Tips for Using SJinn Agent

Welcome to our guide on using SJinn Agent! This document provides practical tips to help you get the most out of SJinn Agent and answer common questions. This guide will be continuously updated, so please stay tuned.

Table of Contents

• [Tips1: How to Communicate Effectively with SJinn Agent]
• [Tips2: When and How to Use Templates]
• [Tips3: About Video Models]
• [Tips4: For New Tasks, Consider Starting a New Chat]

---

Tips1: How to Communicate Effectively with SJinn Agent

Treat it like a creative partner, not just a tool. Clearly expressing your creative intent is crucial.

• What to Avoid: Directly pasting a prompt from another tool (like Midjourney) that only contains an image description. This type of prompt lacks a key instruction: what you want SJinn Agent to do.
• Recommended Approach: Communicate as you would with a partner by giving clear commands. For example: "Please draw a picture of [image description]," or "I want to generate an image that shows [image description]."

Tips2: When and How to Use Templates

What are templates?

In short, templates are pre-configured settings that encapsulate our successful experiences and parameters for specific tasks, providing SJinn Agent with a proven frame of reference.

• Pros: Using a template for a specific task yields more stable and predictable results.
• Cons: If your request does not align with the template's designated task, the pre-set information can confuse SJinn Agent and lead to suboptimal outcomes.

How to Use Templates Correctly

Before using a template, be sure to read its description to understand its intended use case and recommended inputs. This will ensure you achieve the best possible results.

Tips3: About Video Models

We currently support two types of video models:

1. Silent Video Models

This category includes models like Kling, Hailuo, Luma, Seedance, and Runway. While their core function is video generation, each excels at producing different effects and styles.

We have now completed intelligent model selection, which will automatically choose the appropriate model based on your requirements (from Kling, Hailuo, Luma, Seedance).

If you want to manually use a specific type of model, you can go to the tools section, for example: https://sjinn.ai/tools/seedance-lite-image-to-video

2. Video Models with Audio (Veo3)

Due to its current limitations, Veo3 is not the default video model in SJinn Agent. However, you can use it in the following ways:

Explicitly in Your Prompt You can directly instruct SJinn Agent to use Veo3. Here are two successful use cases:

• ASMR Video
• VLog Video

Using Specific Templates Some templates designed for particular tasks will use Veo3 by default, such as:

• Character Speech

Veo3's Advantages and Limitations

Advantages:

1. High realism in scenes.
2. Generates video with synchronized audio.

Limitations:

1. Currently only generates videos in a 16:9 aspect ratio.
2. The voice-over feature only supports English at this time.
3. Vocal timbre may not be consistent across multiple generated clips.

Therefore, if your creative needs conflict with these limitations (e.g., you require a non-16:9 aspect ratio or a non-English voice-over), we recommend that you avoid specifying "Veo3" in your prompts and do not use templates that default to it.

Tips4: For New Tasks, Consider Starting a New Chat

For new tasks that don't require historical information as context, it's recommended to start a new chat window. This can reduce SJinn Agent hallucinations (error probability) and improve response speed. For users who want to create series characters, we are planning a character asset feature. Simply add this asset in a new conversation, and SJinn Agent will be able to understand the character.