Recent diffusion and autoregressive models have substantially improved text-to-video generation, yet producing coherent long-form story videos with consistent characters and aligned audio remains challenging. Existing methods often suffer from narrative drift, unstable character identity, weak cross-scene continuity, and audio-visual mismatch over extended sequences. We propose DramaAgent, a hierarchical, agentic, and model-agnostic framework for long-form text-to-video-and-audio generation. Rather than improving the underlying video backbone itself, DramaAgent introduces an upper-level control layer that decomposes generation into story planning, persistent character conditioning, scene-wise synthesis, and reflection-guided targeted repair. The framework maintains reusable story and character states across scenes, diagnoses failures such as identity drift, missing scene semantics, temporal discontinuity, and cross-modal mismatch, and repairs problematic clips in a stage-specific manner. Experiments across multiple video generation backbones show that DramaAgent improves long-horizon coherence, character consistency, narrative fidelity, and scene-level audio-visual consistency over direct generation and strong baselines. These results suggest that hierarchical agentic control is a practical direction for controllable long-form audiovisual generation. Code: https://github.com/AIGeeksGroup/DramaAgent. Website: https://aigeeksgroup.github.io/DramaAgent.
Figures & tables
Figure 2: Overview of DramaAgent. DramaAgent decomposes a narrative prompt into scene-level targets, constructs persistent character stills, and coordinates image, video, and audio generators through scene-wise generation and reflection-guided repair.
Method
CLIP ↑
Inception ↑
VBench Metrics /% ↑
Subject Cons.
Bg Cons.
Motion Smth.
Dyn. Degree
Aesthetic
Script Synopsis to Keyframe/StoryBoard Generation
StoryDiffusion Zhou et al. (2024)
20.46
6.24
-
-
-
-
-
AutoStory Wang et al. (2025)
20.21
6.01
-
-
-
-
-
MovieAgent Weijia Wu (2025)
22.12
7.23
-
-
-
-
-
DramaAgent (Ours)
23.02
7.64
-
-
-
-
-
Table 1: Main comparison with recent text-to-video generation methods. Results are reported for script synopsis to keyframe/storyboard generation and script synopsis to full movie generation. DramaAgent achieves strong overall performance across both settings.
Figure 3: Qualitative results across three narrative scenes. DramaAgent maintains recurring characters and coherent scene progression across different story contexts. The examples illustrate the role of structured planning, persistent conditioning, and reflection-guided repair in long-form storytelling.
Method
CLIP ↑
Inception ↑
VBench Metrics / % ↑
Subject Cons.
Bg Cons.
Motion Smth.
Dyn. Degree
Aesthetic
Direct Generation (Seedance)
22.90
8.65
92.10
95.60
97.10
64.20
62.30
+ Story Planning
23.55
8.95
93.80
96.30
97.45
65.10
62.90
+ Character Stills
24.12
9.41
96.85
97.33
97.96
66.84
64.37
+ Reflective Repair (Full)
24.34
9.67
97.02
97.54
98.14
67.32
64.91
Table 2: Component ablation on a fixed backbone. Progressively adding story planning, character stills, and reflective repair on Seedance.
Figure 4: Qualitative comparison across four video generation backbones. We compare Hailuo, Keling, Jimeng, and SeedDance under the same narrative prompt and character references, highlighting their differences in identity consistency, temporal stability, and scene realism.
Method
CLIP ↑
Inception ↑
VBench Metrics / % ↑
Subject Cons.
Bg Cons.
Motion Smth.
Dyn. Degree
Aesthetic
DramaAgent (Wan2.2-I2V-Flash)
23.47
9.56
94.12
93.95
96.92
70.15
54.32
DramaAgent (Keling)
23.78
9.35
95.64
95.82
97.43
66.92
58.14
DramaAgent (Jimeng)
23.96
9.38
96.42
96.71
98.02
80.45
61.27
DramaAgent (Hailuo 3.0)
22.91
9.12
96.34
97.27
98.61
84.31
61.32
DramaAgent (Seedance)
24.12
9.41
96.85
97.33
97.96
66.84
64.37
Table 3: Effect of backbone choice. DramaAgent is evaluated with five video generation backbones under the same control framework.
Method
CLIP ↑
Inception ↑
VBench Metrics / % ↑
Subject Cons.
Bg Cons.
Motion Smth.
Dyn. Degree
Aesthetic
Script Synopsis to Keyframe/Storyboard Generation
DramaAgent w/o Reflection
22.05
7.64
-
-
-
-
-
DramaAgent w/ Reflection
23.02
7.64
-
-
-
-
-
Script Synopsis to Movie Generation
DramaAgent w/o Reflection
24.12
9.41
96.85
97.33
97.96
66.84
64.37
Table 4: Effect of reflective refinement. Comparison with and without reflection in storyboard and full movie generation.
Appendix figures & tables8 assets
Supplementary material from the paper’s appendix.
Appendix
Backbone
Version / API Variant
Access Date
Input Type
Candidates / Scene
Repair Rounds
Clip Duration
Resolution
Wan2.2-I2V-Flash
Wan2.2-I2V-Flash
2025-11
image + text
3
2
5s
1280 × 720
Keling
Keling
2025-11
image + text
3
2
5s
1280 × 720
Jimeng
Jimeng
2025-11
text-only / multimodal
3
2
5s
1280 × 720
Hailuo 3.0
Hailuo 3.0
2025-11
image + text
3
2
5s
1280 × 720
Seeddance 3.0
Seeddance 3.0
2025-11
image + text
3
2
5s
1280 × 720
Appendix
Table 5: Reproducibility settings for video generation backbones. We report the backbone/API variant, access date, supported conditioning interface, and scene-level generation budget used in our experiments. All DramaAgent variants share the same upper-level planning, persistent character conditioning, and reflective repair framework; only the underlying video generator changes across rows.
Statistic
Value
Average initial generation time / scene
75 s
Average repaired generation time / scene
28 s
Average end-to-end time / 30–60s script
18–32 min
Average API calls / scene
4.1
Approximate cost / minute of final video
$7.8
Appendix
Table 6: Runtime and cost statistics under the default setting. Reported values are averages over the evaluation set and may vary with backbone API latency and access date.
Parameter
Value
λid
0.35
λsem
0.30
λtemp
0.20
λav
0.15
Acceptance threshold τ
0.80
Early-stop threshold ϵ
0.02
Appendix
Table 7: Default reflection settings. We use the following default hyperparameters for candidate scoring and selective reflective repair.
Metric
DramaAgent Preference
Temporal Coherence
66.7%
Character Consistency
71.3%
Narrative Fidelity
63.3%
Audio-Visual Alignment
68.0%
Inter-rater Agreement (Fleiss’ Kappa)
0.42
Appendix
Table 8: Human evaluation results. DramaAgent is preferred over baseline pipelines across all four evaluation dimensions.
Figure 5: Additional qualitative case: animated family-adventure narrative. DramaAgent preserves recurring character identity, expressive interaction, and multi-scene story continuity in a stylized family-oriented animated story.
Figure 6: Additional qualitative case: action-crime ensemble narrative. The example illustrates multi-character coordination, dynamic scene progression, and stable identity preservation in a high-motion cinematic setting.
Figure 7: Additional qualitative case: fantasy mystery narrative. DramaAgent supports atmosphere-heavy story generation with coherent clue progression, recurring character consistency, and visually stable magical world-building.
Figure 8: Additional qualitative case: mythic fantasy narrative. The generated story maintains stylized visual design and coherent narrative development across ritual, conflict, resolve, and farewell scenes.