AI People Wiki

Builders & open source

Amir Hertz

Research scientist, Google

Amir Hertz is a research scientist at Google, developing algorithms for intuitive image editing and creation and working on the training and distillation of fundamental vision models. Hertz received a PhD at Tel Aviv University under Daniel Cohen-Or and Raja Giryes, with doctoral work on 3D shape generation, texture synthesis, 3D modeling, and meshing. Hertz is first author of "Prompt-to-Prompt Image Editing with Cross Attention Control" and co-first author of "Null-text Inversion for Editing Real Images using Guided Diffusion Models", both presented by authors from Google Research and Tel Aviv University and released in the Apache-2.0-licensed google/prompt-to-prompt repository.123456

At a glance

Current role
Research scientist
Company
Google
Sector
Company
Work listed
3
Honors
2
Sources cited
11
Profile reviewed

Career

Hertz completed a PhD in the Department of Computer Science at Tel Aviv University, supervised by Daniel Cohen-Or and Raja Giryes. The doctoral work produced models and methods for 3D shape generation, texture synthesis, 3D modeling and meshing.1

At Google, Hertz develops algorithms for intuitive image editing and creation and works on improving the training and distillation of fundamental vision models. Hertz describes the broader aim as adapting machine learning practice to computer graphics and vision so that design tools become easier to use.1

Research

Hertz's early papers dealt with geometry: MeshCNN applied convolutions to the edges of triangle meshes, Deep Geometric Texture Synthesis and PointGMM followed in 2020, and SPAGHETTI (SIGGRAPH 2022) let people edit implicit 3D shapes part by part.178

From 2022 the work turned to text-to-image diffusion models. Prompt-to-Prompt showed that images can be edited by changing only the text prompt while controlling cross-attention, and Null-text Inversion extended such editing to real photographs. Later papers include Delta Denoising Score (ICCV 2023), Style Aligned image generation (CVPR 2024) and The Chosen One, on keeping a character consistent across generated images (SIGGRAPH 2024).24910111

Work and contributions

  • Prompt-to-Prompt and Null-text Inversion Open-source code

    First author of Prompt-to-Prompt (2022) and co-first author, with Ron Mokady, of Null-text Inversion (2022); both project pages list the authors with Google Research and Tel Aviv University, and the google/prompt-to-prompt repository, under the Apache License 2.0, implements Prompt-to-Prompt on Latent Diffusion and Stable Diffusion and includes Null-text Inversion for editing real images with Stable Diffusion.23456

  • Style Aligned Image Generation via Shared Attention Published research

    First author, with Andrey Voynov, Shlomi Fruchter and Daniel Cohen-Or, of a method for generating images that share a consistent style using shared attention, presented as an oral paper at CVPR 2024.101

  • MeshCNN Published research

    Co-author, with first author Rana Hanocka, of MeshCNN, a convolutional neural network that operates directly on the edges of triangle meshes, published in ACM Transactions on Graphics (SIGGRAPH 2019).71

Timeline

  1. Co-authored MeshCNN, a neural network that operates on the edges of triangle meshes.7
  2. Published Deep Geometric Texture Synthesis (SIGGRAPH) and PointGMM (CVPR) as first author.1
  3. Released SPAGHETTI, a method for editing implicit shapes through part-aware generation.8
  4. Released Prompt-to-Prompt, on editing generated images through cross-attention control.2
  5. Released Null-text Inversion, co-first-authored with Ron Mokady, for editing real images with diffusion models.45
  6. Released Delta Denoising Score with Kfir Aberman and Daniel Cohen-Or.9
  7. Released Style Aligned Image Generation via Shared Attention, later an oral paper at CVPR 2024.101

Honors and recognition

  • Honorable Mention Award, Eurographics 2024 (SENS) · Eurographics, 20241
  • SIGGRAPH Honorable Mention Award (Word-As-Image for Semantic Typography) · ACM SIGGRAPH1

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

11 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Amir Hertz — homepageAmir Hertz · Official page · accessed Oct 21, 2026
  2. Prompt-to-Prompt Image Editing with Cross Attention Control (arXiv 2208.01626)arXiv · Paper · published Aug 2, 2022 · accessed Oct 21, 2026
  3. Prompt-to-Prompt — project page (Google Research, Tel Aviv University)Prompt-to-Prompt authors · Official page · accessed Oct 21, 2026
  4. Null-text Inversion for Editing Real Images using Guided Diffusion Models (arXiv 2211.09794)arXiv · Paper · published Nov 17, 2022 · accessed Oct 21, 2026
  5. Null-text Inversion — project page (Google Research, Tel Aviv University)Null-text Inversion authors · Official page · accessed Oct 21, 2026
  6. google/prompt-to-prompt README and LICENSE (Apache-2.0)Google (GitHub repository) · Repository · accessed Oct 21, 2026
  7. MeshCNN: A Network with an Edge (arXiv 1809.05910)arXiv · Paper · published Sep 16, 2018 · accessed Oct 21, 2026
  8. SPAGHETTI: Editing Implicit Shapes Through Part Aware Generation (arXiv 2201.13168)arXiv · Paper · published Jan 31, 2022 · accessed Oct 21, 2026
  9. Delta Denoising Score (arXiv 2304.07090)arXiv · Paper · published Apr 14, 2023 · accessed Oct 21, 2026
  10. Style Aligned Image Generation via Shared Attention (arXiv 2312.02133)arXiv · Paper · published Dec 4, 2023 · accessed Oct 21, 2026
  11. The Chosen One: Consistent Characters in Text-to-Image Diffusion Models (arXiv 2311.10093)arXiv · Paper · published Nov 16, 2023 · accessed Oct 21, 2026