PeopleofSuperintelligence

Lab leader

Patrick Esser

Patrick Esser is an AI researcher and co-founder of Black Forest Labs whose work helped shape modern image generation. With collaborators, he developed transformer and latent-diffusion methods for high-resolution synthesis, then coauthored research on diffusion-guided video editing. At Runway, he led applied research on the core model behind Stable Diffusion. A rectified-flow transformer paper he coauthored received the 2024 ICML Best Paper Award. His later papers include FLUX.1 Kontext for image generation and editing and Self-Flow for image, video, and audio synthesis.

Photo of Patrick Esser

Timeline

  1. 2020–2021

    Teaching transformers to draw

    He coauthors Taming Transformers for High-Resolution Image Synthesis with Robin Rombach and Björn Ommer. Their method combines convolutional networks and transformers to synthesize high-resolution images.

  2. 2021–2022

    Generating inside latent space

    With Rombach, Andreas Blattmann, Dominik Lorenz, and Ommer, he coauthors High-Resolution Image Synthesis with Latent Diffusion Models. The paper applies diffusion in pretrained autoencoders' latent space to lower computational demands while retaining image quality.

  3. 2022

    Research behind Stable Diffusion

    At Runway, he serves as a Principal Research Scientist and leads applied research on the core model behind Stable Diffusion.

  4. 2023

    Guiding video with diffusion

    He and four coauthors present a video diffusion method that edits footage from visual or text descriptions.

  5. 2024

    Winning with rectified flow

    He coauthored Scaling Rectified Flow Transformers for High-Resolution Image Synthesis, a transformer approach for high-resolution text-to-image generation. The paper received the ICML 2024 Best Paper Award.

  6. 2024

    Leading a new lab

    He leads Black Forest Labs alongside Robin Rombach and Andreas Blattmann.

  7. 2025–2026

    Generating across media

    In 2025, he coauthored FLUX.1 Kontext, a flow-matching model for image generation and editing. In 2026, he coauthored Self-Flow, a method for multimodal generation across image, video, and audio.

  8. As of October 2026

    A continuing lab role

    He is a co-founder of Black Forest Labs, a research lab building generative image models.

Sources and credits

Photo: ommer-lab.com; People - Computer Vision & Learning Group; creator not stated. License: Not stated by source. Source: ommer-lab.com. Cropped to a square and resized. Ask us to remove or replace this photo.

  1. Taming Transformers for High-Resolution Image Synthesis arxiv.org. Checked 14 October 2026.
  2. High-Resolution Image Synthesis with Latent Diffusion Models arxiv.org. Checked 14 October 2026.
  3. The research origins of Stable Diffusion research.runwayml.com. Checked 14 October 2026.
  4. Structure and Content-Guided Video Synthesis with Diffusion Models arxiv.org. Checked 14 October 2026.
  5. Scaling Rectified Flow Transformers for High-Resolution Image Synthesis arxiv.org. Checked 14 October 2026.
  6. ICML 2024 Awards icml.cc. Checked 14 October 2026.
  7. Latent Space: The AI Engineer Podcast - Generative Video WorldSim, Diffusion, Vision, Reinforcement Learning and Robotics — ICML 2024 Part 1 Transcript and Discussion podscripts.co. Checked 14 October 2026.
  8. Investing in Black Forest Labs a16z.com. Checked 14 October 2026.
  9. FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space arxiv.org. Checked 14 October 2026.
  10. Self-Supervised Flow Matching for Scalable Multi-Modal Synthesis arxiv.org. Checked 14 October 2026.
  11. The 70-Person AI Image Startup Taking on Silicon Valley's Giants wired.com. Checked 14 October 2026.
  12. Black Forest Labs - Northzone northzone.com. Checked 14 October 2026.