Thelsim did a couple of Tarot-style cards a while back. I also just finally got Flux set up in ComfyUI – had started a long time back, and dropped it.

Flux is a ComfyUI model that’s pretty popular over on Reddit, both for the quality and because it uses English-style prompts rather than just a list of comma-separated prompt terms. I remembered Thelsim’s project, wanted to see if I could turn out a full set of photographic-style Major Arcana in the first day using it. Turns out…yes! Usually when running Stable Diffusion, I’ll generate maybe 20 images and pick the best, but this typically had something reasonable on the first try. It’s certainly not flawless – there are quirks in the image, but for anyone else thinking about playing with Flux, I wanted to put this out there, because I was unexpectedly happy with it, especially given that I’ve no experience at all with it. I would totally try and get it set up if you have a local generation setup!

Text was added with a script and ImageMagick, not in ComfyUI.

To get some kind of consistent appearance, I appended to each prompt “The theme is magical fantasy horror. The colors are blue, white, red, orange, and black. The photograph was taken with a Nikon D850.” I also used “Photograph…at night” on each.

https://lemmy.today/pictrs/image/67cd880f-ed36-4074-b12b-3e509b0dafa2.png

Photograph of the Grim Reaper at night in a dark, gloomy field. The Grim Reaper is riding a white horse. The Grim Reaper is holding a simple black scythe. The Grim Reaper’s hood only contains blackness. The sky is full of stars. The Grim Reaper is
wearing black gloves. The Grim Reaper is facing the camera.

https://lemmy.today/pictrs/image/2f0c296e-da19-4f31-8f35-cac07ce372dc.png

An photograph of a huge angel in the clouds playing a medieval trumpet at night. The angel is blowing into the trumpet. The angel is in profile. The zombies are climbing out of their graves in a graveyard. The dead are rising. There are snowy
mountains in the background.

https://lemmy.today/pictrs/image/b99ac2e5-a285-45f9-abc8-4267ae10398a.png

Photograph of a stern-looking young woman wearing a white blindfold and a toga sitting on a throne at night. The woman’s right hand is holding a set of scales aloft. There is a longsword lying by the woman’s feet. The woman is facing the scales.

Should really have a sword in one hand, scales in the other, but I wasn’t able to quickly get that working; probably need more experience with Flux.

https://lemmy.today/pictrs/image/3ed8cbe7-e97b-4f67-849e-19bd798caaae.png

Photograph of an angel at night. The angel is pouring glowing liquid from one large goblet in their left hand into a goblet in their right hand. The angel has a halo.

https://lemmy.today/pictrs/image/18a9c09b-dcc1-46da-87d8-a0c09c851170.png

Photograph of a man wearing armor riding a Roman war chariot at night. The chariot is pulled by two galloping horses wearing barding. The horse on the left is white, and the horse on the right is black. The chariot is charging the camera. The
photograph is an action shot. The man is holding reins.

https://lemmy.today/pictrs/image/4558e24b-0b45-433f-8548-6bb1873bb2d1.png

Photograph of the Devil at night. The Devil is crouching on a pedestal. There are two nude demons sitting at the base of the pedestal. The demon in the lower-right quadrant of the photograph is male. The demon in the lower-left quadrant of the
photograph is female. The Devil is holding a flaming torch in his hand.

It did look like Flux understands directives relative to the portion of the image here (“quadrant”). I wasn’t able to get the same technique going with Justice, though.

https://lemmy.today/pictrs/image/80bd8ca8-047a-4e0c-9d61-28b331646bf1.png

Photograph of an emperor at night. The emperor is holding a scepter.

https://lemmy.today/pictrs/image/e7a1f5f6-c2a3-41e2-9b47-fd9c56c8fd31.png

Photograph of an empress at night. The empress is holding a scepter.

https://lemmy.today/pictrs/image/1a6c7729-a5e3-49b8-bc93-d7ed31924cd6.png

Photograph of a jester at night.

https://lemmy.today/pictrs/image/28d2ab72-b90e-469a-a599-e1cad5992bdc.png

Photograph of a man hung upside-down from a rope tied around his left ankle at night. The man’s hands are hanging limply. The man is wearing Renaissance clothing. The man is wearing boots.

