AI Creation Guide

Guide Overview

This guide is organized into three parts and covers the full Vinabot AI workflow, from generating images and videos with prompts and reference images to creating interactive AI characters that can hold real-time conversations.

Contents

1. What AI Creation Can Do
2. Where to Start
3. Choose a Generation Model
4. Adjust Generation Settings
5. Write Effective Prompts
6. Use Reference Images
7. Use @Image for Precise Multi-Image Control
8. Photo Avatar: Make a Photo Speak
9. Processing Time and Results
10. AI Companion: Create a Real-Time Conversational Persona
11. AI Credits: Usage and Top-Ups
12. Troubleshooting

---

Part 1 | AI Creation

Use prompts and reference images to create images, videos, and talking-photo content.

1. What Can AI Creation Do?

Vinabot turns ideas into images and videos without requiring design skills. Every generated item is saved in Assets, where it can be previewed, downloaded, sent to a frame, or used as source material for an AI Companion.

Tip:
Generated images can be used to create AI Companions. Generated videos can be played directly on a Vinabot frame.

---

2. Where to Start

You can begin a creation in either of these ways:

---

3. Choose a Generation Model

Models differ in visual style, capability, and AI credit cost. Each model card shows its name, description. Choose a model that matches your goal, such as realistic photography, illustration, anime, or cinematic video.

---

4. Adjust Generation Settings

Available settings depend on the selected model. Configure the following options before generating:

Tip:
After you select a model, the app automatically displays the values supported by that model.

---

5. Write Effective Prompts

The more specific your description is, the more predictable the result will be. Use this formula:

Example: A young woman standing beneath cherry blossoms, fresh Japanese editorial style, sunlight filtering through the leaves onto her face, soft warm lighting.

• Be specific. “A golden retriever puppy sitting on a wooden porch at golden hour with a softly blurred background” is stronger than “a dog.”
• Name the style: realistic, oil painting, watercolor, cinematic, anime, or 3D render.
• Describe lighting and mood. Warm or cool, bright or dramatic, and calm or energetic choices strongly affect the result.
• Read the model description first; some models specialize in particular visual styles.

---

6. Use Reference Images

Reference images let the model see the subject, composition, or style you want. Upload from Device Photos or select an existing image from Assets.

After uploading, you can replace an image by selecting another file from Device Photos or Assets, or remove it from the project.

---

7. Use @Image for Precise Multi-Image Control

When several reference images are uploaded, @Image lets you define exactly how each image should be used in the final result.

13. Upload multiple images. The app labels them Image1, Image2, Image3, and so on.
14. Type @ in the description field to open the image-reference menu.
15. Select the image you want to reference, then continue writing the instruction.

Example: @Image1 is sitting in the corner of a cafe. Sunlight falls across @Image2 on the table. @Image1 picks up @Image2 and takes a sip.

Tip:
This method is ideal when a specific person must appear in a specific setting or interact with a particular object.

---

8. Photo Avatar: Make a Photo Speak

Photo Avatar combines a still portrait with a voice script to generate a video in which the person appears to speak. It works well for product introductions, welcome messages, and meaningful memory videos.

Entry: Video Generation > Photo Avatar tab.

Step 1: Upload a Portrait

Tap Upload Image and choose a photo from Device Photos or Assets. For best results, use a clear, front-facing, unobstructed upper-body portrait with even lighting and a simple background.

Step 2: Enter the Voice Script

Write the words the person should say. Use clear, natural, conversational language; avoid rare characters and unnecessary symbols; and treat the script as a short spoken monologue.

Step 3: Choose a Voice

Select an Official Voice or one of My Voices. Preview the voice, then adjust volume and speaking speed if those controls are available.

Important:
The image, voice script, and selected voice are all required.

Step 4: Generate

Tap Generate after all required fields are complete. If your balance is insufficient, the app will ask you to add credits. Otherwise, the required credits are deducted and processing begins. The finished video is saved automatically under Assets > Videos, where it can be previewed, sent to a frame, or downloaded.

---

9. Processing Time and Results

AI generation usually finishes within three minutes. You may return to Home or browse other areas of the app while processing continues. The result appears automatically in Assets when complete; you do not need to remain on the generation page.

