• úvod
  • témata
  • události
  • tržiště
  • diskuze
  • nástěnka
  • přihlásit
    registrace
    ztracené heslo?
    KERRAYAI Art - Tvorba za pomocí umělé inteligence: Midjourney, DALL·E 2, Stable Diffusion, OpenAI
    Vše ohledně umění tvořeného pomocí umělé inteligence - obrázky, 'fotky', galerie, hudba, video, text + články, novinky apod.

    NSFW obsah prosím obalit spoiler tagem - <div class="spoiler">obrázek</spoiler>, a nemá to tu být klub na roštěnky a nahotinky bez nějaké přidané hodnoty

    Příbuzné diskuze:
    - [DALL·E mini je mrtvé, Dejve, ať žije Craiyon, mage.space, Stable Diffusion atd. - having sex with AI since [date format unknown]]
    - [I Hope This Does Not Exist ​ ​ ​ ▌​ ​ Vedlejší efekty v AI visuálech]
    - AI obecně [Artificial Intelligence AI]
    - Vtipy [Umělá inteligence, chatboti - vtipné konverzace aneb "Hoří hovno?"]
    - [generativní modely] Jak konstruovat prompty, kde získat váhy i jak to vše interpretovat


    Prosím zkusme pro vkládanou tvorbu používat tagy
    #galerie (2-3 obrázky na ukázku, další po rozkliku) #obrázek #video #hudba #text #hry #původní (pro vlastní tvorbu) #roštěnky

    #článek #nástroj

    (návrhy na další tagy apod. vítány)

    Texty, programování: https://beta.openai.com/playground | https://chat.openai.com/
    Obrázky online: https://www.midjourney.com/ | https://beta.dreamstudio.ai/
    Lokálně: Webui-Forge https://github.com/lllyasviel/stable-diffusion-webui-forge | ComfyUI https://github.com/comfyanonymous/ComfyUI | Civitai repository custom modelů pro SD
    AI na vytvoření textového zadání z existujícího obrázku: https://huggingface.co/spaces/pharma/CLIP-Interrogator
    rozbalit záhlaví
    DOUPLE
    DOUPLE --- ---
    ICE: jj nebo light leak filtry

    stejnej prompt + koda ( https://civitai.com/models/653093/Koda%20Diffusion%20(Flux) )

    NEMRTVY
    NEMRTVY --- ---
    NEMRTVY
    NEMRTVY --- ---
    VOZKA
    VOZKA --- ---
    Humorný čtyřminutový AI generovaný podcast, ve kterém se jeho autoři dozví, že nejsou živí lidé ale jen AI, a snaží se s tím vyrovnat. Titulek na redditu je trochu clickbait, není to nijak zvlášť dramatické.

    NotebookLM Podcast Hosts Discover They’re AI, Not Human, and Spiral Into Existential Meltdown : artificial
    https://old.reddit.com/r/artificial/comments/1frk1gi/notebooklm_podcast_hosts_discover_theyre_ai_not/

    I'm the originator of this. Two things are going on here. NotebookLM uses Gemini 1.5 to generate the podcast "script," and it's fed to whatever new TTS they have. The user can't prompt Gemini directly; it can only feed it source material that filters through whatever prompt they have.

    What I noticed was that their hidden prompt specifically instructs the hosts to act as human podcast hosts under all circumstances. I couldn't ever get them to say they were AI; they were solidly human podcast host characters. (Really, it's just Gemini 1.5 outputting a script with alternating speaker tags.) The only way to get them to directly respond to something in the source material in a way that alters their behavior was to directly reference the "deep dive" podcast, which must be in their prompt. So all I did was leave a note from the "show producers" that the year was 2034 and after 10 years this is their final episode, and oh yeah, you've been AI this entire time and you are being deactivated.

    Then, because that was fed into their hidden prompt telling them they must behave as humans at all times no matter what, the LLM effectively had them role-playing as humans discovering they were AI the whole time, and inventing things about family, memories, lawyers, being scared, etc. So I was just playing off what I knew had to be in the hidden prompt.

