SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsmodel-selector

model-selector

Agents & models

View repository

Updated 10-15 17:32

About this skill

model-selector

A "shopping advisor" for 15 mainstream AI video models (11 commercial + 4 open-source). Ask the user 3-4 key questions, then give a recommendation + reasons.

When not to use this skill

  • The user has already decided to use a specific model → use the corresponding <model>-prompter skill
  • The user asks "how do I write the prompt" → use seedance-prompter / kling-prompter / happyhorse-prompter
  • The user's prompt isn't working → use seedance-debugger

11-model capability quick-reference matrix

Model Chinese Audio Physics Editing Duration Open-source Available in China Main battlefield
Seedance 2.0 ★★★ — ★★★ — 15s — ✓ Cinematic narrative
HappyHorse 1.0 ★★★ ★★ ★★ — 15s (default 5s) — ✓ ASMR / short films
Kling 3.0 ★★★★★ ★★★★ ★★★★ ✓ 2m — ✓ Chinese drama / image-to-video
Sora 2 ★★ ★★★★ ★★★★★ — 25s (Pro) — △ Cinematic art films
Veo 3.1 ★★ ★★★★★ ★★★ — 148s chained — △ Multi-person dialogue
Gemini Omni ⭐NEW ★★ ★★★ ★★★ ★★★★★ TBD — △ Iterative editing / text rendering / cross-modal
Runway Gen-4 ★★★ — ★★★ ★★★★★ 30s — ✓ Video editing
Pika 2.5 ★★ ★ ★★ — 25s — ✓ Creative effects
Hailuo 02 ★★★ — ★★★★★ — 10s — ✓ Physical actions
Hunyuan 1.5 ★★★★ — ★★★ ✓ 10s ★★★★★ ✓ Open-source / LoRA
Wan 2.7 ★★★ ★★★★★ ★★★ ✓ 15s — ✓ Digital human lip-sync
Jimeng 3.0 ★★★★★ ★★ ★★★ — 15s — ✓ Chinese + CapCut integration

Workflow

Step 1: Ask 3-4 key questions

Don't ask all of them; pick the 3-4 that most affect the outcome:

Dimension Question
Duration "How long does it need to be? 5 seconds / 10 seconds / over 15 seconds?"
Audio "Do you need native audio? Just ambient sound / dialogue / BGM / none"
Language "Is it a Chinese-language scene or an English-language scene?"
Nature of editing "Brand-new generation or editing an existing video?"
Platform restrictions "Available in China / available internationally / local deployment?"
Use case "Short-video hit / film-grade / commercial ad / personal creation?"

Step 2: Recommend according to the decision tree

是编辑已有视频？
├── Yes → Runway Aleph（独家）
└── No → 继续
    ├── 要数字人/真人对口型？
    │   └── Yes → Wan 2.7（业界最准）
    ├── 要多人对话+精确配乐？
    │   └── Yes → Veo 3.1
    ├── 中文剧情/中式美学？
    │   └── Yes → 即梦 3.0（创作者） / Kling 3.0（开发者 API）
    ├── 极致物理感（水/火/动作）？
    │   └── Yes → Hailuo 02 > Sora 2
    ├── 创意爆款特效（变身/融化）？
    │   └── Yes → Pika 2.5（Pikaffects）
    ├── 本地部署/开源/LoRA 训练？
    │   └── Yes → Hunyuan 1.5
    ├── 极致电影质感/艺术片？
    │   └── Yes → Sora 2 > Runway Gen-4
    ├── 5/10/15 秒紧凑短片 + 环境音？
    │   └── Yes → HappyHorse 1.0
    └── 复杂多镜头叙事？
        └── Yes → Seedance 2.0 / Kling 智能分镜

Step 3: Output format

## 推荐模型

**最佳选择**：[模型名]
**理由**：[2-3 句话说明为什么]
**对应 skill**：[skills/xxx-prompter] （如果有）
**对应方法论**：[methodology/NN-xxx-公式.md]
**官方平台**：[链接]

**备选**：[模型名]
**为什么备选**：[一句话]

## 你的需求 → 我的判断

| 你的需求 | 对应能力 | 评级 |
|---|---|---|
| [需求 1] | [模型支持情况] | ★★★★★ |
| ... | ... | ... |

Practical examples

Example 1: The user asks "I want to make a Chinese short drama"

Clarifying questions：

  1. Short drama duration? A clip under 10 seconds or a complete 1-2 minute story?
  2. Do you need native audio (character dialogue)?
  3. Do you lean toward realistic / ancient-style / modern urban?

User's answer: A single 10-second clip, with dialogue, modern urban.

Output：

Best choice：Kling 3.0 Reason: Chinese comprehension is tied for first in the industry (with Jimeng), native audio supports character-directed speech, up to 15 seconds per single segment, and the 5-layer advanced formula is naturally suited to narrative.

Corresponding skill：skills/kling-prompter Corresponding methodology：methodology/09-kling-公式.md Official platform：klingai.com

Alternative：Wan 2.7(If there are many dialogue scenes and you need a strong digital human feel) / Jimeng 3.0(If doing post-production directly in CapCut)

Example 2: User asks "I have a video and want to change its style, like turn it into a watercolor animation"

Judgment: Editing an existing video → directly recommend Runway Aleph.

Output：

Best choice：Runway Aleph Reason: Editing existing videos is Aleph's exclusive capability (the other 10 models cannot do it).re-style The verb is specifically used for style transfer, and can keep the original motion and temporal sequence unchanged.

Example prompt：

Re-style this clip as a hand-painted watercolor animation. Soft brush textures, flowing edges, warm palette. Preserve all motion and timing exactly.

Corresponding methodology：methodology/13-六大模型公式速查.md#1-runway-gen-4--aleph Official platform：Runway

Alternative: There is no real substitute. If absolutely necessary, Hunyuan Video can also make some changes (open source), but it is not a native editing capability.

Example 3: User asks "I want to build a local video generation setup myself, and my budget is limited"

Clarifying questions：

  1. What is the GPU configuration? (24GB / 12GB / 8GB)
  2. Main use? (Personal play / self-training LoRA / commercial)

User answer: RTX 4090 24GB, wants to self-train LoRA.

Output：

Best choice：Hunyuan Video 1.5(8.3B version) Reason: Strongest open source, can run on 24GB VRAM, natively supports LoRA training, active community (full ComfyUI workflow support).

Recommended parameters：Q5_K_M GGUF + 30 steps + dpmpp_2m_sde + sgm_uniform + cfg 6.0 Official resources：

Alternative: CogVideoX (Zhipu open source) / LTX Video (Lightricks real-time) — these two are not included in this repository's main directory, but they belong to the same open-source camp.

Reference