CitronLegacy/StarfireTT_H3

Starfire (Teen Titans) — H3

Creator: CitronLegacy

Dataset Description

A small, hand-captioned LoRA training set for Starfire from the animated series Teen Titans. It mixes still frames and short video clips, each paired with a sidecar .txt caption.

Intended for:

• Character / style LoRA training (image and video models)
• Mixed image+video captioning experiments
• Small-dataset fine-tuning tests

Dataset Structure

The dataset is uploaded as a single zip (StarfireTT_H3.zip) containing one flat folder of 36 files (~5 MB): 10 images, 8 videos, and 18 captions, no subfolders.

StarfireTT_H3.zip
└── StarfireTT_H3/
    ├── starfire_clips_v1_1.mp4
    ├── starfire_clips_v1_1.txt
    ├── starfire_clips_v1_2.mp4
    ├── starfire_clips_v1_2.txt
    ├── starfire_clips_v1_3.mp4
    ├── starfire_clips_v1_3.txt
    ├── starfire_clips_v1_4.mp4
    ├── starfire_clips_v1_4.txt
    ├── starfire_clips_v1_5.mp4
    ├── starfire_clips_v1_5.txt
    ├── starfire_clips_v1_6.mp4
    ├── starfire_clips_v1_6.txt
    ├── starfire_clips_v1_7.mp4
    ├── starfire_clips_v1_7.txt
    ├── starfire_clips_v1_8.mp4
    ├── starfire_clips_v1_8.txt
    ├── strfr_001.jpg
    ├── strfr_001.txt
    ├── strfr_005.jpg
    ├── strfr_005.txt
    ├── strfr_007.jpg
    ├── strfr_007.txt
    ├── strfr_010.jpg
    ├── strfr_010.txt
    ├── strfr_011.jpg
    ├── strfr_011.txt
    ├── strfr_015.jpg
    ├── strfr_015.txt
    ├── strfr_027.jpg
    ├── strfr_027.txt
    ├── strfr_032.jpg
    ├── strfr_032.txt
    ├── strfr_036.jpg
    ├── strfr_036.txt
    ├── strfr_040.jpg
    └── strfr_040.txt

Part	Files	Naming	Notes
Images	10 .jpg	strfr_<NNN>.jpg	Mixed sizes, from 416×934 to 1920×1080
Videos	8 .mp4	starfire_clips_v1_<1–8>.mp4	1.6–6.2 s at 25 fps (one clip is ~4 s at 1920×1060)
Captions	18 .txt	same basename as the media file	One caption per file

Image numbers are not contiguous (001, 005, 007, 010, 011, 015, 027, 032, 036, 040) because they were selected from a larger set.

Video properties

• Resolution: mostly 640×480 or 640×472; narrower crops at 286×472 and 382×472; one at 1920×1060
• Frame rate: 25 fps (one clip is ~28 fps)
• Duration: 1.6 s to 6.2 s

Caption format

Every caption starts with the trigger word strfr, followed by natural-language description of expression, outfit, action, and background. Video captions also include dialogue and camera/motion notes.

Image caption:

strfr, Close-up of her tilting her head slightly with a soft, faint smile, her bright green eyes looking off to the side. She wears a light lavender sleeveless top with long orange-toned red hair falling over her shoulders, set against a starry night sky, cinematic lighting

Video caption:

strfr, A female animated character and a light purple top with a green circular emblem on the chest, is shown in a close-up shot ... she says, "This symbol is not of the earth." ... set against a dark background with circuit board patterns and a distant city skyline.

Data Collection

The images and video clips were all taken from the official Teen Titans TV show and captioned by hand and with AI assistance.

Intended Uses

• Training a Starfire character LoRA
• Testing mixed image/video captioning
• Non-commercial research and education

Limitations

• Very small (10 images, 8 clips) — suited to LoRA training only.
• Resolution and aspect ratio are not uniform; resize/bucket as needed.
• Some captions have grammatical slips (e.g. "A female animated character and a light purple top").
• Single character.

Licensing

Important

All images and video clips are official content from the Teen Titans TV show and are NOT owned or created by CitronLegacy. Teen Titans and Starfire are the intellectual property of Warner Bros. Discovery / DC. This repository is shared as a research/training-format example. Users are responsible for ensuring their use complies with applicable copyright law and the rights of the original holders.

Citation

If you use this dataset, I'd appreciate if you cite this Hugging Face repository.