Datasets:
Add README and metadata.jsonl for all configurations
Browse files
README.md
ADDED
|
@@ -0,0 +1,144 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
language:
|
| 4 |
+
- multilingual
|
| 5 |
+
pretty_name: Veo 3 Video Generation Dataset
|
| 6 |
+
size_categories:
|
| 7 |
+
- 1K<n<10K
|
| 8 |
+
task_categories:
|
| 9 |
+
- text-to-video
|
| 10 |
+
- image-to-video
|
| 11 |
+
tags:
|
| 12 |
+
- video
|
| 13 |
+
- veo3
|
| 14 |
+
- google
|
| 15 |
+
- text-to-video
|
| 16 |
+
- image-to-video
|
| 17 |
+
- generation
|
| 18 |
+
- prompts
|
| 19 |
+
configs:
|
| 20 |
+
- config_name: veo3
|
| 21 |
+
data_files:
|
| 22 |
+
- split: train
|
| 23 |
+
path: videos/veo3/train/metadata.jsonl
|
| 24 |
+
- config_name: veo3-audio
|
| 25 |
+
data_files:
|
| 26 |
+
- split: train
|
| 27 |
+
path: videos/veo3-audio/train/metadata.jsonl
|
| 28 |
+
- config_name: veo3-audio-off
|
| 29 |
+
data_files:
|
| 30 |
+
- split: train
|
| 31 |
+
path: videos/veo3-audio-off/train/metadata.jsonl
|
| 32 |
+
- config_name: veo3-fast
|
| 33 |
+
data_files:
|
| 34 |
+
- split: train
|
| 35 |
+
path: videos/veo3-fast/train/metadata.jsonl
|
| 36 |
+
- config_name: veo3-fast-audio
|
| 37 |
+
data_files:
|
| 38 |
+
- split: train
|
| 39 |
+
path: videos/veo3-fast-audio/train/metadata.jsonl
|
| 40 |
+
- config_name: veo3-fast-audio-off
|
| 41 |
+
data_files:
|
| 42 |
+
- split: train
|
| 43 |
+
path: videos/veo3-fast-audio-off/train/metadata.jsonl
|
| 44 |
+
---
|
| 45 |
+
|
| 46 |
+
# Veo 3 Video Generation Dataset
|
| 47 |
+
|
| 48 |
+
[English](#english) | [Português do Brasil](#portugues-do-brasil)
|
| 49 |
+
|
| 50 |
+
## English
|
| 51 |
+
|
| 52 |
+
### Summary
|
| 53 |
+
|
| 54 |
+
A collection of AI-generated videos created with Google's Veo 3 family of models. Each record contains the original text prompt, the model variant used, the generated video, and (when applicable) the input reference image. Videos are organized into one configuration per model variant.
|
| 55 |
+
|
| 56 |
+
- Videos: **5,811**
|
| 57 |
+
- Input images: **1,354**
|
| 58 |
+
- Configurations: **6**
|
| 59 |
+
- Language of prompts: multilingual (primarily English)
|
| 60 |
+
|
| 61 |
+
### Configurations
|
| 62 |
+
|
| 63 |
+
| Config | Videos | With input image |
|
| 64 |
+
|---|--:|--:|
|
| 65 |
+
| `veo3` | 1,333 | 309 |
|
| 66 |
+
| `veo3-audio` | 1,081 | 355 |
|
| 67 |
+
| `veo3-audio-off` | 416 | 0 |
|
| 68 |
+
| `veo3-fast` | 1,342 | 287 |
|
| 69 |
+
| `veo3-fast-audio` | 1,223 | 403 |
|
| 70 |
+
| `veo3-fast-audio-off` | 416 | 0 |
|
| 71 |
+
|
| 72 |
+
### Schema
|
| 73 |
+
|
| 74 |
+
Each configuration has a `metadata.jsonl` file alongside the video files:
|
| 75 |
+
|
| 76 |
+
```json
|
| 77 |
+
{"file_name":"videos/veo3/train/00000001.mp4","prompt":"A cinematic...","model":"veo3","input_image_file_name":"images/veo3/train/00000001_input.png"}
|
| 78 |
+
```
|
| 79 |
+
|
| 80 |
+
- **file_name** — relative path to the video file.
|
| 81 |
+
- **prompt** — the text prompt used for generation.
|
| 82 |
+
- **model** — Veo 3 model variant.
|
| 83 |
+
- **input_image_file_name** — path to the reference image (null for text-to-video).
