Bing AI generator lets you create any kind of image you want, EDIT: Video AI generator here (Page 175 of 177)
|
Originally Posted By Klee: Added some Spanish Moss. I think i'm getting the hang of this AI stuff.  Twin boys fishing on the Mississippi River circa 1870. https://i.imgur.com/jb9ejhX.png Edit: Here is the prompt I created it with. "a american civil war era sepia-toned black and white photograph from 1862,wet plate collodion photograph, on the shore of the mississippi river looking towards the river, two male children sitting on a small wooden boxes fishing a medium distance away, paddle steamer in the far distance, medium depth of field, soft focus, silver halide grain, vignetted edges, faded tones, natural dappled sunlight filtering through ancient cypress trees, spanish moss hanging off the trees, atmospheric haze and morning mist, authentic 19th century photographic realism, grainy matte finish paper print texture, medium width white slightly undefined faded rounded edge borders, documentary style" You are getting it. Prompts are everything and every model uses different prompts. One thing to keep in mind when you can't get the changes you want, having AI create something is straight forward but having AI create something specific can be a royal pain. For example the boat it created looks like it has sailing masts but they would be way too short for that boat and there are no exhaust stacks or black plum of smoke coming from them. Try to make that specific change, replace the sailing masts with smoke stacks. Depending on the engine/model it might be easy or it might take hours. When it comes to reproducible person, none of the AI engines know how to manage freckles/moles. No matter the wording they will still shift and move. I'm getting closer but still not perfect with my prompts. I literally spent an entire day reworking prompts just to get freckles/moles to stay consistent. I'm running ComfyUI at home as well, trying to do more with it as I burn through all my other AI creates really quick. Hope you don't mind, I brought your picture to life. Still couldn't get the smokestacks I wanted. But that should be fixed in the reference image before turning it into a video. A great starting image is a requirement for a good video.  River Steamboat Boys Fishing - 1800s |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Originally Posted By SWIRE: You are getting it. Prompts are everything and every model uses different prompts. One thing to keep in mind when you can't get the changes you want, having AI create something is straight forward but having AI create something specific can be a royal pain. For example the boat it created looks like it has sailing masts but they would be way too short for that boat and there are no exhaust stacks or black plum of smoke coming from them. Try to make that specific change, replace the sailing masts with smoke stacks. Depending on the engine/model it might be easy or it might take hours. When it comes to reproducible person, none of the AI engines know how to manage freckles/moles. No matter the wording they will still shift and move. I'm getting closer but still not perfect with my prompts. I literally spent an entire day reworking prompts just to get freckles/moles to stay consistent. I'm running ComfyUI at home as well, trying to do more with it as I burn through all my other AI creates really quick. Hope you don't mind, I brought your picture to life. Still couldn't get the smokestacks I wanted. But that should be fixed in the reference image before turning it into a video. A great starting image is a requirement for a good video.  https://www.youtube.com/watch?v=87rjrRTzvgoThat's cool! I plan on eventually graduating to more complex stuff including videos. About the prompts, its a little tricky and sometimes the results are completely wrong. An example, I was tweaking the prompts and it started adding baseball caps that I did not specify and when I told it "without hats" it removed one but not the other. After I removed the "without hats" it stopped giving the two boys hats. Sometimes adding or taking away something in the prompts does nothing or it could do something unexpected. I probably go through three or four before I have one that I like. Also when I do hot women in bikinis on the beach it usually has the ribs showing and sometimes they just do not look correct, no i'm not going to make them fat just to hide them  , a little too low or maybe a couple extra.EDIT: Also I noticed if you do not add something like " looking at the camera seductively, cute, pouty, angry" ect, the face and eyes have a dead look. Dead eyes/expression. I guess its the "Thousand AI Stare"  that seems like she is looking through you.  More real looking. Definatly looking at you.   |
|
What workflow, nodes, and models are using in Comfy? Changing all those can have a big effect in on how it responds. Are you using both a Positive and Negative prompt feeding into your Apply or Guider? Having both is a must. Then strength and steps will be a factor along with whatever model you using in the Sampler. Most of my image generation is with either Grok or Firefly which is run by Adobe. I had a AI credits with the Creative Cloud subscription and never used them. Once I got my first lesson in AI I started burning through quick. Then they offered a promotion too good to pass up on. Their plan allows me to generated 4k images, which is crazy in the AI world. Grok I think is closer to a 1.5K image. For my videos 720p is about the best I can do with Comfy and my PC. I haven't tried pushing the limits with Image generation in Comfy yet because Adobe is going to be hands down better. But I'm playing around with it. Both the Facebook and YouTube algorithms absolutely hate the videos that Comfy just created for me, even though I think they are fairly close to what HeyGen can put out. There is something they are picking up on that I'm not seeing. Mine are 720p and HeyGen is 1080p but that alone shouldn't make much difference. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Finally got the Naughty children where they need to be Messing up the Monday Morning Assembly. Everybody is still giggling from time to time, recalling everything. If the adjusted lyrics weren't Rated-R they wouldn't be punished so severely. THE dullest and boring babysitters around while Mom and Dad have a Badly needed Night Out. Attached File |
"He should have killed me. I would have killed me."