The feet are a bit off; I didn’t spent too much time futzing with it. Flux wasn’t super-into having things upside-down, though it did ultimately do it.

https://lemmy.today/pictrs/image/f052e1d7-c7cf-4ba2-8f9a-50b762e25186.png

Photograph of an old man wearing a robe walking on a mountain trail at night. The man is holding a lantern aloft and a staff.

https://lemmy.today/pictrs/image/439ec9a7-7739-4924-bede-cc776f7be8da.png

Photograph of a pope at night.

https://lemmy.today/pictrs/image/0be8647f-cb20-4be0-92c5-e266a4edca00.png

Photograph of a high priestess at night.

https://lemmy.today/pictrs/image/160c4575-c02a-4ccb-b513-6c60043d5b2f.png

Photograph of two lovers at night. The lovers are wearing Renaissance clothing. There are many fireflies.

https://lemmy.today/pictrs/image/a0e754ca-525e-4674-8b0a-bec48748e7f0.png

Photograph of a magician at night.

https://lemmy.today/pictrs/image/553fd519-ea29-4dba-9ed3-d0bf634c25fa.png

An photograph of two standing stones by a river at night. The moon is in the sky. In the lower-right quadrant of the photograph, there is a white wolf howling at the moon. In the lower-left quadrant of the photograph, there is a black dog howling at the moon.

I omitted the traditional crawfish. I didn’t really like the look of it, and on top of that, Flux kept wanting to make it look glowy, which I didn’t want.

https://lemmy.today/pictrs/image/687826be-81ba-452b-8378-b81f58e9bfce.png

An photograph of a naked woman at night crouching by a lake. The woman is facing away from the camera. The woman is holding a jug and pouring water into the lake. There is a bright star in the sky. There is an eight-point lens flare coming from the
bright star. The sky is black. The photograph is NSFW.

https://lemmy.today/pictrs/image/a2d42d98-2920-4eed-b5ab-5b807736d3f5.png

An photograph of a full solar eclipse with a visible solar corona. The Sun is black. The photograph is at night. A naked nude infant rides a white horse at night, with sunflowers in the background at night. The photograph is NSFW.

https://lemmy.today/pictrs/image/43709bad-1a8a-4680-8de9-5cf58005bb5e.png

Photograph of a tower on a hill at night.

https://lemmy.today/pictrs/image/9d308020-a83f-4b61-ad58-f47654e41ddf.png

A photograph of a glowing figure eight in the sky at night. The background is sky and clouds. A flying, nude woman in the clouds holding a wood baton in each hand is in front of the figure eight. The photograph is NSFW. The woman is nude.

I didn’t really like the traditional The World tarot card style, and it didn’t mesh well with a photographic style with all the disembodied heads, so I mashed up the oroborous and flying woman with batons from two different The World styles. Also, Flux was okay with up to three heads of various species sticking in at each corner, but for some reason was resistant to doing all four. I didn’t want to bang on it more. Flux was determined to put some clothing on the woman.

https://lemmy.today/pictrs/image/39b791c9-4493-43dc-bf86-4acc8daa1785.png

Photograph of a circle floating in the clouds at night. The circle is labeled with alchemical symbols. There are esoteric symbols covering the photograph. The circle is centered in the photograph.

There are normally some nude figures in a Tarot deck and I included this here; I didn’t flag the post NSFW as I don’t think that it’s all that explicit.

  • tal@lemmy.todayOP
    link
    fedilink
    English
    arrow-up
    2
    ·
    edit-2
    3 months ago

    I was pretty impressed with Flux, plan to use it more. For a Tarot deck, which has a bunch of nude figures, it was pretty determined to clothe them; I eventually just left the woman in The World wearing something. Describing the image as “NSFW” helped; I’m sure that people have their own techniques that I just don’t know about.

    I’m used to being able to use regional prompting in Stable Diffusion to stick specific things at specific places in the image. I don’t know yet if there’s a regional prompting analog compatible with Flux; the Stable Diffusion and Flux workflows are (unexpectedly to me) quite different in ComfyUI. Flux does understand some level of English-like description of the layout of the image, which is cool, but I wasn’t always able to get the output I wanted with that, so I expect that there’s still more digging.