jhc13
commented
5 months ago Commas will be removed from the generated captions if you check Remove tag separators in caption (which is checked by default).
You can use Find and Replace on . to convert them to ,.
I did a captioning and i received for example this prompt:
metal print, vibrant colors, neon lights, futuristic vehicles, skull motifs, intricate designs, action-packed scene, dynamic composition, low-angle viewpoint, cyberpunk aesthetic, detailed textures, and stylized portrayal of characters and machinery. The image captures a dystopian future where characters are equipped with advanced weaponry and technology, set against a backdrop of neon-lit streets and high-tech vehicles. The color palette is dominated by purples, reds, and yellows, giving it a nocturnal ambiance. The composition is dynamic, with characters positioned in action poses, suggesting movement and urgency. The lighting plays a crucial role in setting the mood, casting shadows and highlights that add depth and dimension to the scene. Overall, the image exudes a sense of chaos, adventure, and technological advancement.
I want to avoid setting commas for example "The color palette is dominated by purples, reds, and yellows, giving it a..." should be "The color palette is dominated by purples reds and yellows giving it a nocturnal ambiance."
When there is a new sentence like "...characters and machinery. The image captures..." should be "...characters and machinery, (comma) The image captures..." to make it a new tag.
model used: THUDM/cogvlm-chat-hf