Lights, Camera, Consistency: A Multistage Pipeline for Character-Stable AI Video Stories
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917156327784448 |
|---|---|
| author | Jain, Chayan Sharma, Rishant Garg, Archit Bhanuka, Ishan Narang, Pratik Kumar, Dhruv |
| author_facet | Jain, Chayan Sharma, Rishant Garg, Archit Bhanuka, Ishan Narang, Pratik Kumar, Dhruv |
| contents | Generating long, cohesive video stories with consistent characters is a significant challenge for current text-to-video AI. We introduce a method that approaches video generation in a filmmaker-like manner. Instead of creating a video in one step, our proposed pipeline first uses a large language model to generate a detailed production script. This script guides a text-to-image model in creating consistent visuals for each character, which then serve as anchors for a video generation model to synthesize each scene individually. Our baseline comparisons validate the necessity of this multi-stage decomposition; specifically, we observe that removing the visual anchoring mechanism results in a catastrophic drop in character consistency scores (from 7.99 to 0.55), confirming that visual priors are essential for identity preservation. Furthermore, we analyze cultural disparities in current models, revealing distinct biases in subject consistency and dynamic degree between Indian vs Western-themed generations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_16954 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Lights, Camera, Consistency: A Multistage Pipeline for Character-Stable AI Video Stories Jain, Chayan Sharma, Rishant Garg, Archit Bhanuka, Ishan Narang, Pratik Kumar, Dhruv Computer Vision and Pattern Recognition Artificial Intelligence Generating long, cohesive video stories with consistent characters is a significant challenge for current text-to-video AI. We introduce a method that approaches video generation in a filmmaker-like manner. Instead of creating a video in one step, our proposed pipeline first uses a large language model to generate a detailed production script. This script guides a text-to-image model in creating consistent visuals for each character, which then serve as anchors for a video generation model to synthesize each scene individually. Our baseline comparisons validate the necessity of this multi-stage decomposition; specifically, we observe that removing the visual anchoring mechanism results in a catastrophic drop in character consistency scores (from 7.99 to 0.55), confirming that visual priors are essential for identity preservation. Furthermore, we analyze cultural disparities in current models, revealing distinct biases in subject consistency and dynamic degree between Indian vs Western-themed generations. |
| title | Lights, Camera, Consistency: A Multistage Pipeline for Character-Stable AI Video Stories |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2512.16954 |