OpenAI Updates DALL·E 3 To Edit Images with Prompts

Updated on April 6 2024
image

OpenAI, the Microsoft backed Artificial Intelligence Org, announced extremely user friendly upgrades to DALL·E 3, there text -to-image AI. When using DALL-E in ChatGPT on the web or mobile, users will now have access to built-in image editing tools. This new feature aims to simplify the creative process by allowing users to directly edit generated images within the familiar ChatGPT interface.

Let’s get into more specific of these DALL·E 3 updates.

DALL·E 3: Whats New?

New features in dall-e 3
New features in dall-e 3

The recent update of OpenAI’s DALL·E 3 introduces significant enhancements to the image generation and editing capabilities, making it more accessible and user-friendly for both creators and consumers. This update builds upon the previous versions by integrating DALL·E 3 with ChatGPT, allowing users to generate detailed prompts for images with ease and refine their creations without leaving the chat interface.

ChatGPT Integration

DALL-E 3 now integrates seamlessly with ChatGPT. Users just need to describe their desired image in a natural language based on ChatGPT and DALL-E 3 itself will generate personalized prompts. This not only simplify the work but also leverage ChatGPT’s nuances and details to create more detailed and expressive images.

Editor Interface

The creation of a new interface for editors will result in the improvement of their AI-powered images. The tool is equipped both with pixel-wise adjustments of individual image spots and for general composition corrections. Users can pinpoint some areas and type text messages which will erase, add or refine those features accordingly. This level of fine detailed control, therefore, gives users the perfect opportunity to accomplish their exact artistic presumptions.

Preset Style Suggestions

For novice users, DALL-E 3 now offers a selection of preset style suggestions. This library includes options like woodcut, gothic, synthwave, and hand-drawn, providing a variety of artistic starting points for image creation. This feature makes DALL-E 3 more accessible by offering pre-configured stylistic options to spark user creativity.

Addressing Trustworthiness

OpenAI recognizes the importance of trust in AI-generated content. To address concerns about image authenticity, DALL-E 3 now incorporates visible watermarks and metadata to identify AI-generated images.

Read More: OpenAI Watermarks Images Created by Dall-E 3

Fine-Tuning API Improvements

The fine-tuning API, launched for GPT-3.5 in August 2023, has seen a series of improvements aimed at giving developers greater control over their fine-tuning jobs. These enhancements include:

Epoch-based Checkpoint Creation: This functionality trains the full checkpoint of fine-tuning model while one epoch passes, which eliminates the need for subsequent retraining of the different epochs and reduces overfitting.

Comparative Playground: A newly designed interface to have previews where a person can now do multiple models comparing or fine-tune entire scenarios to a single statement.

Third-party Integration: Integrations for popular platforms suit for weighty task (i.e. Weights and Biases), will be supported, and the development team will be able to share complex training data with the existing tech stack they use.

Comprehensive Validation Metrics: The advantage to evaluate metrics such as loss and accuracy once across the whole dataset, while providing an overview of model performance across a larger sample.

Hyper-parameter Configuration: The developers can now configure available hyper-parameters directly from the dashboard without need of basic coding while still getting the job done right, which is especially suitable for the users.

Fine-Tuning Dashboard Improvements: The Dashboard improvements include an option for the user to configure hyper-parameters, view more assessed training metrics as well as the return of the past jobs based on already defined configurations, which, generally, improves the quality of the training process.

DALL·E 3 and Custom Models Program

The latest model – OpenAI’s DALL·E 3 – offers better image detail, more complex scenes, and overall better capability to turn the text into the image. Additionally, it upgrades the aspect ratio to be more flexible and comes with a prompt rewriting built in to make the images better than before while reducing bias and introducing more natural variation. The process is scheduled to become available in OpenAI Studio and via the REST API.

The Custom Models Program, revealed at DevDay at the end of November, has been enhanced with an option of assisted fine-tuning. This provided an opportunity to jointly work with OpenAI’s experts who implemented techniques beyond the fine tuning API that entail use of additional hyper-parameters and parameter efficient fine-tuning (PEFT) at a larger scale. This is especially useful for enterprises that are requiring the best in training data pipelines, evaluation systems and the development of a model that is to their specification targets and use cases.

Also Read: OpenAI Secretly released GPT 4.5 – Fake or Real?

Conclusion

To access the DALL-E and its new updates, you will need to be a ChatGPT Plus subscriber. The subscription costs $20 per month. Additionally, with DALL-E 3, you have the added benefit of generating images within the ChatGPT interface, allowing you to seamlessly switch between text and image prompts without exiting the platform. However, if you prefer not to spend money, you can use DALL·E 3 in Microsoft Designer at no cost.

As OpenAI continues to innovate, the potential for further customization and optimization of AI models, including DALL·E 3, remains vast, offering exciting possibilities for the future of AI development.

Featured Tools

creatify-logo

Creatify

Humanize AI

Air Chat

Beatoven

Notably

Spice

Unhinged

Thumbmachine

Related Articles