Tutorial Stable Diffusion AI: Automatic1111 Beginner's Guide

[riniji]

Ganito kasikat yang AI image na yan, kaya pwede mo ngang mapagkamalang tao he he.

View attachment 2673901 View attachment 2673902 View attachment 2673903 View attachment 2673904 View attachment 2673905 View attachment 2673906 View attachment 2673907 View attachment 2673908 View attachment 2673909 View attachment 2673910 View attachment 2673911 View attachment 2673912 View attachment 2673913 View attachment 2673914 View attachment 2673915 View attachment 2673916 View attachment 2673917 View attachment 2673918 View attachment 2673919 View attachment 2673920 View attachment 2673921 View attachment 2673922 View attachment 2673923 View attachment 2673924 View attachment 2673925 View attachment 2673926 View attachment 2673927 View attachment 2673928 View attachment 2673929 View attachment 2673930 View attachment 2673931 View attachment 2673932 View attachment 2673933 View attachment 2673934 View attachment 2673935 View attachment 2673936 View attachment 2673937 View attachment 2673938 View attachment 2673939 View attachment 2673940

ang kyut ng seed na to

best quality, masterpiece, (photorealistic:1.4), 1girl, light smile, shirt with collars, waist up, dramatic lighting, from below
Negative prompt: nsfw, ng_deepnegative_v1_75t,badhandv4, (worst quality:2), (low quality:2), (normal quality:2), lowres,watermark, monochrome
Size: 512x768, Seed: 3698311310, Model: majicmix_realv609, Steps: 30, Sampler: Euler, CFG scale: 7, Clip skip: 2, Model hash: bbbb7f92f0, Hires steps: 15, Hires upscale: 2, ADetailer conf: 48, Hires upscaler: 8x_NMKD-Superscale_150000_G, ADetailer model: face_yolov8s.pt, ADetailer version: 23.5.6.post0, ADetailer x offset: 0, ADetailer y offset: 0, Denoising strength: 0.12, ADetailer CFG scale: 7.0, ADetailer mask blur: 4, ADetailer dilate/erode: 12, ADetailer inpaint full: True, ADetailer inpaint width: 512, ADetailer inpaint height: 512, ADetailer inpaint padding: 0, ADetailer denoising strength: 0.15, ADetailer use inpaint width/height: True

 

[alist1986]

majicmix_realv609

Oo, thanks to the various majicMIX checkpoints for SD GUI you'll get that typical girl look. Kung gagamitin sa default SD 1.x - 2.x yang prompts ay malayo at super-variable ang results kahit pa may Seed. Pero sa Dreamshaper 5 and 6 kuha yung resemblance as an example maging yung AbsoluteReality checkpoints kahit default prompts lang (aside from the other settings mentioned).

Using available ckpt helps the user to choose which satisfies the art result he wants to create using available ckpt, safetensors, etc. Though madali kung may resources ka using known checkpoint mergers. Pero kung wala, magpapakahirap kang mag-train pa at mag-experimento ng weights, etc. Sa image creation pa lang yan. Kaya malaki ang advantage sa normal users kahit wala pang knowhow sa image generation.

Hinahanapan ko pa ng ibang gamit yang AI generated art tools aside from this stuff. Maraming ipinasok na sa SD at iba pang similar AI generators na machine learning models (You do not have permission to view the full content of this post. Log in or register now.) like GSGAN, ESRGAN, atbp. na pwedeng pakinabangan.
Ito yung maaaring next learning study after familiarizing installation and usage of these available applications.
You do not have permission to view the full content of this post. Log in or register now.
You do not have permission to view the full content of this post. Log in or register now.
Siyempre, kung mag-prompt ka sa sarili mo. Ang tanong ay " What's next?". He he. I-career mo na. Go deeper (kaysa cloning of same art lang) while continuing practice of your favorite art tool.

 

[Myexistence]

Naaddict na ako sa Majicmix model, ang gaganda ng mga nakukuha ko.

You do not have permission to view the full content of this post. Log in or register now. You do not have permission to view the full content of this post. Log in or register now. You do not have permission to view the full content of this post. Log in or register now. You do not have permission to view the full content of this post. Log in or register now.
You do not have permission to view the full content of this post. Log in or register now. You do not have permission to view the full content of this post. Log in or register now.

Pero malupit yung gumamit ng EBSynth para gumawa ng sizzling shortform videos....

 

[alist1986]

EBSynth para gumawa ng sizzling shortform videos

Eh di friends mo na si xxxHolic he he.

