# guandeh17/Self-Forcing

Official codebase for "Self Forcing: Bridging Training and Inference in Autoregressive Video Diffusion" (NeurIPS 2025 Spotlight)

Repository: https://github.com/guandeh17/Self-Forcing
Canonical: https://ross.abutalabs.com/products/self-forcing
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2025-09-12T03:31:30+00:00

## Health v2 (maintenance only)
Score: 37/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 41, release rhythm 35, longevity 32
- inputs: {"age_days": 450, "days_push": 355, "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 3488, forks 282 (observed 2026-08-28T04:08:07.114724+00:00)

## What it is
Official implementation of Self Forcing, a training method for autoregressive video diffusion models that simulates inference during training to close the train-test gap. It enables real-time streaming text-to-video generation on a single RTX 4090 with quality matching state-of-the-art diffusion models.

## Use cases
- generate video from text prompts in real time
- stream video generation on a single consumer GPU
- train autoregressive video diffusion models
- run a text-to-video demo locally
- experiment with diffusion model training techniques
- generate videos chunk-by-chunk with KV caching

## When to choose
- you need real-time or streaming video generation on a single 24GB GPU
- you want to reproduce or build on the Self Forcing research
- you need autoregressive chunk-wise video generation with low latency

## When to avoid
- you need production-grade video generation without GPU hardware
- you want a polished end-user video generation product rather than research code
- you lack a Linux machine with an NVIDIA GPU and 64GB RAM

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, video-processing, llm-inference, gpu-computing
- domain: deep-learning, artificial-intelligence, image-processing
- platform: python
- tags: video-diffusion, autoregressive-generation, text-to-video, streaming-generation, diffusion-models, research-code, neurips-2025, video, linux, gpu

## Member repositories
- guandeh17/Self-Forcing (main) score 37

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:07.114724+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 2026-08-29T18:36:06.567354+00:00, confidence not recorded.
  - readme: https://github.com/guandeh17/Self-Forcing (fetched 2026-08-28T04:08:07.114724+00:00, sha 2aaf89d9fcb7)
- Data as of 2026-08-30T08:39:29.467469+00:00.