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
|
Originally Posted By Klee: Looks like she got married without my permission, that was NOT in the prompt.....  https://i.imgur.com/CYE89Fs.png |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
| YouTube must have change their algorithm or policy in AI videos, could be punishment for not posting consistently as well...but my shorts went from 400-900 views to 0-4 views this whole last week. YouTube was never the target audience anyway but dropping views by that much is a kick in the balls. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Nice, I have videos like that in the works but that starts with static images. I'm trying to build multiple video layers to create a moving scene with objects in it moving at different speeds. I got the idea from another youtube channel. Spaceship, with space girls, moving through space, with objects outside the ship moving. It isn't terribly difficult but very time consuming. The hard part is what you think would be the simplest. Create the space girl with a solid green screen background . Instead it gives me some weird patched background which makes trying to turn it transparent a major pain, especially around and through the hair.  |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
This is all I have so far. The concept works. I created the binaural music to go with it. The ship, the planets, the girl...ect. I think it is garbage but it is a start. Space Girl Binaural Music Asteroid |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Hybrid Content, AI mixed with real is also a thing. It seems to be more accepted that an AI alone. Here is one hybrid video I made by putting my AI character into the flight suit and green screening it. Then chroma key remove the green screen and placed that video layer on top of my a video of my step-daughter flying.  iFly VA Beach - Glowing Review. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
I have confirmation that YouTube is going after slop AI content and stopped recommending a lot of Ai video. I even gave my character an updated studio and added music...about 5 views so far. My stuff far from slop but YouTube is still pushing back on it. I'm able to reproduce the character with 99% accuracy, the freckles and moles will shift. Outfit and hair change is easy. HeyGen is the main engine I use for the lip sync video but I've made and published videos I created in Grok and Comfyui. Grok will only create 6-10 seconds clips, then changes appearance and voice for each clip. But it is getting better and good enough to make a video from...but you will burn up a lot of your credits trying. Comfy is the way to go but it tries to cache/buffer the whole video in memory, it can handle about 1 second of video per GB of GPU ram before it starts to degrade the quality and morph the character. I have a workflow that is supposed to chunk the video creation process and create a 8-10 second clip, grab the last few frame of it and use it as a reference for the start of the next clip. I'll be playing with that next.  They only notice what you didn’t do… never what you actually did. Anyone else living this? ?? |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
 MiniMax H3 BF16 T2V_2 Minimax H3 BF16 text-to-video 530 second generation time Slightly higher resolution (which youtube nerfs, sadly)  MiniMax H3 BF16 T2V_3 |
And then, he arrives at the last item on his bucket list...
|
Originally Posted By GySgt_D: https://www.youtube.com/watch?v=UAbkqwCPGKgMinimax H3 BF16 text-to-video 530 second generation time I plan on trying Minimax H3 as my first try at text to video. |
|
Aria defends the Texarrakis BBQ from "changing consumer habits" |
And then, he arrives at the last item on his bucket list...
