Video Models Wiki
A practical field guide to the best AI video models of 2026. Each page covers what a model does well and the prompt patterns that get results across text-to-video, image-to-video, and reference-to-video.
xAI locked down Grok Imagine's uncensored mode through 2026. This wiki helps Venice users find alternatives that give more control and stay permissive. Seedance 2.0 is the top pick for high-quality mature content. It allows real-face references when the subject gives active consent and passes identity verification.
The models
Kling 3
Kuaishou's cinematic powerhouse. Native 4K, 15s clips, multi-shot "AI Director."
Seedance 2.0
ByteDance's multimodal director. Up to 12 references (image+video+audio), native sound.
Grok Imagine
xAI's fast generator. Was the go-to for uncensored content. Now moderated.
Wan 2.7
Alibaba's open-weight suite. Run it locally with full control, including Spicy/uncensored.
Happy Horse 1.0
Alibaba's #1-ranked open model. Joint video+audio, 7-language lip-sync, fast.
Not sure?
Answer a few questions and jump to the right model for your use case.
Quick comparison
| Model | Max res | Max length | Native audio | Modes | Open weights | Mature content |
|---|---|---|---|---|---|---|
| Kling 3 | 4K | 15s | Yes | T2V · I2V · R2V | No | Filtered |
| Seedance 2.0 | 2K | 15s | Yes | T2V · I2V · R2V · V2V | No | Permissive* |
| Grok Imagine | 1080p | ~6–15s | Yes | T2V · I2V | No | Blocked |
| Wan 2.7 | 1080p | 15s | Yes | T2V · I2V · R2V · edit | Yes | Uncensored** |
| Happy Horse 1.0 | 1080p | 15s | Yes | T2V · I2V · R2V · edit | Yes | Varies |
*Seedance: the official Dreamina app filters explicit content. Suggestive work in the "gray zone" still gets through, and third-party API access allows more. Real faces require consent and identity verification.
**Wan: the base model ships with filters. You can run the open weights with community "Spicy" LoRAs that strip them. See each model page for specifics and the full comparison table.
Start here
- T2V · I2V · R2V explained. Start with the three generation modes.
- Prompting fundamentals. The director formula that works across these models.
- Leaving Grok Imagine. What changed, and where to go.
- NSFW best practices. Tasteful techniques for high-quality adult content.
- Consent & responsible use. Read this before you use a real person's likeness.
Compiled from public documentation, vendor release notes, and the ByteDance Seedance 2.0 prompt guides. These models change fast. Verify current limits and policies before production use. Last researched: 2026.