Trade show filmMOSS Yacht · 2022

Music and live performance

2014–2019

I spent three seasons at JoyTurk Akustik as gaffer and cinematographer; most of Turkey’s best-known musicians passed through that studio. Başka Şarkılar was an archive channel recording live performances by independent musicians; I was a co-founder and its programme director.

Levent Yüksel — live performanceJoyTurk Akustik · 2014–2016
Demir Demirkan — Bana Müsademusic video · gaffer
Başka Şarkılar2017–2019
Derdiyoklar — concert recording2018

Short film

2012–2020

Between 2012 and 2019 I worked on many short films as cinematographer, gaffer and camera operator.

Veysel'in Kanatlarıdir. Metin Hasgül · 2020 · 13 min

Brand identity

A brand identity built from scratch for a personal brand based in Germany. The job did not start with a logo. First came a 34-page creative direction document, in German and Turkish.

The full guide · 34 pages

Moodboards and visual direction frames were generated with AI, 39 frames.

The logo was drawn last: exploration variants, the final mark, and the mark locked up with the brand name.

Aysel Hacer Sünger · 2026

How I work

method

I built an AI film crew

this year

I studied film at İstanbul Bilgi University and worked on sets in the lighting and camera departments. I started experimenting to see whether the way a set works could be applied to AI.

I did not write prompts. I built a film crew: every role is a separate step, the order never changes, no step is skipped.

ScreenwriterThe script. A list of every character, location and prop that has to stay consistent.
DirectorBreaks the scene into shots. Decides how long each shot runs and what counts as good enough.
Location managerDraws the floor plan. Decides what the camera sees in every direction, and what must never appear.
Art directorLocks wardrobe and props. Nothing changes colour or shape between shots.
GafferWrites the light: where it comes from, how hard it is, how deep the shadows go.
CinematographerFrames every shot. Keeps the camera on one side of the action so the cuts never disorient the viewer.
SupervisorChecks the plan with fresh eyes before anything is generated. One error and production does not start.

Opmar Optik · AFF Reklam · 2026
Veo · Flow · Nano Banana

I measure first

I did research for an English-language YouTube channel. I settled on the psychology niche, and before opening the channel I decided to work out how the videos that succeed in it are written.

The reason was this: I found the texts AI wrote too artificial. I wanted the rules I would hand the writing system to be measurements, not guesses.

I gathered the rules from three separate literature reviews: what persuades people, how a story is built, and how attention is held. More than sixty candidate rules came out. The story review started from two questions of my own: where is the conflict in this scene, and what does this piece do for the narrative.

Then I did not ask anyone whether they liked the rules. I tested them one by one against the actual transcripts of videos that had worked.

I wrote the question firstI wrote down what I was looking for and locked it before seeing the data, so I could not bend it to fit later.
I hid the sourceWhoever read the texts did not know which channel a video came from or whether it had performed.
Two separate readersTwo independent readers assessed the same texts without seeing each other’s work.
I planted a fake splitI split some of the videos at random instead of by performance. If differences turn up there too, what was found is not real.
I never opened half of itI set aside half the texts, keeping them to test whatever I found in the first half.
317,319videos scanned
1,396channels scanned
851videos transcribed
2.6Mwords of text
303hours of footage
856thumbnails reviewed

Of more than sixty rules, only one was confirmed by measurement. Titles, thumbnails and publishing frequency pointed nowhere: large channels and small ones alike were putting out nine to twelve videos a month, and what separated them was the age of the channel. The fake split produced differences at the same rate as the real one, which means the differences found were not real.

One result survived, and it was worth more than what I was looking for. When videos on the same subject are compared against each other, the subject stops being a variable and only the telling is left. The upper half had been watched 8.43 times more than the lower half.

I closed the research here. The problem turned out to be judgement, not measurement: two readers gave the same answer to all eight questions on only 40 percent of 308 texts. The same model had no such problem producing text. I built the channel accordingly: the model writes, the human decides.

The whole study took four days. The transcribed material runs to 303 hours; one person watching eight hours a day would need thirty-eight days just to watch it. The scanning, the transcription and the reading were done by AI.

What I learned is this: gathering data at this scale is no longer the hard part. The hard part is deciding what the data means.

Who I work with

2019 – 2026

I have worked with AFF Reklam since 2019, and with BREX.CO and Res Publica since 2022. Almost all of the work above came out of those three relationships.

20192020202120222023202420252026
AFF Reklamstaff first, then freelance
BREX.COfreelance
Res Publicafreelance
AdKingsfreelance
Storytel Türkiye63 months
Bitaksi11 months

What I work with

tools
Editing and designAfter Effects · Premiere Pro · Illustrator · Photoshop
Image and videoVeo · Flow · Higgsfield · Nano Banana · local generation and model training in ComfyUI
AudioElevenLabs · Udio · Whisper
AutomationI turn repeating work into scripts: I decide what it should do and whether it ran correctly, and have the AI write the code.
CrewA fixed three-person crew with their own cameras, drone, lenses and sound kit

Contact

Istanbul

ahmet@ahmetyildirim.studio
LinkedIn
Vimeo