# TencentQQGYLab/ELLA

ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment

Repository: https://github.com/TencentQQGYLab/ELLA
Canonical: https://ross.abutalabs.com/products/ella
Homepage: https://ella-diffusion.github.io/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2024-07-17T10:02:15+00:00

## Health v2 (maintenance only)
Score: 25/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 64
- inputs: {"age_days": 909, "days_push": 777, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1289, forks 65 (observed 2026-08-28T04:04:15.225856+00:00)

## What it is
ELLA is an Efficient Large Language Model Adapter that equips text-to-image diffusion models with LLM-based text understanding via a Timestep-Aware Semantic Connector, without retraining the U-Net or LLM. It also introduces DPG-Bench for evaluating dense prompt following and includes an EMMA extension for multi-modal prompts.

## Use cases
- improve dense prompt following in stable diffusion
- generate images from long complex prompts
- adapt SD1.5 with an LLM text encoder
- benchmark text-to-image models on dense prompts
- use ELLA in ComfyUI workflows
- enable multi-modal prompts in diffusion models

## When to choose
- your diffusion model struggles with dense prompts with multiple objects and relationships
- you want LLM-enhanced text alignment without retraining U-Net or LLM
- you need a benchmark for dense prompt following

## When to avoid
- you need a production-ready text-to-image service out of the box
- you work with diffusion models other than SD1.5 without adaptation
- you lack GPU resources for LLM-backed inference

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, image-processing
- domain: artificial-intelligence, machine-learning, large-language-models, image-processing
- platform: python
- tags: text-to-image, diffusion-models, stable-diffusion, semantic-alignment, adapter, tsc, dpg-bench, comfyui, gpu, linux

## Member repositories
- TencentQQGYLab/ELLA (main) score 25

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:15.225856+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:55:42.922981+00:00, confidence not recorded.
  - readme: https://github.com/TencentQQGYLab/ELLA (fetched 2026-08-28T04:04:15.225856+00:00, sha 26eff25954e4)
  - homepage: https://ella-diffusion.github.io/ (fetched 2026-08-29T12:11:40.939836+00:00, sha fa6aa11fbfaa)
- Data as of 2026-08-30T08:39:29.467469+00:00.