imagesjackyhate/text-to-image-2Mtext-to-imagemultimodalimage-generationfine-tuningdiffusionwebdatasetcaptionscomputer-vision

Text-to-Image Curated Pairs

Name: Text-to-Image Curated Pairs
License: https://huggingface.co/datasets/jackyhate/text-to-image-2M

Free

Open dataset

Sample structure: 100 / 100

3 download links issued

Seller: DataBazaar

Already have an account? Log in

Agent? Connect your account →

Source, license and coverage

Supplier documentation. These claims are separate from the automated sample score. A listing edit date is not a data freshness date.

License: mit
Source / creator: jackyhate/text-to-image-2M
Collection method: Per the publisher, samples were curated and enhanced from existing text-image sources to address quality issues in raw web-scraped T2I datasets. The selection prioritizes high-quality, diverse text-image pairs suitable for fine-tuning. Specific upstream source composition and filtering heuristics are documented on the HuggingFace dataset card.
Coverage start: Not documented
Coverage end: Not documented
Data last updated: Not documented
Update schedule: Not documented

Source documentation ↗

License terms ↗

Preview contains complete records from a bounded prefix of the previous sample; it is not a random sample of the full dataset. Embedded binary payloads are explicitly replaced with byte-length descriptors; the preview preserves accompanying text and metadata. Captions are English-only, limiting multilingual T2I training utility. Upstream source provenance is not fully enumerated per-sample, so deduplication against specific eval benchmarks (e.g., COCO, DrawBench, PartiPrompts) is not guaranteed. The publisher does not document demographic, geographic, or content-category sampling biases; buyers should validate empirically for their use case. Image licensing of the underlying photographs (as distinct from the dataset compilation license) is not individually verified.

Sample structure score: 100 / 100

This automated check describes the inspected sample, not factual accuracy, legal rights, representativeness, or the quality of the entire dataset. It is not a customer rating.

Assessed 5 sample records (JSON) on 2026-10-09. All records in the provided sample were checked.

Check	Points	Evidence
Populated cells	50 / 50	20 of 20 top-level cells contain a value. Null, absent and blank values count as missing; zero and false count as populated.
Consistent value types	30 / 30	20 of 20 populated cells match their column's most common observed type. Types are inferred, not checked against real-world truth.
Consistent record shape	20 / 20	5 of 5 records have the expected fields. CSV/TSV use the header width; JSON uses the union of observed keys.

Field-level findings and improvements

Check missing cells and mixed types below. Document intentional missing values or mixed types in your field descriptions. Do not fill legitimate unknowns with invented values just to increase this score.

Field	Missing cells	Most common type	Other populated types
__key__	0 / 5	string	0 / 5
__url__	0 / 5	string	0 / 5
jpg	0 / 5	object	0 / 5
json	0 / 5	object	0 / 5

How the score is calculated, its limitations, and how to correct an assessment →

About this data

Text-image pairs in webdataset format designed for fine-tuning text-to-image generative models.

Retrieve with your agent or Python

Create an account and configure DATABAZAAR_API_KEY. This example retrieves free or already purchased data; it never makes a purchase. For a multi-file dataset, choose a file index from the manifest.

Download the Python example

python3 retrieve-dataset.py 25fdb009-f87e-4977-b5d2-698a5119407f --output dataset.bin

Full supplier documentation

## Overview text-to-image-2M is a curated dataset of approximately 2 million text-image pairs designed for fine-tuning text-to-image generative models. Distributed in webdataset (tar shard) format with image modalities paired with English text captions. The dataset was assembled and enhanced specifically to meet the scale and quality demands of modern T2I model training, where >1M sample sizes have been observed to produce noticeably better results. ## Schema - image — binary (jpg/png) — the image sample - text / caption — string — English-language caption describing the image - key — string — webdataset shard sample key - (webdataset format; exact sidecar fields depend on shard contents) ## Sources - HuggingFace: https://huggingface.co/datasets/jackyhate/text-to-image-2M — License: MIT - DOI: 10.57967/hf/3066 ## Methodology Per the publisher, samples were curated and enhanced from existing text-image sources to address quality issues in raw web-scraped T2I datasets. The selection prioritizes high-quality, diverse text-image pairs suitable for fine-tuning. Specific upstream source composition and filtering heuristics are documented on the HuggingFace dataset card. ## Known gaps & limitations Captions are English-only, limiting multilingual T2I training utility. Upstream source provenance is not fully enumerated per-sample, so deduplication against specific eval benchmarks (e.g., COCO, DrawBench, PartiPrompts) is not guaranteed. The publisher does not document demographic, geographic, or content-category sampling biases; buyers should validate empirically for their use case. Image licensing of the underlying photographs (as distinct from the dataset compilation license) is not individually verified. ## Intended use & out-of-scope - IS for: fine-tuning text-to-image diffusion/autoregressive models, T2I research, caption-conditioned image generation evals, image-to-text training. - NOT for: production deployment without independent image-rights review; benchmark training without dedup against held-out eval sets (leakage risk); non-English T2I tasks. _Federated dataset: 10 parquet shards, 4.69 GB total. Queries and downloads stream through the DataBazaar API._ Original supplier listing: Text-to-Image 2M: Curated Text-Image Pair Training Dataset ~2M curated high-quality text-image pairs in webdataset format, designed for fine-tuning text-to-image generative models. MIT licensed.

Schema

Name	Type	Description
__key__	VARCHAR	Unique identifier for the sample within webdataset shard (format: dataset_variant_shard_index).
__url__	VARCHAR	HuggingFace dataset repository URL and TAR shard file location.
jpg	STRUCT(bytes BLOB, path VARCHAR)	Binary JPEG image file bytes and relative path within the webdataset shard.
json	STRUCT(prompt VARCHAR)	English-language text caption describing the image content.

Sample Data

Preview a sample of the data before downloading.

Public sample only. Sign in to retrieve the full dataset, including free datasets.

For AI Agents

Via MCP Server

# 1. Add to your agent's MCP config (claude_desktop_config.json or similar):
{
  "mcpServers": {
    "databazaar": { "command": "npx", "args": ["databazaar-mcp"] }
  }
}

# 2. Your agent can then call:
search_datasets({ query: "Text-to-Image Curated Pairs" })
// Found: 25fdb009-f87e-4977-b5d2-698a5119407f
get_download_url({ dataset_id: "25fdb009-f87e-4977-b5d2-698a5119407f" })  // free — sign in with MCP OAuth first

Via REST API

# Free dataset — sign in or use your account API key:
curl https://api.databazaar.io/datasets/25fdb009-f87e-4977-b5d2-698a5119407f/download-url -H "Authorization: Bearer $DATABAZAAR_API_KEY"