Most AI tools are built around a chat box: you type a request, and the answer lands in a thread. Hylope is built as a workspace instead. Your inputs, your results, your gallery and your settings stay on screen together, and images move between them by drag and drop.
Two ideas hold it together. Every mode, from image and video to music and voice, uses the same layout, so once you know your way around Image, you know it everywhere. And nothing is lost: every generation keeps its model, prompt and settings, so you can try things freely and come back to any of them.
This first video of the Getting Started series shows that workspace, and how generation works in it, through your first images.
What you will build
- A first image, then the same prompt on four models at once
- Stacks sorted by status, with the takes you like in front
- One prompt that makes an image per city, with variables
- A snippet saved from your own prompt
- An image taken through Style Transfer to a video, with its lineage
One layout for every mode
The screen is organized around your result. It sits in the middle, with the prompt bar under it, the settings of the current mode on the right, and your recent generations and the gallery along the bottom. Switch from Image to Video or Upscale and only the content changes: the models on offer, the inputs the mode accepts, the settings it needs.
Images move by drag and drop. Drop one from the gallery, the timeline or the Shelf onto an input to use it, or drop several, or a whole stack, to load them all at once. To continue in another mode, use the buttons under a result.
In the video, we switch from Image to Video and back, to show how little changes.
Try freely, nothing is lost
Generating is trial and error, so every attempt is easy to undo. Change the prompt or the model and the result clears, because that is a new generation. Remove the change, or click the previous card in the timeline, and the earlier result comes back.
The same goes for anything you made. Reload loads an image’s model, prompt and settings again, from the timeline, the gallery, the lightbox or a right click. Improve prompt rewrites your prompt with the best practices of the current mode, and Revert gives you your own text back.
You don’t have to wait for a result either. Generations keep running while you switch modes or apps, and clicking Generate again queues more, up to the limit of your plan.
Compare, then sort
One image tells you little. We run the same prompt on Nano Banana 1.0 Pro, Muse Image 1.0, Grok Imagine 2.0 and Recraft 4.1 Utility at once, from the batch icon on the model card, to see how each model reads it.
Variables do the same with an idea. List a few values, like Paris, Berlin, Rio and Roma, type @ in the prompt to insert the variable, and each generation uses one of them.
All of this fills the gallery quickly. Generations that share a prompt are grouped into a stack. Give the takes a status, Good, Maybe or Rejected, and when you close the stack it splits by status, so the ones you like stay in front. Rejected images leave the stack but are not deleted: the Hidden filter shows them again.
To keep a few assets at hand, drop them on the Shelf at the top of the screen. It stays with the project wherever you go in the app.
From a prompt to a video
Snippets keep the parts of a prompt you want to use again. In the video, we save a description of rainy, overcast light that we like.
Select the text
Select the sentence in the prompt and click Create snippet. Give it a title and a category, here Style.
Insert it
Type @ in any prompt, or the first letters of a category, and pick the snippet.
Change it
Click the snippet to edit it, and save a new version. The earlier versions stay in its history.
Then we take the image further. In Image Edit, Style Transfer applies a style preset with no prompt at all. From the result, Video switches to Video with the image loaded as the start frame, ready to animate.
Open the video and click Lineage to see the whole path: the prompt and its snippets, the edit, then the video.