Maganda nga yang EBSynth sa mahilig din sa Tiktok. Galing ng videos.
Free pa for commercial use he he, at maliit lang yung hardware requirements ng binasa ko.
The minimum system requirements for EbSynth are:

• Windows 7 or macOS 10.11 or newer
• 4GB of RAM
• 2GB of available disk space
• A graphics card with support for OpenGL 3.3 or newer

The recommended system requirements for EbSynth are:

• Windows 10 or macOS 10.13 or newer
• 8GB of RAM
• 4GB of available disk space
• A graphics card with support for OpenGL 4.0 or newer

Buti binanggit mo.

Napako ako sa kaka-testing ng sangkaterbang GPT offline apps kaya wala muna graphics. Napapagana ko pero mabagal din tulad ng SD din. Low hardware kasi.
Magandang kalikutin din yan at ang dami ko na rin sources ng mga galleries ng AI prints na pwedeng gawan ng videos. Buti sinabi mo at nalampasan ko yan sa aking searches at di ako masyado sa graphics area.

 

[madhatter89]

ang cool naman neto! power talaga basta AI

 

[alist1986]

You do not have permission to view the full content of this post. Log in or register now., Pre, madali na online ito at marami ng crowdsourced clusters na nag-provide nito with assorted models - mga online volunteers. Di lang masyadong customized using controlnet pero good as original yang "stable horde" service - free to use pa. Di lang mabilis pero good for people who wants to use the most available models and common settings using latest A1111 UI and comparble with Colab.
Napuno na pc ko sa kaka-test ng mga GUI din kaya makatipid ako space pag binura ko na he he.
Yung ToonYou related pic sa thread ay halos same din using CIVITAI settings.

Pero may iba rin online AI Generator service na mas OK, but SH sa github is just right for normal use w/o Colab.
Maganda rin sa landscape at pang-competition yung results. Maghanda ka lang ng Topaz AI apps, pang-commercial na.
Bukod sa image, may TXT generation (using opensource GPT) din sila with decent Google AI voice assistance atbp. Basta may volunteers ng hardware online, pwede kang gumamit. Para yang P2P service to share your hardware online with the rest of the world he he. Dati kong gawain noon kaya naalala ko lang kung merong ganyan using GPT search. May nakaisip na ng ganito sa opensource community. Good.

 

[Myexistence]

Yes marami nang mga alternative apps/programs/ generation sites ang nag sisilabasan ngayon. Ishare ko lang yung mga favorites.

You do not have permission to view the full content of this post. Log in or register now. - Bago ko ito nasubukan at ang kinagandahan nito ay ay may controlnet siya na kasama at pwede ka mag generate ng images sa #forbidden# server nila
You do not have permission to view the full content of this post. Log in or register now. - alternative ko sa Leonardo AI

May isa pa ako inaabangan yung MAKEAYO, desktop program siya na parang automatic1111 pero made for newbies.

 

[alist1986]

Marami nga dyan sa underground dens ng disc@rd atbp. ang gathering he he. One to sawa yung generation. Sa GUI naman, ganoon din, may portable versions for desktop and USB stick - readymade na. Pero ang importante, kung may hardware ka to run it without crashing. Ganyan talaga pag sikat, i-google mo na lang. Maiiwan ka kung magtatanong sa sa thread.

Habang tumatagal, mas tumataas yung GPU requirement sa dami ng features and plugin/extensions ng AI ART generators. Yung mga unang labas lang yung napagtiyagaan ko sa CPU mode kahit matagal mag-generate he he.

Ang addiction ko, hindi yung generation ng graphics, kundi papaano paganahin kahit wala sa specs/OS. Sa online ako naghanap ng testing ground (aside from Colab notebooks) sa graphic generation. Pwede naman hiwalay yung ibang features ng SD basta alam mo kung saan merong service nila like img2img, upscaling, ebsynth, etc. Kabisado mo dapat yung development ng app na ito at iba pang kamukha niya. Basta well-versed ka sa AI GPT technology (online and offline), yung branches ng text and art generation ay mapapadali sa'yo. Matagal na pagbabasa, pagsali sa AI art forums at hands-on. Kahit wala kang programming background pwede na ngayon.

Sa beginners, mag-umpisa na kayong kausapin yang malalaking GPT services ng OPENAI, Google at Microsoft thru voice or text, malalaman ninyo sa madaling panahon kung papaano sila gamitin. It all starts with a "prompt", and discover its powers. Mahirap sa umpisa, pero casual na lang sa kalaunan. Use this technology for your own benefit. Full-featured na yan ngayon kahit free versions.

 

[jessicalanye]

ganda nya

 

[alist1986]

ganda nya

View attachment 2679288

Medyo di normal yung alignment ng ulo noong nasa kaliwa. Kailangan pang itulak patalikod sa tingin ko. Yung sa kanan, medyo mahaba yung leeg. Pero sa visual ay perfect babes he he.