|
Originally Posted By GySgt_D: https://www.youtube.com/watch?v=UAbkqwCPGKgMinimax H3 BF16 text-to-video 530 second generation time Slightly higher resolution (which youtube nerfs, sadly) https://www.youtube.com/watch?v=vpiXRJhHI40Nice, what is H3? I've seen it mentioned a couple places recently but it isn't anything I've come across as far as using yet. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Originally Posted By SWIRE: YouTube really hates AI now. https://www.ar15.com/media/mediaFiles/680/AI-3808869.png Are you uploading the video as it was outputted by AI? If so maybe it's reading the embedded meta data? I know in ComfyUI you can drag and drop a previously generated image into a new install of comfui, Hermes told me that, and it can automatically recreate the workflow. I did that with one of the first images from a comfy install a couple of months before, I dragged and dropped it into comfy and it replicated the workflow and the original prompt from the info in the image. Maybe run the videos through a converter that strips all of the info. |
|
Originally Posted By GySgt_D: Aria defends the Texarrakis BBQ from "changing consumer habits" https://www.youtube.com/watch?v=lCKOjHc3_aYHer hair is Supposed to be Chestnut Brown.... |
"He should have killed me. I would have killed me."
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
Back to Mundania Tuesday Morning, four more morose children would be very difficult to find. Justin just waved 'Hello' to Lilly Ann. The only words The Principal said to them was, "People are still laughing about your stunt." Attached File |
"He should have killed me. I would have killed me."
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
|
Originally Posted By Klee: Are you uploading the video as it was outputted by AI? If so maybe it's reading the embedded meta data? I know in ComfyUI you can drag and drop a previously generated image into a new install of comfui, Hermes told me that, and it can automatically recreate the workflow. I did that with one of the first images from a comfy install a couple of months before, I dragged and dropped it into comfy and it replicated the workflow and the original prompt from the info in the image. Maybe run the videos through a converter that strips all of the info. Originally Posted By Klee: Originally Posted By SWIRE: YouTube really hates AI now. https://www.ar15.com/media/mediaFiles/680/AI-3808869.png Are you uploading the video as it was outputted by AI? If so maybe it's reading the embedded meta data? I know in ComfyUI you can drag and drop a previously generated image into a new install of comfui, Hermes told me that, and it can automatically recreate the workflow. I did that with one of the first images from a comfy install a couple of months before, I dragged and dropped it into comfy and it replicated the workflow and the original prompt from the info in the image. Maybe run the videos through a converter that strips all of the info. It's a wild time to be working with AI right now. One of the pro's I'm watching was doing a video covering the results of creating a new AI influencer for 30 days. He got them up to 45,000 followers and then Instagram completely nuked the account for being AI even though he tagged it as AI in their own system settings. Facebook doesn't seem to care or at least not as much. They are slow rolling my AI videos before then 12 hours later they start getting views. I'm going to keep posting to all three. Haven't gotten on to TikTok yet but learning that platform is something I should do. I'm tempted to run a promotion on one the Shorts that YouTube is sandbagging and see what happens. Maybe it is a "pay to play" scam. Since it is AI maybe they figure the people have money to spend on promotions. That would suck if that is the case. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Originally Posted By SWIRE: Their AI spotting algorithm is all over the place. 90% of the videos I uploaded it did not tag as AI even though most came from HeyGen. I've had others than I ran through Adobe to add a music track and it still came up as AI. But Adobe is one of the major players in pushing tracking AI. Apparently they are now embedding data into the video itself, not just metadata, so even saving it in a different program doesn't delete it. It's a wild time to be working with AI right now. One of the pro's I'm watching was doing a video covering the results of creating a new AI influencer for 30 days. He got them up to 45,000 followers and then Instagram completely nuked the account for being AI even though he tagged it as AI in their own system settings. Facebook doesn't seem to care or at least not as much. They are slow rolling my AI videos before then 12 hours later they start getting views. I'm going to keep posting to all three. Haven't gotten on to TikTok yet but learning that platform is something I should do. I'm tempted to run a promotion on one the Shorts that YouTube is sandbagging and see what happens. Maybe it is a "pay to play" scam. Since it is AI maybe they figure the people have money to spend on promotions. That would suck if that is the case. Do a experiment, play the video on your computer and capture it with a video capture program and that should get rid of any tags or metadata. Then upload the captured video. Or maybe display it on a 4k screen and video it with your phone then upload that as the metadata would only show that it was captured on a phone not created by AI. |
|
Originally Posted By SWIRE: Nice, what is H3? A newly-released open-source model with day 1 support in ComfyUI. I would say that it has been better received than any that came before. Much better than the previous models, imho. The BF16 version that I use is about 60+ GB in size. The VAE and text encoder are also large. The vast majority of people stick with the nerfed fp8 versions. |
And then, he arrives at the last item on his bucket list...
