OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Text To Image

Text To Image is an experimental Python and TensorFlow implementation that converts captions into Skip-Thought Vectors and synthesizes images with the GAN-CLS algorithm. Built on a TensorFlow DCGAN implementation, it includes documented workflows for processing flowers data, training the model and generating images from captions.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Text To Image — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you want to examine or run the repository’s experimental flowers-focused caption-to-image workflow, rather than use a hosted generation tool.

Who it is for

  • Researchers and developers exploring caption-conditioned image generation.
  • Python users who want to inspect or adapt a GAN-CLS and Skip-Thought Vector implementation.

Documented uses

  • Process flower captions into Skip-Thought Vectors for model training.
  • Train or resume the caption-conditioned image model.
  • Generate image samples from captions using a trained or downloaded model.
Before you choose

Requirements & limits

The repository is public and licensed under MIT. The documented workflow is code-based and lists Python 2.7.6 and TensorFlow among its requirements.

  • The README identifies the implementation as experimental and the flowers dataset as its current training dataset.
  • Generation depends on caption vectors and a trained or downloaded image model; the repository does not describe a one-step hosted service.
  • The documented output size is 64 × 64 pixels.
  • The stated core stack includes Python 2.7.6, TensorFlow and h5py, with Theano, scikit-learn and NLTK for Skip-Thought Vectors.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Generate images from captions

Prepare caption vectors and generate images using a trained or downloaded model.

Before you start

  • Python 2.7.6, TensorFlow and h5py; Theano, scikit-learn and NLTK for Skip-Thought Vectors.
  • A caption text file, Skip-Thought model and vocabulary files, and a trained or downloaded image model.

Steps

  1. Save captions in Data/sample_captions.txt.

  2. Run the command below to create their thought vectors.

    Documented command
    python generate_thought_vectors.py --caption_file="Data/sample_captions.txt"

  3. Run the command below, substituting the model path.

    Documented command
    python generate_images.py --model_path= --n_images=8

Expected outcome

Generated images are saved in `Data/val_samples/`; `n_images` sets the number generated per caption.

Guide limitations

  • The repository describes this as experimental and reports 64 × 64 generated images.
  • The example generation workflow requires the appropriate model and caption-vector resources; no execution or compatibility guarantee is implied.
Documentation 1 ↗
Source-based practical guide

Prepare and train the flowers model

Process the flowers captions into Skip-Thought vectors and start model training.

Before you start

  • Python 2.7.6, TensorFlow, h5py, Theano, scikit-learn and NLTK.
  • Flowers images and captions, Skip-Thought pretrained models and vocabulary, and the required Data directories.

Steps

  1. Place flower images in Data/flowers/jpg and the extracted text_c10 captions folder in Data/flowers.

  2. Save Skip-Thought model and vocabulary files in Data/skipthoughts; create Data/samples, Data/val_samples and Data/Models.

  3. Run the command below to extract vectors.

    Documented command
    python data_loader.py --data_set="flowers"

  4. Start training with the command below.

    Documented command
    python train.py --data_set="flowers"

Expected outcome

The documented workflow processes flowers captions and starts training the GAN-CLS implementation on the flowers dataset.

Guide limitations

  • The README says downloading the datasets and models can involve several gigabytes.
  • The author reports a training run of around 200 epochs taking roughly two to three days on a GPU; this is not a runtime guarantee.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
paarthneekhara/text-to-image · Official homepage title · retained in its source language
How we review sourcesSuggest a correction