Good concept
Good concept, but character consistency and instruction-following need major improvement
I have been using LongStories.ai to create longer AI story videos, including children's stories and talking-character videos. The platform has a very good concept and one major advantage: it can take a script and create a complete long-form video with characters, narration, voices, scenes, animation and lip-sync in one place. This saves a considerable amount of time compared with creating many individual AI clips and combining them manually.
However, after producing several videos, I experienced significant problems with character consistency, scene continuity and understanding even fairly clear instructions.
For example, characters sometimes disappeared when another character was speaking, character positions changed between scenes, objects were placed incorrectly, and some generated scenes did not logically match the story. In one royal story, a character appeared with an unnatural duplicate/“ghost” face, a crown was oriented incorrectly when the character was viewed from behind, and a coronation was initially staged like a private interaction instead of the public royal ceremony requested.
The biggest frustration came when trying to correct these problems. Sometimes regenerating or editing a scene fixed the original issue but introduced a completely different problem. In another scene, I corrected the visual successfully, but when I re-animated it, an English-speaking character unexpectedly spoke in another language. Repeating the generation with explicit instructions to use English did not solve it.
This means users can spend additional credits trying to repair AI mistakes without knowing whether the regeneration will actually improve the scene.
What I like: LongStories.ai makes long-form AI video creation remarkably convenient. The interface brings script, characters, voices, animation and scenes together, and it can produce attractive visuals without requiring professional video-editing knowledge.
What needs improvement: The platform needs much stronger character locking, spatial and scene continuity, prompt understanding, dialogue/language control, and especially reliable scene-level editing. When a user asks to change one specific thing, the system should preserve everything else instead of potentially redesigning the scene.
I believe LongStories.ai has considerable potential, and I hope the developers continue improving it. But users considering the platform—especially those who require consistent recurring characters and precise storytelling across many scenes—should understand these limitations before purchasing a large credit package.
My suggestion to the developers: Focus less on adding more features for now and more on making the AI understand context across the entire story. A long-form storytelling platform needs to remember who each character is, where they are standing, what they are wearing, which direction they are facing, who is speaking, and what happened in the previous scene.
If these issues are substantially improved, LongStories.ai could become a very strong long-form AI storytelling platform.
17 août 2026
Avis spontané