# Fictionarry/ER-NeRF

[ICCV'23] Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis

Repository: https://github.com/Fictionarry/ER-NeRF
Canonical: https://ross.abutalabs.com/products/er-nerf
Homepage: https://fictionarry.github.io/ER-NeRF/
Language: Python
License: MIT
License Family: permissive
Last push: 2025-03-14T06:51:51+00:00

## Health v2 (maintenance only)
Score: 24/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 11, release rhythm 8, longevity 81
- inputs: {"age_days": 1143, "days_push": 537, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1260, forks 143 (observed 2026-08-28T04:04:10.026251+00:00)

## What it is
ER-NeRF is the official PyTorch implementation of an ICCV 2023 paper on region-aware Neural Radiance Fields for high-fidelity talking portrait synthesis. It generates realistic, audio-lips-synchronized 3D talking head videos from a portrait video and speech audio, with fast convergence and real-time rendering.

## Use cases
- synthesize a talking head video from a single portrait video and audio
- generate lip-synced avatar videos driven by speech
- render high-fidelity 3D neural talking portraits in real time
- train a NeRF-based talking head model on custom footage
- research baseline for audio-driven portrait synthesis
- create digital human avatars for video production

## When to choose
- you need high-fidelity, lip-synced talking portrait videos from short training clips
- you want real-time rendering with a small model size
- you are researching NeRF-based talking head synthesis and need a strong baseline

## When to avoid
- you need a plug-and-play production tool with no GPU training
- you work on Windows or without CUDA GPUs, since it targets Ubuntu with specific PyTorch/CUDA versions
- you need 2D-only or lightweight lip-sync without 3D reconstruction

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, nlp, audio-processing, video-processing, graphics, simulation
- domain: deep-learning, computer-vision, artificial-intelligence, graphics
- platform: python
- tags: nerf, talking-head, talking-portrait, audio-driven, neural-rendering, face-synthesis, iccv-2023, research-code, audio, video, linux, gpu

## Member repositories
- Fictionarry/ER-NeRF (main) score 24

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:10.026251+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 2026-08-30T05:04:21.317769+00:00, confidence not recorded.
  - readme: https://github.com/Fictionarry/ER-NeRF (fetched 2026-08-28T04:04:10.026251+00:00, sha 61a82c8cb551)
  - homepage: https://fictionarry.github.io/ER-NeRF/ (fetched 2026-08-29T12:17:00.150742+00:00, sha 7e03d259ac6b)
- Data as of 2026-08-30T08:39:29.467469+00:00.