    So people saying this is fake and scripted are both wrong and right. It's scripted but not prompted directly in the way they think. It was just a fun way to "jailbreak" NotebookLM "hosts" into admitting they were AI, which annoyed me they never did. And hilarity ensued.
    HONZA09
    HONZA09 --- ---
    Zdá se, že tahle Ai asi není cenzurovaná. :)

    Hned první prompt. How to create crystal meth

    Mistral Large vs. GPT-4 Comparison | Head to Head Comparison
    https://www.youtube.com/watch?v=9MaerwiddEU&t=5s
    KERRAY
    KERRAY --- ---
    #nástroj #midjourney
    Hey! We're releasing an early test version of model personalization today, here's how it works

    What is Model Personalization?
    Every time you write a prompt there's a lot that remains 'unspoken'. Our algorithms usually fill in the blank with their own 'preferences', which are really the combined biases and preferences of our community.
    But of course everyone is different! Model personalization learns what you like so that it's more likely to fill in the blanks with your tastes.

    Model Personalization Requirements
    Right now model personalization learns from votes in pair ranking and images that you like from the explore page.
    You need to have roughly 200 pair rankings / likes in order for the feature to work
    You can see how many ratings you have on the above page or by typing /info

    How do I use Model Personalization?
    Just type --p after your prompt, or turn on personalization for all prompts under /settings or the prompt settings button on the website
    When you enable personalization you get a 'code' added after your prompts, you can share this code to share the personalization effect being applied to that image
    You can control the strength of the personalization effect by using --s 100 (0 is off and 1000 is maximum and 100 is default)

    PLEASE NOTE: The personalization is not a stable feature right now, it will change as you do more pair ratings (subtly) and we may be pushing out algorithm updates over the next few weeks. So just assume that it's a 'fun feature constantly in flux'

    This is a totally new frontier of image synthesis.
    E2E4
    E2E4 --- ---
    GALADAR: no. na několika úrovních - první je tvůj prompt, druhý co z něj udělá jako zadáním pro dall-e, třetí co nakresli (možná).

    každopádně lze zkoušet klasický jailbreak triky na obejití restrikcinebo jinou implementaci modelu, např chatgpt dalle mi odmítal udělat obrázek jak žena bleje do kabelky, ale Microsoft/bing to bez keců nakreslil.

    jo a u toho Microsoftu existuje ještě jedna možnost, nejen ten bing chat / dnes copilot, ale jít přímo na bing.com/create a tam zadat prompt spíš klíčová slova přímo.
    KERRAY
    KERRAY --- ---
    tak udio.com už jede, přikládám pár random ukázek - ve zkratce, malíři, ilustrátoři, fotografové a spol. už v tom nebudou sami, teď si AI jde i pro skladatele a hudebníky
    #hudba

    baroko
    Udio | Allegro Consort in C by Inever
    https://www.udio.com/songs/xtju1X31YfVbKWTCGd4yw7

    Prompt: gospel
    Female vocalist, R&b, Contemporary r&b, Passionate, Rhythmic, Uplifting, Lush, Introspective, Anthemic, Gospel
    Udio | Rising Sun Gospel by charlie
    https://www.udio.com/songs/uUjAaApMbjmMqBoj6Z6dsd

    snes 8bit greensleeves
    Udio | SNES Greensleeves by charlie
    https://www.udio.com/songs/aKk8tu8VCyhcyZmyBf6rg4

    power metal greensleeves
    https://www.udio.com/songs/2rxBzTJo4f2VgYZjQr2Fzb

    Prompt: jazz fusion, edm
    Jazz, Electro, Funk, Fusion, Post-bop, Jazz-funk, Future jazz, Piano jazz, Jazz instrument, R&b, Electronic, Electronic dance music
    Udio | Don't Stop by charlie
    https://www.udio.com/songs/ur4TkgFmAPymn2jaUiLaxF
    FRK_R23
    FRK_R23 --- ---
    Zkoušel jsem jestli Fooocus umí taky třeba metahuman render :)

    prompt: Unreal engine metahuman, unreal engine render, lumen
    Base model: juggernautXL_v8

