Creating Tech Packs 4x Faster with AI Image Gen at One Jeanswear Group

With 50+ tech packs due and almost no time, I built a ChatGPT workflow that sketched garments 4x faster. The workflow spread across the team, saving 20+ hours a week, and we went to market on time with our strongest Summer season in five years.

Project Overview

A tech pack is the blueprint a factory uses to cost and produce a garment. It includes a technical sketch, construction callouts, measurements (specs), and a bill of materials (BOM). Each one takes about an hour, and the sketch is the slowest part. I built an AI workflow that made sketching 4x faster and helped our team finish 50+ tech packs in under a week.

Role: Senior Designer, Wovens
Team: Design Director, Senior Designer Wovens (me), Senior Designer Knits
Tools: ChatGPT, Adobe Illustrator
Timeline: 1 week


Outcome:

  • 50+ tech packs in under a week.

  • Sketch time cut from 30 minutes to about 7.

  • 20+ hours saved per week across the team.

  • Our most successful Summer season in five years.

 

The Problem

It’s Summer 2025. Our buyer meeting is two weeks away, and we don’t have a collection. We go sample shopping and pull together a line of 50 styles. The line looks great, but there are no tech packs behind it. Without tech packs, production can’t get costing and sales can’t present the line.

What follows is days of overtime to hit the deadline. The team burns out, tension builds between colleagues, and eventually one team member quits.

Now it’s Summer 2026, and we’re facing the same problem: more than 50 tech packs to finish and no time to do them. Over the past year, I’ve been experimenting heavily with image generation tools, including ChatGPT, Midjourney, Gemini, and Grok. Determined not to repeat last year, I set out to build a faster way.

 

Finding The Bottleneck

I started by mapping how long each part of a tech pack took:

  • Sketching: 30 minutes

  • Specs and BOMs: 20 minutes

  • Callouts: 10 minutes

Sketching took the most time, and it’s also the most critical step. Factories need detailed sketches to cost garments accurately, and a costing mistake can cause major problems once styles are bought.

 

Using ChatGPT to Create Fashion Sketches

I tested AI image generation tools like Midjourney, Grok, and ChatGPT and chose ChatGPT because it produced the most technically accurate sketches. I asked it to write a sketch prompt using a Task–Context–Reference structure and gave it examples of my existing sketches so the results would match my style. I firmly believe image outputs need image references to get the best results.

Once the chat was set up with the prompt, the workflow was simple: upload a photo of the sample and get a sketch back.

“Task: Create a flat technical sketch of the garment in the attached photo, showing front and back views, for use in a factory tech pack.

Context: Factories will use this sketch to cost and produce the garment, so construction accuracy matters more than style. Use clean black line art on a white background, with no shading, color, or model. Show every seam, dart, topstitch, closure, pocket, and trim exactly as it appears. Keep proportions true to the garment, and don’t add or remove details.

Reference: Match the line weight, proportions, and layout of the attached example sketches.”
 

Handling Regressions

About half of the sketches needed at least one revision, usually to refine construction details. I handled regressions in two ways.

1. Re-prompting in ChatGPT. I used ChatGPT’s edit function to describe the changes I needed, and added photos wherever I could.

[Example: before/after of a revision]

2. Manual override. When several rounds of edits didn’t fix a sketch, I brought it into Adobe Illustrator and drew over it by hand. It isn’t a perfect solution, but it’s still faster than sketching from a blank canvas.

 

Outcome

The Knits designer adopted my workflow too, so the time savings doubled across the team.

  • Sketch time: 30 minutes cut to about 7 (4x faster)

  • Full tech pack: about 60 minutes cut to about 37

  • 50+ tech packs completed in under a week

  • 20+ hours saved per week across the team (10+ each)

Factories costed the styles accurately, and with sketches and costing ready on time, we went to market fully prepared. Buyers bought more styles than in any Summer season in the past five years.

 

Learnings

  • Image outputs need image references. My existing sketches did more for quality and consistency than any wording in the prompt.

  • Keep a human in the loop. With half the sketches needing revision, having a clear fallback kept speed from costing accuracy.

  • Build it to be shared. A ready-made chat that only needed a photo upload made it easy for a teammate to adopt.