Anyone have success generating more photorealistic images that don't have the slightly warped look without training their own models? If so, how are you generating the prompts, and what kind of service are you using for that? We tried DALL-E, but we want to focus on more realistic rather than artistic images
Our use case is generating images to add to custom work instructions content for the manufacturing industry
05/03/2023, 7:22 AM
I used the AUTOMATIC1111 gui for installing stable diffusion locally on my macbook, here:
you can find example prompts on civitai.com to get a feeling for writing prompts, the first site I linked also has a guide
you can download lots of other models on huggingface for stable diffusion, and you can also use the AUTOMATIC1111 gui interface to hybridize models.
there are models that try to closely emulate midjourney, for example.
05/04/2023, 4:28 PM
Thanks we'll check those out. We'd ideally like to be able to automatically have GPT generate prompts from the work instructions text, and right now, we can't guarantee that it'll be what we expect. Images generated using prompts may still still warp-y and more artistic, so we'll be testing out more models
05/04/2023, 4:30 PM
Sounds very doable based on my limited toying around with prompt engineering
05/04/2023, 8:33 PM
Midjourney has been giving the least warped images so far with different prompt experiments but they don't have an API afaik - only the Discord bot