User Guide
AI Models
Complete reference of all AI models available in OpenStory
OpenStory integrates with a wide range of AI models across four categories: script analysis, image generation, motion/video generation, and music/audio generation. All media models are accessed via Fal.ai, while script analysis uses OpenRouter.
Script Analysis Models
These LLM models analyze your script, extract scenes, characters, and locations, and generate prompts. You can select multiple models to generate parallel sequences for comparison. Next to Generate, Turbo is the default (Luna, Nano Banana 2 Lite, MiniMax H3 Max, ElevenLabs). Quality selects the quality-ranked defaults. Both modes show the full catalog, grouped Fast / Quality.
| Model | Vendor | Context Window | License |
|---|---|---|---|
| GPT-5.6 Luna | OpenAI | 1M tokens | Proprietary (default) |
| Claude Fable 5 | Anthropic | 1M tokens | Proprietary |
| Claude Opus 5 | Anthropic | 1M tokens | Proprietary |
| Claude Opus 5 Fast | Anthropic | 1M tokens | Proprietary (scene-split) |
| Gemini 3.7 Flash | 1M tokens | Proprietary | |
| Gemini 3.1 Pro | 1M tokens | Proprietary | |
| GPT-5.6 Sol | OpenAI | 1M tokens | Proprietary |
| GLM-5.3 Flash | Z.ai | 1M tokens | Open Weight (MIT) |
| GPT-5.6 Terra | OpenAI | 1M tokens | Proprietary |
| DeepSeek V4 Pro | DeepSeek | 1M tokens | Open Weight (MIT) |
| Claude Sonnet 5 | Anthropic | 1M tokens | Proprietary |
| Grok 4.6 | SpaceXAI | 500K tokens | Proprietary |
| Mistral Small 4 | Mistral | 262K tokens | Open Weight (Apache 2.0) |
| Seed 2.0 Mini | ByteDance | 262K tokens | Proprietary |
Image Generation Models
These models create the visual images for each scene. You can select multiple models to generate variant images for comparison.
| Model | Vendor | License | Notes |
|---|---|---|---|
| Nano Banana 2 Lite | Proprietary | Turbo default — fastest Google tier, references, fixed 1K | |
| GPT Image 2 | OpenAI | Proprietary | Quality default — text rendering, UI fidelity, up to 4K |
| Nano Banana 2 | Proprietary | Fast generation and editing | |
| Nano Banana Pro | Proprietary | Enhanced realism and typography | |
| Grok Imagine Image 2.0 | SpaceXAI | Proprietary | Newest Imagine image model, 1K/2K, edit up to 3 refs |
| Grok Imagine Image Quality | SpaceXAI | Proprietary | Quality Mode — higher fidelity, stronger text |
| FLUX.2 Max | Black Forest Labs | Proprietary | Exceptional realism |
| Phota | Phota | Proprietary | Character consistency via profiles |
| Hunyuan Image v3 | Tencent | Open Weight | Strong composition |
| FLUX.2 Dev | Black Forest Labs | Open Weight | 32B open weights with native editing |
| Qwen Image 2 Pro | Alibaba | Open Weight (Apache 2.0) | Native 2K, text rendering |
| HiDream I1 | HiDream | Open Weight (MIT) | 17B parameters |
| Seedream 5.0 Pro | ByteDance | Proprietary | Flagship generation and editing |
| FLUX.2 Flash | Black Forest Labs | Open Weight | Cheapest distilled FLUX.2 — sub-second, edit up to 4 refs |
| FLUX.2 Turbo | Black Forest Labs | Open Weight | Distilled FLUX.2 — ~2s, edit up to 4 refs |
Edit Endpoints
Most image models support reference image editing via dedicated edit endpoints. This allows the AI to use character and location reference images when generating scenes, improving visual consistency.
Motion/Video Models
These models animate still images into video clips.
| Model | Vendor | Est. Time | License | Notes |
|---|---|---|---|---|
| MiniMax H3 Max | MiniMax | ~10s | Proprietary | Turbo default; native audio |
| Seedance 2.0 | ByteDance | ~3.5 min | Proprietary | Quality default; native audio |
| Grok Imagine Video 1.5 | SpaceXAI | ~30s | Proprietary | Highest quality ranking |
| LTX 2.3 Pro | Lightricks | ~2 min | Open Weight | |
| Veo 3.1 | ~2.5 min | Proprietary | 20K max prompt length | |
| MiniMax Hailuo 2.3 | MiniMax | ~3 min | Proprietary | In the Turbo picker; Seedance-class latency |
| Kling v3 Pro | Kling | ~5 min | Proprietary |
Aspect Ratio Compatibility
Not all motion models support all aspect ratios. OpenStory automatically filters to show only compatible models and will switch to a compatible default if your current model doesn't support the selected ratio.
Audio Support
Some motion models can generate audio alongside video. OpenStory checks each model's capabilities to determine audio support.
Music & Audio Models
| Model | Vendor | Max Duration | Type | License |
|---|---|---|---|---|
| ElevenLabs Music | ElevenLabs | 600s (10 min) | Music | Proprietary (default) |
| ACE-Step 1.5 | ACE Studio | 600s (10 min) | Music | Open Weight |
| ACE-Step | ACE Studio | 240s (4 min) | Music | Open Weight |
Music vs. Sound Effects
Music models generate background music tracks from text prompts and optional tags.
Capabilities
| Feature | ElevenLabs Music | ACE-Step 1.5 | ACE-Step |
|---|---|---|---|
| Prompt-based | Yes | Yes | Yes |
| Lyrics support | No | Yes | Yes |
| Instrumental | Yes | Yes | Yes |
| Long-form | Yes (10 min) | Yes (10 min) | Yes (4 min) |