Uma ferramenta de engenharia rápida baseada em vicunha para difusão estável
gerar imagens incríveis sem conhecimento imediato
Exemplos - Instalação - Uso
O arquivo Stablediffy.txt é um arquivo de prompt para o modelo Vicuna, semelhante ao arquivo chat-with-bob.txt. Primeiro, a situação é descrita. Em segundo lugar, algumas palavras-chave como “relâmpago dramático” são listadas para ensiná-las à Vicuna. Terceiro, alguns exemplos de respostas do Stablediffy às perguntas aproximadas do humano são listados para ensinar a Vicuna como responder.
Você também pode experimentar o arquivo de prompt com outros modelos de linguagem de IA.
### Human:suggest a prompt for an mysterious house
### Assistant: Mysterious house, haunted mansion, spooky, intricate, sharp details, octane render, unreal engine, photograph, realistic skin texture, photorealistic, hyper realism, highly detailed, 85mm portrait photography, award winning, hard rim lighting photography, dramatic lighting, cinematic lighting, gritty atmosphere, urban, noir style, night scene"
### Human:suggest a prompt for a mysterious lake in the night
### Assistant: Mysterious lake, night scene, dark and moody, atmospheric, spooky, gothic, creepy, haunting, eerie, suspenseful, dramatic lightning, cinematic lightning, octane render, unreal engine, realistic reflections, water reflections, moonlight, starry sky, mysterious atmosphere, dark and gritty
### Human:suggest a prompt for a ghost ship on the ocean
### Assistant: Ghost ship, abandoned ship, creepy atmosphere, spooky, dark, ominous, mysterious, cinematic lightning, dramatic lightning, octane render, unreal engine, realistic, highly detailed, textured, 3D modeling, digital art, 3D illustration
### Human:suggest a prompt for a portrait of Danial Craig
### Assistant: Daniel Craig portrait, actor, famous, well-known, blue eyes, chiseled jawline, suave, sophisticated, realistic, hyper-realistic, highly detailed, 85mm lens, photography, realistic skin texture, photorealistic, noir style, cinematic lightning, dramatic lightning, portrait photography, sharp focus, colorful"
Para instalar o Vicuna, siga as instruções neste repositório: https://github.com/vicuna-tools/vicuna-installation-guide/
Feito isso, navegue até a pasta llama.cpp/prompts
Agora, baixe o arquivo Stablediffy.txt com wget
wget "https://raw.githubusercontent.com/vicuna-tools/Stablediffy.txt/main/Stablediffy.txt"
Navegue de volta para a pasta llama.cpp
Execute o modelo de vicunha com o arquivo de prompt Stablediffy.txt (13B)
./main -m models/ggml-vic13b-q5_1.bin -f 'prompts/Stablediffy.txt' -r 'USER:'
Execute o modelo de vicunha com o arquivo de prompt Stablediffy.txt (7B)
./main -m models/ggml-vic7b-q5_1.bin -f 'prompts/Stablediffy.txt' -r 'USER:'
Isso deve iniciar o processamento do arquivo Stablediffy.txt após o último exemplo de conversa processado:
USER: Car in 1950, highly detailed, classic car, 1950's...
USER:
, onde você pode começar a solicitar prompts de difusão estável