Miju Labs

← All buyers · Krea

V

Victor Perez

Co-founder & CEO · Krea

medium confidenceTier A · Talk to firstImage, video & audio modelsimage · video

Perez co-founded Krea with Diego Rodriguez and runs it as CEO. Krea's whole positioning is taste: an AI creative platform that amplifies a user's own aesthetic rather than averaging everyone's, with stated plans for hyper-personalisation so creators get models that understand their individual preferences. That is the same thesis as Julian's business, stated by a CEO — which makes Krea either a natural partner or a competitor for the same idea.

Background

Perez and Diego Rodriguez are both from Spain and met at Cornell. They built Krea as a real-time AI design tool and grew it into a full creative workspace covering image, video and animation, backed by Bain Capital Ventures and others, with reported ARR in the mid single-digit millions by 2025 and around $83M raised.

Krea's differentiator is aesthetic opinion: it post-trained FLUX.1 Krea in collaboration with Black Forest Labs specifically to escape the generic 'AI look', and shipped Krea 1, Krea 2 and Krea Realtime on the same principle.

What they run now

Krea's product and strategy: a single workspace for designers and creative teams, models trained to a deliberate aesthetic, and a stated roadmap toward hyper-personalisation where a creator's own taste becomes the model's. Head of Design Annie Clark has framed the product as amplifying a user's creative identity through refinement and personalisation.

What they buy

Inference. As CEO he approves spend; his team has publicly said they do preference labelling in-house and distrust external preference datasets, so a straight data sale is unlikely. What Krea's roadmap actually implies is the opposite trade: if they want per-creator personalised models, they need per-person revealed-preference capture at scale, which is a product and infrastructure problem they may prefer to own. Krea is a small company — reported ARR in the millions, not hundreds of millions — so budget is a real constraint.

Career

  1. 2022 – presentCo-founder & CEO, Krea — San Francisco
  2. c. 2022Student, Cornell University
  3. to c. 2022Cornell University, Undergraduate (with co-founder Diego Rodriguez)

On the record

creator control · 2025-04

Says the mission is to put creators in control with an AI collaborator that adapts to their vision. source

amplify not automate · 2025-04

Says AI should amplify human imagination rather than automate it away — made by you with AI, never just by AI. source

personalised taste · 2025-04

Krea has described plans for hyper-personalisation so creators get models that understand their own taste and aesthetic preferences. source

Interviews & talks

Why it matters here

Moderate as a buyer, high as an intellectual benchmark. Perez has articulated Julian's thesis from the buyer's chair, which is useful evidence that the market is real. But Krea's stated preference is to keep taste work in-house, and its budget is small. Treat Krea as a proof point to cite to bigger buyers (BFL, Luma, Adobe-class companies) rather than a revenue target.

How to reach

Speculation: if approaching at all, approach as a peer with a shared thesis rather than a supplier — e.g. a short written argument about how to measure whether a personalised model actually matches an individual's taste, which is the unsolved problem in Krea's own roadmap and the thing Chung's LiteraryTaste work quantifies. He is active in startup media and reachable on LinkedIn.

What we could not establish
  • Krea's current funding, valuation and headcount in 2026.
  • Whether Krea has any external data spend.
  • Whether Diego Rodriguez or Perez owns model-side decisions.

Why they are on the map

CEO; listed on the Krea 2 technical report; sets the aesthetic direction. source — evidence dated 2026-06

Decides whether the problem is funded.

How Krea buys data

Krea 2 tech report (2026-06-23) says preference annotations were 'collected entirely in house, by people familiar with the specific strengths, weaknesses, and quirks of the model', and that they prototyped reward models with Thinking Machines' Tinker. FLUX.1 Krea post (2025-07) argues that generic preference datasets hurt aesthetics and that under 1M high-quality examples is enough. May 2026: a Contra Labs study using working creatives benchmarked Krea 2 Large on style fidelity. 2026 'ML Researcher - Posttraining' role will 'coordinate with data teams and partners to manage collection of preference data' and 'train custom VLM as reward models'.

Small team (~17 in April 2025, grown since). Founders lead. Research covers Diffusion and Posttraining (RL, reward VLMs, evals). There is an internal data team that does annotation, and a 'Personalization & Recommendation' ML role models user taste.

Known vendors: contra-labs, Thinking Machines Tinker (infra, not data)

Sources

  1. A Powerful New Platform for Kreatives (Bain Capital Ventures, 2025-04-07)
  2. Krea revenue and funding
  3. Releasing Open Weights for FLUX.1 Krea (Sangwu Lee & Erwann Millon, 2025-07-31)
  4. Krea 2 Technical Report