Image extend
A
Abundant Possum
Given an image as input, allow us to choose to create a new image based on it, with the original content, but with a new aspect ratio (or resolution), and fill in the blank region with AI-generated content that would extend what we gave as input.
This is similar to various other websites that offer it, such as this one:
As an example, suppose I want to create a landscape image out of this image that is in portrait orientation:
Log In
Olga
Thank you for the detailed feedback! We’re constantly improving both the service and the model, and insights like yours help us focus on what matters most to users.
For the best results, it often helps to try a few variations - tweaking the prompt, adding more detail, or rerunning the generation. Neural networks sometimes need those extra attempts to dial in exactly what you’re looking for.
Olga
Hi Liran,
Thanks for your feedback!
We have the 'Frame' tool and 'Variate' features that can help you achieve your goal.
To use frame click on the 'Frame' tool, create a frame, place your image inside it, and then write your prompt. Then click on "Recraft frame" to get the enhanced image, to which you can also apply the 'Variate' feature afterward.
A
Abundant Possum
Olga
- That's not the officially, known name for this operation.
- It seems it has put the original image on the side instead of letting us choose where and how to put it. If I want 9:16 aspect ratio image to become 16:9 aspect ratio, for example, isn't there a way to center-fit it automatically? I need to drag things manually hoping I do it right?
- The "Frame" doesn't let me choose an image for some reason. How exactly did you put the image? There is no button to do it.
- At first, I tried dragging an image file from outside the web browser says I'm out of times to do it. That's despite the fact I have 88 credits... Can you please change the credits system to let us use the website in all places, and also when it reaches 0, to just put us in low priority like some websites do?
- The "recraft" seems very buggy: The skin color of the hand of the person on the right doesn't match the skin color of the head, and it holds an unknown stick object for some reason that doesn't match being in a straight line... Even the hair color doesn't match... And, when I tried it out, it changed the image completely. I've provided it with 2 cats with their tails downwards, and on one image it provided 3 cats, and on another it changed the cats to have the tails upwards... There is no trace to the image I've provided.
So, in short, this feature of outpainting doesn't exist on the website. Please provide it.
Olga
Abundant Possum:
I’m afraid I didn’t quite understand your first two questions. Could you please clarify what you’re trying to achieve?
- Simply drag the image into the frame to position it.
- With the free plan, you can upload up to 3 images. After reaching this limit, you'll be able to upload new images again 24 hours after uploading your first image.
- Image generation relies on neural networks, and while we aim to deliver high-quality results, AI-generated outputs may sometimes require adjustments to align with your expectations. This is a natural limitation of the technology rather than an issue with the platform. Experimenting with prompt variations can often help achieve the desired outcome.
To assist you better, please share a link to your project (at help@recraft.ai). Also, let us know in more detail what result you’re aiming for.
A
Abundant Possum
Olga It's not intuitive at all. I wanted to just extend an image, meaning to add empty space to it that will be filled using AI.
The AI should try its best to fill it with logical things that could be there.
For example, if my image has sky at the top, and I tell it to add space there, it should extend to have what's possible to see in the sky: same color of the sky, and maybe some clouds.
Same goes for if the bottom has grass and I expand the image there.
Dragging the image manually to move it should be a feature, but not something I have to do, as I might move things too much, and I prefer to also have buttons to adjust it automatically (center/top/bottom/left/right, including fitting of course).
Your screenshot showed it doesn't provide a good understanding of what's possible to add (wrong skin color, wrong hair color, illogical object in the hand).
A prompt should also be optional, but without it, it should handle things automatically in a way that makes sense.
Does your website offer such a thing, for free?