In comic art there are some artists who use a lot of photo references, to the extent that a panel looks like a collage of photos that they traced and added stuff to. Greg Land is a famous example. Their work tends to have an Uncanny valley look. They do work faster than traditional artists though.
I suspect that AI is not at a point where it can draw a comic by itself, but it can generate images that an artist can combine and trace over to form a complete comic. This might be able to speed up the work of artists considerably, while eliminating much of the Uncanny valley traced photo look.
Sounds fun and interesting; I for one would be really interested in how much one can currently push AI art to create a complete work, and not just one-off pictures. Many thanks to Steve for funding it!
As an aside, this comes at a slightly awkward time as the new much-improved version of Stable Diffusion (which one pretty much has to use to craft a work of comic-length consistency) will release this month.
After that is released, I expect there to be some weeks before all the tooling (ControlNets, checkpoints...) are updated to support the new model. This means there might be a really short window to submit to the first stage of the contest if you want "current gen" quality!
Yeah, the final SDXL checkpoint is supposed to be released in a week or so, IIRC, and then everything has to be updated, finetunes trained for weeks, LoRAs created... Also, the Imagen API only just opened up, so that might be viable but will take time to roll out to more people & learn to use. 30 days from today seems premature - what's the rush? It's been sitting around since 2005/2020.
A real shame that Aljion Dave resigned. The 36 pages he did look great! I'd probably have paid some money into a Kickstarter to stuff his mouth with gold to finish it! Look forward to seeing the winning entries to the contest.
Thanks for doing this! I have a novella I’d like to turn into a graphic novel with generative AI, so I’ll be excited to see what the outcome of this is.
In comic art there are some artists who use a lot of photo references, to the extent that a panel looks like a collage of photos that they traced and added stuff to. Greg Land is a famous example. Their work tends to have an Uncanny valley look. They do work faster than traditional artists though.
I suspect that AI is not at a point where it can draw a comic by itself, but it can generate images that an artist can combine and trace over to form a complete comic. This might be able to speed up the work of artists considerably, while eliminating much of the Uncanny valley traced photo look.
Sounds fun and interesting; I for one would be really interested in how much one can currently push AI art to create a complete work, and not just one-off pictures. Many thanks to Steve for funding it!
As an aside, this comes at a slightly awkward time as the new much-improved version of Stable Diffusion (which one pretty much has to use to craft a work of comic-length consistency) will release this month.
After that is released, I expect there to be some weeks before all the tooling (ControlNets, checkpoints...) are updated to support the new model. This means there might be a really short window to submit to the first stage of the contest if you want "current gen" quality!
Yeah, the final SDXL checkpoint is supposed to be released in a week or so, IIRC, and then everything has to be updated, finetunes trained for weeks, LoRAs created... Also, the Imagen API only just opened up, so that might be viable but will take time to roll out to more people & learn to use. 30 days from today seems premature - what's the rush? It's been sitting around since 2005/2020.
Were the submissions / details / winners of stages 1 and 2 posted anywhere?
A real shame that Aljion Dave resigned. The 36 pages he did look great! I'd probably have paid some money into a Kickstarter to stuff his mouth with gold to finish it! Look forward to seeing the winning entries to the contest.
Good luck
Thanks for doing this! I have a novella I’d like to turn into a graphic novel with generative AI, so I’ll be excited to see what the outcome of this is.