---
title: TimeChat-Captioner
emoji: 🎬
colorFrom: purple
colorTo: yellow
sdk: gradio
sdk_version: 6.19.0
app_file: app.py
short_description: Time-aware audio-visual captions for multi-scene videos
python_version: "3.12"
startup_duration_timeout: 1h
---

# TimeChat-Captioner

Generate detailed, time-aware, and structurally coherent audio-visual captions for
multi-scene videos. Upload a short video clip (~60 seconds recommended) and get a
script-like description with timestamps that coordinates visual and audio information.

## Model

[yaolily/TimeChat-Captioner-GRPO-7B](https://huggingface.co/yaolily/TimeChat-Captioner-GRPO-7B) —
a 7B VLM fine-tuned from Qwen2.5-Omni-7B for dense video captioning.

## Paper

[TimeChat-Captioner: Scripting Multi-Scene Videos with Time-Aware and Structural Audio-Visual Captions](https://arxiv.org/abs/2602.08711) (ICML 2026)

## Usage

1. Upload a video (recommended ~60 second clips for best timestamps).
2. Optionally customize the prompt.
3. Click "Generate Caption" to get a structured, time-aware description.