 

[riniji]

Spoiler: generation data

Spoiler contents are visible only to Established Members.

eto model from civitai

Spoiler: ICBINP

Spoiler contents are visible only to Established Members.

 

[johnking52]

ano bang checkpoint ito nakita ko lang sa website

 

[alist1986]

ano bang checkpoint ito nakita ko lang sa website

View attachment 2680061

Hanapin mo na lang sa civitai.com yan. Gamitin mo yung "magicmix realistic" as search query at tingnan from version 2 - 6.
Mata lang yung ginamit ko at familiar ako sa mukha. Mag-experiment ka sa age para lumapit doon sa pic mo. Medyo 5 yrs younger yung nakita kong checkpoints compared sa pic mo.
Ang daming channel sa patreon na nagbebenta ng mukha na yan he he. Pag may oras, ihanap kita ng exact settings. Sa GPT na kasi yung hobby ko, wala na dyan, though connected.

 

[alist1986]

ano bang checkpoint ito nakita ko lang sa website

View attachment 2680061

Nag-test lang ako ng ilan para mapalapit dyan pero di ko pa makuha yung tamang setting. Nasa attachment yung output. Nairapan ako't kulang ako ng mga extensions at kalimitan, misaligned yung mata he he. Medyo ok naman. Ito yung original settings, pero di ko nasunod lahat. Ang models ko "SD+magicmix realistic" tapos sa upscaling hinaluan ko ng GFPGAN para sa corrections ng mata.

Spoiler

Spoiler contents are visible only to Established Members.

 

[johnking52]

Hanapin mo na lang sa civitai.com yan. Gamitin mo yung "magicmix realistic" as search query at tingnan from version 2 - 6.
Mata lang yung ginamit ko at familiar ako sa mukha. Mag-experiment ka sa age para lumapit doon sa pic mo. Medyo 5 yrs younger yung nakita kong checkpoints compared sa pic mo.
Ang daming channel sa patreon na nagbebenta ng mukha na yan he he. Pag may oras, ihanap kita ng exact settings. Sa GPT na kasi yung hobby ko, wala na dyan, though connected.

binebenta pala itong klaseng art grabe.

Hanapin mo na lang sa civitai.com yan. Gamitin mo yung "magicmix realistic" as search query at tingnan from version 2 - 6.
Mata lang yung ginamit ko at familiar ako sa mukha. Mag-experiment ka sa age para lumapit doon sa pic mo. Medyo 5 yrs younger yung nakita kong checkpoints compared sa pic mo.
Ang daming channel sa patreon na nagbebenta ng mukha na yan he he. Pag may oras, ihanap kita ng exact settings. Sa GPT na kasi yung hobby ko, wala na dyan, though connected.

salamat paps

 

[alist1986]

binebenta pala itong klaseng art grabe.


salamat paps

Paborito kasi yung mga ganyang "SD ckpt" AI pics na asian beauty ngayon as avatar at profile pictures he he. Ang daming dedicated channels para sa ganyang face features sa net. Kaya mo rin mag-generate nyan. Sa'kim kasi mga 900 sec per pic kaya nakakainip din he he. Di ako makagalaw sa pc during art generation at cpu mode yung gamit ko using the GUI. Mamaya, subukan ko ulit habng nagluluto he he, pero best is medyo unique ng kaunti sa gusto mong source pic.

Maraming tools din sa hugginface to check possible prompts based on a pic aside from SD prompt reader. Tipong image to prompt. halos standard na yung negative prompts base sa normal use ng specific checkpoints.

Bukod sa MagicMiX series, yung ChilloutMix versions ang isa rin popular na madalas haluhan ng Cute_girl_mix4 para for a Chinese feel or yung Normal Korean girl face, Chilloutmix base lora. Check mo rin yung origins ng ckpt para makapag-customize ka ng sarili mo. Kalimitan dyan merge na with others to fit their needs. Marami rin custom versions ng mga yan na mahirap hanapin sa civitai or huggingface na hindi na kailangan ng VAE or LoRas. Sa mga SD forums mo yan makukuha with links. Talagang matagal na pag-aaral at generation para makuha mo yung aesthethic ng mga pics na gusto mong i-control. Ito yung ibang results na na-retrieve ko kanina na kasama noong una sa last post. Not perfect pero naayos na man yan ng Prhotoshop,Topaz, atbp.


Ang last tip kong iiwan na mapapakinabangan ninyo, don't rely on Google search. Sobrang filtered yan na halos wala kang makuhang "specific" results mapa-SD accessories or image search. Yandex ang gamitin ninyo para maturo kayo agad especially for this AI category.

