# Picsart-AI-Research/StreamingT2V

[CVPR 2025] StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text

Repository: https://github.com/Picsart-AI-Research/StreamingT2V
Canonical: https://ross.abutalabs.com/products/streamingt2v
Homepage: https://streamingt2v.github.io/
Language: Python
License Family: other
Topics: long-video-generation
Last push: 2025-03-27T08:38:48+00:00

## Health v2 (maintenance only)
Score: 31/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 13, release rhythm 35, longevity 64
- inputs: {"age_days": 898, "days_push": 524, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1630, forks 157 (observed 2026-08-28T04:05:13.975693+00:00)

## What it is
StreamingT2V is a research codebase (CVPR 2025) implementing an autoregressive technique that turns short text-to-video diffusion models like SVD and ModelScope into long, temporally consistent video generators. It supports generating videos of 200 to 1200+ frames with smooth transitions, rich motion, and text/image alignment.

## Use cases
- generate long videos from a text prompt
- extend a short text-to-video model to minutes-long output
- generate video from an image with consistent motion
- produce temporally consistent AI video without hard cuts
- run diffusion-based video generation on a single GPU
- experiment with autoregressive video diffusion research

## When to choose
- you need long (8s to 2min+) AI-generated videos with smooth transitions
- you want to build on SVD or ModelScope as a base video model
- you have a CUDA GPU with at least 24 GB VRAM and can run Python research code
- you're reproducing or extending CVPR 2025 video generation research

## When to avoid
- you need a production-ready product with a polished UI or API
- you only need short 16-24 frame clips, where standard T2V models suffice
- you lack a high-VRAM GPU (default config needs 60 GB)
- you require a permissive or clearly defined license - the repo has none
- you need real-time or low-latency video generation

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

## Member repositories
- Picsart-AI-Research/StreamingT2V (main) score 31

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:13.975693+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 2026-08-30T03:47:52.641813+00:00, confidence not recorded.
  - readme: https://github.com/Picsart-AI-Research/StreamingT2V (fetched 2026-08-28T04:05:13.975693+00:00, sha 5ae58168d2eb)
  - homepage: https://streamingt2v.github.io/ (fetched 2026-08-29T11:20:38.693869+00:00, sha 4a8425be77ab)
- Data as of 2026-08-30T08:39:29.467469+00:00.