|
| 84 |
+
|
| 85 |
+
### Usage
|
| 86 |
+
|
| 87 |
+
```python
|
| 88 |
+
from datasets import load_dataset
|
| 89 |
+
|
| 90 |
+
# Load a specific model configuration
|
| 91 |
+
dataset = load_dataset("artificialguybr/veo3-video-prompts", "veo3-fast")
|
| 92 |
+
```
|
| 93 |
+
|
| 94 |
+
### Provenance and rights
|
| 95 |
+
|
| 96 |
+
Videos were generated using Google's Veo 3 models via prompts submitted by users in an online community. Account identifiers, message identifiers, timestamps, and source-specific references were removed from the metadata. Google's Gemini API Additional Terms state that Google does not claim ownership of generated content, but this does not guarantee copyrightability or absence of third-party rights. Some prompts reference copyrighted characters, trademarks, or public figures. The dataset is labeled `license: other`. Users are responsible for assessing applicable rights and intended use. Removal requests can be sent to the dataset maintainer through the Hugging Face profile.
|
| 97 |
+
|
| 98 |
+
## Português do Brasil
|
| 99 |
+
|
| 100 |
+
### Resumo
|
| 101 |
+
|
| 102 |
+
Coleção de vídeos gerados por IA criados com a família de modelos Veo 3 do Google. Cada registro contém o prompt de texto original, a variante do modelo usada, o vídeo gerado e (quando aplicável) a imagem de referência de entrada. Os vídeos estão organizados em uma configuração por variante de modelo.
|
| 103 |
+
|
| 104 |
+
- Vídeos: **5,811**
|
| 105 |
+
- Imagens de entrada: **1,354**
|
| 106 |
+
- Configurações: **6**
|
| 107 |
+
- Idioma dos prompts: multilíngue (principalmente inglês)
|
| 108 |
+
|
| 109 |
+
### Configurações
|
| 110 |
+
|
| 111 |
+
| Config | Vídeos | Com imagem de entrada |
|
| 112 |
+
|---|--:|--:|
|
| 113 |
+
| `veo3` | 1,333 | 309 |
|
| 114 |
+
| `veo3-audio` | 1,081 | 355 |
|
| 115 |
+
| `veo3-audio-off` | 416 | 0 |
|
| 116 |
+
| `veo3-fast` | 1,342 | 287 |
|
| 117 |
+
| `veo3-fast-audio` | 1,223 | 403 |
|
| 118 |
+
| `veo3-fast-audio-off` | 416 | 0 |
|
| 119 |
+
|
| 120 |
+
### Esquema
|
| 121 |
+
|
| 122 |
+
Cada configuração possui um arquivo `metadata.jsonl` junto aos arquivos de vídeo:
|
| 123 |
+
|
| 124 |
+
```json
|
| 125 |
+
{"file_name":"videos/veo3/train/00000001.mp4","prompt":"A cinematic...","model":"veo3","input_image_file_name":"images/veo3/train/00000001_input.png"}
|
| 126 |
+
```
|
| 127 |
+
|
| 128 |
+
- **file_name** — caminho relativo do arquivo de vídeo.
|
| 129 |
+
- **prompt** — o prompt de texto usado na geração.
|
| 130 |
+
- **model** — variante do modelo Veo 3.
|
| 131 |
+
- **input_image_file_name** — caminho da imagem de referência (null para text-to-video).
|
| 132 |
+
|
| 133 |
+
### Uso
|
| 134 |
+
|
| 135 |
+
```python
|
| 136 |
+
from datasets import load_dataset
|
| 137 |
+
|
| 138 |
+
# Carregar uma configuração específica
|
| 139 |
+
dataset = load_dataset("artificialguybr/veo3-video-prompts", "veo3-fast")
|
| 140 |
+
```
|
| 141 |
+
|
| 142 |
+
### Proveniência e direitos
|
| 143 |
+
|
| 144 |
+
Os vídeos foram gerados usando os modelos Veo 3 do Google a partir de prompts enviados por usuários em uma comunidade online. Identificadores de conta, identificadores de mensagem, timestamps e referências específicas à fonte foram removidos dos metadados. Os Termos Adicionais do Gemini API do Google afirmam que o Google não reivindica propriedade do conteúdo gerado, mas isso não garante protegibilidade por direitos autorais ou ausência de direitos de terceiros. Alguns prompts referenciam personagens protegidos, marcas ou figuras públicas. O dataset usa `license: other`. Cada usuário é responsável por avaliar os direitos aplicáveis e o uso pretendido. Pedidos de remoção podem ser enviados ao mantenedor pelo perfil do Hugging Face.
|
videos/veo3-audio-off/train/metadata.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
videos/veo3-audio/train/metadata.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
videos/veo3-fast-audio-off/train/metadata.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
videos/veo3-fast-audio/train/metadata.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
videos/veo3-fast/train/metadata.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
videos/veo3/train/metadata.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|