Tip:
If processing takes significantly longer than three minutes, see Section 12, Troubleshooting.

---

Part 2 | AI Companions

Upload a portrait and create a digital persona that can hold real-time conversations.

10. AI Companion: Create a Real-Time Conversational Characters

AI Companion is a core Vinabot feature. A character created from your portrait can speak naturally, display facial motion, use an official or cloned voice, follow a defined personality, reference uploaded documents, and respond to a custom wake word.

• Real-time voice conversation with natural responses
• Facial expressions and motion such as blinking, nodding, smiling, and lip synchronization
• Official voices or a compatible cloned voice
• A custom background, personality, knowledge scope, and speaking style
• Answers grounded in uploaded Knowledge Base files
• Hands-free activation with a custom wake word

You may create a character based on a real family member, fictional persona, historical figure, or business assistant, provided that you own or are authorized to use the reference image.

Step 1: Create a Character

Start from Character on the Home screen or from Assets > Characters, then complete the portrait, name, profile, and voice fields.

After uploading, you can optionally select a preset image style to influence the character’s overall appearance.

Step 2: Character Profile

The profile defines who the character is and directly affects speaking style, knowledge scope, and conversational behavior. The maximum length is 5,000 characters. For consistent results, describe:

Example: She is a 60-year-old retired English teacher who speaks gently and in an organized way. She taught for 30 years, loves classical literature, and occasionally uses idioms and lines of poetry. She often addresses the listener as “my dear” or “young friend,” creating a warm and reassuring presence.

Step 3: Voice and Name

Every character requires a voice. Choose an Official Voice or a compatible voice from My Voices. Preview the selection and adjust volume or speaking speed where supported.

Important:
A character without a configured voice cannot be sent to a device.

Character names may contain up to 20 characters. Short, natural names such as Linda, Dad, or Grandma work best.

Step 4: Advanced Settings

Conversation Language: Configure the language the character can understand and speak. An interactive frame can use one conversation language at a time; switching languages within a single device session is not currently supported.

Wake Word: A spoken keyword that activates a character without touching the screen. Use 4-6 Chinese characters or 2-4 English words. Choose words that are clear, easy to pronounce, and unlikely to be confused with everyday speech.

Preset Q&A: Define fixed question-and-answer pairs. When a question matches a preset, Vinabot returns the approved answer instead of generating a new one. This improves consistency and is especially useful for business information that must remain accurate.

Knowledge Base: Upload one or more supported documents. Vinabot parses the files and prioritizes relevant Knowledge Base content when answering related questions. Use it for product manuals, company information, training material, personal stories, or specialist references.

Step 5: Edit and Manage Characters

You can edit a character after creation. Editing does not consume AI credits, so you can refine the name, profile, voice, and advanced settings whenever needed. Under Assets > Characters, use filters to find characters by status or category. Tap Manage to enter batch mode and send or delete multiple characters. Characters still being generated cannot be included in batch actions.

Step 6: Talk with a Character

Vinabot uses AI language models to understand spoken questions and generate natural responses in real time. Each response considers the current user input, the Character Profile, Knowledge Base content, matched Preset Q&A, and the current-session conversation context.

Conversation Memory: Vinabot does not retain conversation history across separate sessions. Use a detailed Character Profile, Knowledge Base, and Preset Q&A to maintain a consistent experience without long-term memory.

Real-Time Information: When an eligible online service is available, Vinabot can answer topics such as weather, news, and general knowledge. Availability and freshness depend on the current AI service.

Multilingual Support: The app can store up to five configured languages. The interactive frame uses one conversation language at a time. In-session language switching on the device is planned for a future version.

---

Part 3 | AI Credits and Support

11. AI Credits: Usage and Top-Ups

The credits required for each generation depend on the selected model, resolution, duration, and other parameters. The app displays the required amount on the model card before generation begins.

Monthly AI credits can be used across all supported generation types. Higher resolution, longer video duration, and more complex generation settings generally consume more credits.

---

12. Troubleshooting

If failures continue, contact support@vinabot.ai and include the device information, a description of the steps taken, and screenshots of the error. These details help the support team investigate more quickly.