If SOLO can be used to generate fully-functional websites, does it mean it can also create the image assets for the site - eg., images for hero banners, placeholder testimonial headshots, blog article illustrations, etc.?
If not, how are Trae developers doing it?
Is there a recommended 'workflow' for generating images with Kilo CLI?
I've setup Kilo and am using Kilo CLI on a Mac. I've subscribed to KiloPass, so I believe I do have access to frontier models that can generate images. But without a dedicated 'chat app' - like ChatGPT has - that allows the agent to show me the image and get feedback, what is the workflow for asking Kilo to generate an image, viewing the image and getting Kilo to make changes? Is there a "recommended workflow"?
I thought it would be super easy to get ChatGPT to make me a reasonably good looking flowchart, that I could simply download from the chat to use elsewhere. But apparently it isn't.
It tried to generate the flowchart using mermaid. It was able to preview the mermaid diagram within the chat, which looked perfect. But it was unable to provide it to me as an image: it wanted me to take the mermaid code to the mermaid live editor website, and download an image from there. Sure, I could do it, but that's not what I asked: I wanted it to be able to generate a downloadable image for me, directly.
So it tried generating an SVG. Multiple times. It looked nothing like the mermaid diagram. Flow lines were wrong. After multiple attempts, it finally gave me a usable SVG file, but still lots of things needed to be corrected: line weights, text alignment (not centred), and more. Still not a 'download and use' solution.
Finally, I tried to ask it to just generate a bitmap image of the flowchart — a PNG or JPEG would do. It screwed up the formatting even more. Text that didn't fit into the node boxes, arrows pointing in the wrong direction, etc...
Is it at all possible for ChatGPT to generate a simple, usable flowchart (image)? Or do I have to fork out subscription money to an external tool for that, too?
I wanted to create an AI voice for an AI singer I’m trying to develop. I didn’t want the voice to be MY voice. Currently, it seems the only way to do that is to ‘extract’ the voice from a song - i.e., create the voice based off a song. I did that, but now every time I create a new song with that voice, very recognisable melody elements of the original song repeat in the new one - to the point that it’s incredibly difficult to create a whole new song that doesn’t “remind you” of the original.
I tried using a totally different song as “inspiration”, and also remastering it as a “cover”: if the “audio style” is set too low, I end up with something quite different melodically from the base song, and if it’s set too high then the melody elements from the voice sample get introduced…
My plan doesn’t give me access to Studio, so it took me well over 50 generations - including multiple regenerations of sections on the song in the Editor - to get a usable 2nd track with the same AI voice… Am I missing something basic? Or is this something that is supposed to be done in Studio only?