|
Originally Posted By GySgt_D: The BF16 version that I use is about 60+ GB in size. The VAE and text encoder are also large. The vast majority of people stick with the nerfed fp8 versions. Luck was finally on my side for a change, I bought a beefy computer with a 5090, 4TB M.2, and 64GB of ram about a month before all the chip makers decided to switch to AI chips. Still for AI video that is barely touching the compute needed. A 50 second video at 720p is about the max before it crashes or starts deforming the character. BTW, Comfy 29.2 has some big dynamic vram updates compared to 28.8. At first my startup file would no longer work and I had to specify low vram for it to run " --lowvram --reserve-vram 1" and then everything ran slower. However Comfy used to crash around the 30 second video mark because it ran out of vram. Now I can get close to a minute and it doesn't crash. Going over a minute it still doesn't crash but it sits there grinding and would take days to finish. It is an overall improvement but still not great. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
Well I did it and created a AI thread in Babes of the day in team.  https://www.ar15.com/forums/topic.html?b=1&f=142&t=2854467&tl=Well-I-guess-its-time-to-post-some-of-my-AI-pics-please-look-before-hating-LOL |
 3D girls are gross 14 steps H3, followed by 3 upscaling/refining steps with LTX2.3 |
And then, he arrives at the last item on his bucket list...
|
Originally Posted By GySgt_D: https://www.youtube.com/watch?v=GRec7VrdTmk14 steps H3, followed by 3 upscaling/refining steps with LTX2.3  LOL, looks great!! Only thing I would do is have her speak in mandarin or at least have an accent. Been wanting to try H3 but having too much fun with creating stills. |
 Ignore the hole in my head. It's for batteries... 10x better than the LTX2.3 that I had been using for last several months. |
And then, he arrives at the last item on his bucket list...
|
Mundania's favorite Little Leaguers. And a clue why Jennifer cut most of her hair off. Attached File |
"He should have killed me. I would have killed me."
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
 H3 15 steps Spent 11 hours today tweaking settings, and replacing the final steps LTX2.3 treatment. The latter was badly changing the output. Now using Nvidia VFX Super Resolution. *Much* faster, with little or no bad effects. |
And then, he arrives at the last item on his bucket list...
|
Originally Posted By GySgt_D: https://www.youtube.com/watch?v=a3MOiRI_9JASpent 11 hours today tweaking settings, and replacing the final steps LTX2.3 treatment. The latter was badly changing the output. Now using Nvidia VFX Super Resolution. *Much* faster, with little or no bad effects. What video card are you using? |
|
Aria-R, (22) talks with The Great Texarrakis, Texas Chainsaw BBQ and Amusement Park. His smile is because he just learned she is also Licensed and Registered CPA in the Great State of Texas. “Dis is going to be FUNNNNNNN!” the Agent gleefully said. Aria-R signs a Letter of Intent, agreeing in principle to join The BBQ soon after her contract with The Troupe expired. She never told the Agent that she didn't receive her CPA and License until about two weeks previously. Attached File |
"He should have killed me. I would have killed me."
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.
For God and Country: Geronimo! Geronimo! Geronimo!
Michael Moore: Trump’s election is going to be the biggest Fuck You ever recorded in human history….And it will feel good.