Video generation is a challenging yet pivotal task in various industries,
such as gaming, e-commerce, and advertising. One significant unresolved aspect
within T2V is the effective visualization of text within generated videos.
Despite the progress achieved in Text-to-Video~(T2V) generation, current
methods still cannot effectively visualize texts in videos directly, as they
mainly focus on summarizing semantic scene information, understanding, and
depicting actions. While recent advances in image-level visual text generation
show promise, transitioning these techniques into the video domain faces
problems, notably in preserving textual fidelity and motion coherence. In this
paper, we propose an innovative approach termed Text-Animator for visual text
video generation. Text-Animator contains a text embedding injection module to
precisely depict the structures of visual text in generated videos. Besides, we
develop a camera control module and a text refinement module to improve the
stability of generated visual text by controlling the camera movement as well
as the motion of visualized text. Quantitative and qualitative experimental
results demonstrate the superiority of our approach to the accuracy of
generated visual text over state-of-the-art video generation methods. The
project page can be found at this https URL

我们提出了一种创新的方法 Text-Animator，用于视觉文本视频生成，通过精确描述生成视频中视觉文本的结构，并通过控制摄像机移动和文本运动来改善生成视觉文本的稳定性，实验证明了我们方法在生成视觉文本准确性上的优越性。

Text-Animator: 可控的视觉文本视频生成

Text-Animator: Controllable Visual Text Video Generation

A common practice in Natural Language Processing (NLP) is to visualize the
text corpus without reading through the entire literature, still grasping the
central idea and key points described. For a long time, researchers focused on
extracting topics from the text and visualizing them based on their relative
significance in the corpus. However, recently, researchers started coming up
with more complex systems that not only expose the topics of the corpus but
also word closely related to the topic to give users a holistic view. These
detailed visualizations spawned research on comparing text corpora based on
their visualization. Topics are often compared to idealize the difference
between corpora. However, to capture greater semantics from different corpora,
researchers have started to compare texts based on the sentiment of the topics
related to the text. Comparing the words carrying the most weightage, we can
get an idea about the important topics for corpus. There are multiple existing
texts comparing methods present that compare topics rather than sentiments but
we feel that focusing on sentiment-carrying words would better compare the two
corpora. Since only sentiments can explain the real feeling of the text and not
just the topic, topics without sentiments are just nouns. We aim to
differentiate the corpus with a focus on sentiment, as opposed to comparing all
the words appearing in the two corpora. The rationale behind this is, that the
two corpora do not many have identical words for side-by-side comparison, so
comparing the sentiment words gives us an idea of how the corpora are appealing
to the emotions of the reader. We can argue that the entropy or the
unexpectedness and divergence of topics should also be of importance and help
us to identify key pivot points and the importance of certain topics in the
corpus alongside relative sentiment.

本篇研究论文探讨了利用自然语言处理技术在文本可视化方面的应用，除了传统的利用文本主题抽取与可视化的方法，更深入地进行了针对单词情感价值的比较，探究不同文本间的差异与重点。