README.md

---
license: apache-2.0
base_model: Qwen/Qwen2.5-0.5B-Instruct
tags:
  - gguf
  - ffmpeg
  - command-generation
  - lm-studio
  - ollama
---

# FFMPEG Command Generator (GGUF)

Fine-tuned Qwen2.5-0.5B that generates FFMPEG commands from natural language with chain-of-thought reasoning.

## Quick Start

### LM Studio
```bash
lms import kingjux/ffmpeg-command-generator-gguf
```

### Ollama
```bash
ollama run hf.co/kingjux/ffmpeg-command-generator-gguf
```

## Example

**Input:** "Convert video.mp4 to webm format"

**Output:**
```
<think>
Task: Convert MP4 to WebM
- WebM uses VP9 video + Opus audio
- Use -c:v libvpx-vp9 for video
- Use -c:a libopus for audio
</think>

ffmpeg -i video.mp4 -c:v libvpx-vp9 -c:a libopus output.webm
```

## Training
- Base: Qwen2.5-0.5B-Instruct
- Method: LoRA fine-tuning (r=16, alpha=32)
- Dataset: 30 FFMPEG command examples with CoT reasoning
- Trained on HuggingFace Jobs (T4 GPU)