 

[johnking52]

Paborito kasi yung mga ganyang "SD ckpt" AI pics na asian beauty ngayon as avatar at profile pictures he he. Ang daming dedicated channels para sa ganyang face features sa net. Kaya mo rin mag-generate nyan. Sa'kim kasi mga 900 sec per pic kaya nakakainip din he he. Di ako makagalaw sa pc during art generation at cpu mode yung gamit ko using the GUI. Mamaya, subukan ko ulit habng nagluluto he he, pero best is medyo unique ng kaunti sa gusto mong source pic.

Maraming tools din sa hugginface to check possible prompts based on a pic aside from SD prompt reader. Tipong image to prompt. halos standard na yung negative prompts base sa normal use ng specific checkpoints.

Bukod sa MagicMiX series, yung ChilloutMix versions ang isa rin popular na madalas haluhan ng Cute_girl_mix4 para for a Chinese feel or yung Normal Korean girl face, Chilloutmix base lora. Check mo rin yung origins ng ckpt para makapag-customize ka ng sarili mo. Kalimitan dyan merge na with others to fit their needs. Marami rin custom versions ng mga yan na mahirap hanapin sa civitai or huggingface na hindi na kailangan ng VAE or LoRas. Sa mga SD forums mo yan makukuha with links. Talagang matagal na pag-aaral at generation para makuha mo yung aesthethic ng mga pics na gusto mong i-control. Ito yung ibang results na na-retrieve ko kanina na kasama noong una sa last post. Not perfect pero naayos na man yan ng Prhotoshop,Topaz, atbp.

View attachment 2680769 View attachment 2680770 View attachment 2680771 View attachment 2680772 View attachment 2680773 View attachment 2680774 View attachment 2680775 View attachment 2680776
Ang last tip kong iiwan na mapapakinabangan ninyo, don't rely on Google search. Sobrang filtered yan na halos wala kang makuhang "specific" results mapa-SD accessories or image search. Yandex ang gamitin ninyo para maturo kayo agad especially for this AI category.

ano gamit mo na stable diffusion na site?

 

[alist1986]

ano gamit mo na stable diffusion na site?

Dyan sa pics, pc ko gamit ko na cpu mode. Gumawa ako ng thread to make it work noon:
You do not have permission to view the full content of this post. Log in or register now.
Check mo yung packs dyan kung kaya ng pc mo.
Sa SD sites, minsan nag-test ako pero madalas galing sa "There's an AI for That" site under "Stable Diffusion" query. Ini-isa-isa ko sila hanggang meron akong makitang ok. Wala akong preferences. Test and go lang. Sinabi ka na yan sa mga previous posts ko.
Marami ryan nasa dic@rd channels na wan-to-sawa ka mag-generate pero di ako mahilig sa ganyang running posts na ang daming members na wala akong kilala - nahihilo ako sa kakapindot he he. Sa developer ako nakikipag-usap ng direkta para mas madali hangga't maari via email pag gusto ko talaga yung application at active sa source code. Kung as user lang, sa net lang ako nagbabasa ng guides via search engines. Kung "no holds bard" na forum, join ka sa reddit. Marami tips dyan at kung saan ka makapag-test.
Kung gusto mong talagang gayahin yung "ai art", talagang magtitiyaga ka sa image search hanggang makuha mo yung link ng artist, at kung meron siyang nilabas na tips (via html pages, embedded txt sa pics, videos) same like civitai kung napansin mo. Para kang imbestigador he he, gagamitin mo yung iniwang bakas at doon ka mag-umpisa. Sa hands-on na lang makikita, depende kung mai-apply mo lahat ng prompts, settings, extensions sa iyong gamit na SD GUI. Sa'kin, civitai and huggingface lang for Stable Diffusion ay ok na,

 

[johnking52]

cge po salamat po try ko

 

[alist1986]

cge po salamat po try ko

Yung HappyAccidents link sa post#87 ni Myexistence ay OK naman; medyo tricky lang kung papaano i-link yung gusto mong ckpt dyan via civitai. Parehas ng style siya using Stadio. Pati yung opensource project ng Stable Horde ayos din pagtiyagaan ng balikan ko. Gamit ka ng GPT na lang for the infos para mas klaro. Umikot ka sa mga ckpt ng MagicMIX, ChilloutMIX, and Realistic Vision to create your unique AI Asian character. Kahit walang seed, mas mainam para makapili ka ng pinupuntirya mong looks.
Dito lang ako natuto sa ngayon:
1. You do not have permission to view the full content of this post. Log in or register now. - sa tutorials
2. You do not have permission to view the full content of this post. Log in or register now. - sa ebook
Yung mga samples sa civitai ckpt/loras pics na lang yung guides sa testing.