ChenyangQiQi/FateZero

Health v2 · maintenance only

Adoption not part of the score

1163 stars · 109 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

FateZero is a zero-shot text-based video editing framework built on pretrained Stable Diffusion models, introduced in an ICCV 2023 Oral paper. It fuses attention maps captured during DDIM inversion to edit real-world videos from text prompts without per-prompt training or user-provided masks.

Use cases

• edit a real video with a text prompt without training
• apply Van Gogh or anime style to an existing video
• swap objects in a video like cat to lion
• change video attributes while preserving motion and structure
• zero-shot video style transfer with stable diffusion
• edit video frames consistently using diffusion models

When to choose

• you want text-driven edits on an existing video without training a model
• you need temporal consistency in diffusion-based video editing
• you want a research reference implementation for attention-fusion editing

When to avoid

• you need fast, production-grade video editing with low compute
• you want a polished end-user application rather than research code
• you need support for the latest diffusion models or active maintenance

Facets

library · maturity maintenance

video-processing image-processing machine-learning deep-learning image-processing artificial-intelligence computer-vision python cross-platform stable-diffusion diffusion-models zero-shot-editing video-editing text-driven-editing style-transfer research-paper iccv-2023 video gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="ChenyangQiQi/FateZero")