Meta tags:
description=DiffSketcher, text-to-sketch, text2sketch, text-to-svg, text2svg, vector sketch;

Headings (most frequently used words):

results, scene, bibtex, abstract, methodology, demo, more, still, life, cartoon, character, complex, painterly, rendering, colorful, oil, painting, style, transfer, citation,

Text of the page (most frequently used words):
the (20), diffusion (11), sketch (10), diffsketcher (8), that (8), text (7), model (7), and (6), synthesis (5), #through (5), latent (5), models (5), guided (4), vector (4), sketches (4), ximing (3), xing (3), prompt (3), free (3), hand (3), paper (3), vectorized (3), this (2), close (2), copy (2), bibtex (2), chuang (2), wang (2), haitao (2), zhou (2), jing (2), zhang (2), qian (2), dong (2), 2023 (2), painting (2), results (2), scene (2), detailed (2), demo (2), from (2), such (2), raster (2), frozen (2), directly (2), predicts (2), image (2), trained (2), show (2), innovative (2), algorithm (2), creates (2), using (2), natural (2), language (2), input (2), optimizing (2), prior (2), for (2), university (2), webpage, template, was, originally, made, sdiffsketcher, success, inproceedings, xing2023diffsketcher, title, author, booktitle, thirty, seventh, conference, neural, information, processing, systems, year, url, https, openreview, net, forum, cy1xatveqj, citation, style, transfer, oil, colorful, painterly, rendering, complex, cartoon, character, still, life, more, sailboat, sailing, sea, clear, day, very, masterpiece, baby, yoda, hoding, lightsaber, macaw, full, color, ultra, realistic, insanely, beautiful, generates, synthesize, matches, given, optimize, parameters, differentiable, rasterizer, produces, resulting, sample, blue, part, picture, since, update, direction, not, need, backpropagate, simply, acts, like, efficient, critic, space, edits, photo, eiffel, tower, methodology, even, mainly, images, discover, pretrained, impressive, power, guiding, present, developed, based, pre, performs, task, set, bézier, curves, with, which, allows, use, level, parametric, generator, furthermore, explore, attention, maps, embedded, speed, generation, process, generated, demonstrate, multiple, levels, abstraction, while, maintaining, recognizability, underlying, structure, essential, visual, details, subject, drawn, our, experiments, achieves, greater, quality, than, work, effective, stroke, initialization, extended, version, score, distillation, sampling, sds, loss, abstract, render, code, status, accepted, nips, beihang, hong, kong,

Text of the page (random words):
diffsketcher text guided vector sketch synthesis through latent diffusion models diffsketcher text guided vector sketch synthesis through latent diffusion models ximing xing 1 chuang wang 1 haitao zhou 1 jing zhang 1 dong xu 2 qian yu 1 1 beihang university 2 the university of hong kong paper status accepted by nips 2023 paper code demo render tl dr diffsketcher an innovative algorithm that creates vectorized free hand sketches using natural language input abstract even trained mainly on images we discover that pretrained diffusion models show impressive power in guiding sketch synthesis in this paper we present diffsketcher an innovative algorithm that creates vectorized free hand sketches using natural language input diffsketcher is developed based on a pre trained text to image diffusion model it performs the task by directly optimizing a set of bézier curves with an extended version of the score distillation sampling sds loss which allows us to use a raster level diffusion model as a prior for optimizing a parametric vectorized sketch generator furthermore we explore attention maps embedded in the diffusion model for effective stroke initialization to speed up the generation process the generated sketches demonstrate multiple levels of abstraction while maintaining recognizability underlying structure and essential visual details of the subject drawn our experiments show that diffsketcher achieves greater quality than prior work methodology diffsketcher generates free hand sketches from a text prompt such as a photo of eiffel tower to synthesize a sketch that matches the given text prompt we optimize the parameters of the differentiable rasterizer that produces the raster sketch such that the resulting sketch is close to a sample from the frozen latent diffusion model the blue part of the picture since the diffusion model directly predicts the update direction we do not need to backpropagate through the diffusion model the model simply acts like an efficient frozen critic that predicts image space edits demo prompt macaw full color ultra detailed realistic insanely beautiful very detailed masterpiece painting of baby yoda hoding a lightsaber sailboat sailing in the sea on a clear day sketch more results 1 still life 2 cartoon character 3 scene 4 complex scene painterly rendering colorful results oil painting style transfer citation inproceedings xing2023diffsketcher title diffsketcher text guided vector sketch synthesis through latent diffusion models author ximing xing and chuang wang and haitao zhou and jing zhang and qian yu and dong xu booktitle thirty seventh conference on neural information processing systems year 2023 url https openreview net forum id cy1xatveqj copy bibtex bibtex copy success close sdiffsketcher text guided vector sketch synthesis through latent diffusion models this webpage template was originally made by ximing xing