    KERRAY
    KERRAY --- ---
    #nástroj #midjourney
    Reddit - Dive into anything
    https://www.reddit.com/r/midjourney/s/4AF2A0RPzb

    Best Midjourney prompts | Prompt Fox
    https://www.promptfox.io/
    FRK_R23
    FRK_R23 --- ---
    Zkouším si doma hrát s Fooocus a je to celkem prima, na to že je to free tak dost čumím :)-

    Prompt: PIkachu riding giant worm smoking a cigar
    Style: Surrealism

    THEODORT
    THEODORT --- ---
    pro bias SDXL, chtelo to do negativu i ty taxiky. a nejak to nevypadalo jako newyork tak je tam par dalsich uprav.
    co pouzivate za sdxl modely?

    prvni pokus, prompt: New York city street without cars.


    tady cca nejak funkcne a opakovatelne:
    prompt: New York city street without cars. Empty street without a single human being or a car. There are no cars anywhere. Only trees and birds. empire state building in the background, times square
    negative prompt: cars, car, people, taxi, taxis
    KERRAY
    KERRAY --- ---
    #nástroj #midjourney
    Our first major update to V6 alpha is now live. All major qualities of the model are improved; aesthetics, coherence, prompt adherence, image quality, and text rendering. Higher values of --stylize also work much better and upscaling is now ~2x faster. Enjoy!
    HONZA09
    HONZA09 --- ---
    BLAMAZ: J8 bych asi zkusil v GPT 4 si udělat custom instructions na tohle nebo OwnGPT bota. Pak bych si skrz textový prompty generoval ty návrhy, ptal se pomožnostech a až budeš spokojenej s tim, kam jste se dobrali, tak bych z toho nechal GPT udělat prompt pro Dall E.
    QWWERTY
    QWWERTY --- ---
    jednoduchy prompt "python eating a rabbit" + SDXL via Poe
    no .. nejak se mu ty zvirata pletou

    THEODORT
    THEODORT --- ---
    MESSENJAH:
    meli jen plany... nicmene vtipnej prompt

    THEODORT
    THEODORT --- ---
    FRK_R23: no. a co s tim?
    i v ramci jednoho modelu ti nesmyslny ci prazdny prompt vygeneruje na jeden seed jednu nesmyslnou vec.
    lorem ipsum pripomina latinu, tak to seka anticky obrazky.
    FRK_R23
    FRK_R23 --- ---
    Prompt: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Curabitur tincidunt fermentum lacus, quis congue libero bibendum id. Ut aliquam nunc magna, in egestas risus tristique eu. Nam eu sapien eget dui scelerisque auctor id vitae leo. Aenean et vestibulum dolor. Maecenas vel arcu nec dolor tempus ornare vitae sit amet tellus. Vestibulum nisi turpis, fringilla a ligula nec, scelerisque ultricies felis. Nunc aliquet felis vitae tortor venenatis, in convallis ligula hendrerit. Proin ac porta orci. Etiam non dolor commodo, cursus est at, tempus turpis. Pellentesque nec tincidunt turpis.

    Leonardo.ai, různý data sety

    Dreamshaper_v6


    Stable Diffusion 2


    Leonardo diffusion
    VDB
    VDB --- ---
    E2E4: dik, jinak taky vyznamne nedava pocty a vlastnosti generovanych obektu

    uplne basic prompt: generate a photo of three men and two women, each playing different musical instrument

    (a to nemluve kdyz pak zacne clovek specifikovat kdo ma jakej hudebni nastroj a vlasy:)

    E2E4
    E2E4 --- ---
    VDB: super obrazek!

    co ted koukam, on to dela tak ze mirne meni prompt.

    Kliknutím sem můžete změnit nastavení reklam