Source: AI风格资料收集清单使用指南 · Category: Interface — AI Image Generation (接口-AI生图) · Updated 2026-09-29 08:42:45
English Translation
AI Style Material Collection Checklist Usage Guide
This guide is written in plain language for operators and agents: it explains how to collect and fill in the material for an AI style so the system can correctly generate the image effect you want.
I. Why Collect and Organize Materials
In short: to have the AI generate images in a specific style (e.g. a person posing with a lion, or checking in at West Lake), you need to tell the AI two things —
- What image(s) to generate from (upload the person's photo, the scene photo)
- What effect to generate (describe it clearly in text)
This checklist is a table that lets you fill in those two things clearly. Once filled, hand it to the administrator to configure; the system can then automatically generate images per your requirements.
II. What to Fill In
The checklist has 6 columns in total. Their meanings are as follows:
| Col | Column Name | Plain Explanation |
|---|---|---|
| A | Style Name | Give the style a name, e.g. "Lion Photo" or "West Lake Check-in" |
| B | Type | Choose "Single Image" or "Dual Image" (explained in detail below) |
| C | Upload Image (Image 1) | The customer/user's portrait photo, labeled "Image 1" |
| D | Scene Image (Image 2) | The background scene photo, labeled "Image 2" (for image-to-image only) |
| E | Prompt | Describe in text what image you want to generate (most important) |
| F | Composite Ratio | Final image aspect ratio; optional 1:1 / 9:16 / 16:9 |
If you don't know how to fill it in, check the "Reference" tab at the bottom of the Excel file — it has filled-in examples.
III. What's the Difference Between Single and Dual Image
This is the most critical concept — it directly determines how many images you need to provide:
| Type | Images Required | Details |
|---|---|---|
| Single Image | Only 1 (Image 1) | Upload only the portrait; no scene image needed. The AI auto-generates the scene from the prompt. |
| Dual Image | 2 (Image 1 + Image 2) | Upload the portrait (Image 1) and the scene photo (Image 2). The AI places the person from Image 1 into the scene of Image 2. |
How to Choose
- Single Image: suited to scenes that need no specific background. E.g. "photo with a lion" — the AI generates the lion and background itself; you only provide the portrait.
- Dual Image: suited to scenes needing a specific check-in location. E.g. "check in at West Lake" — you provide the portrait and the West Lake scene photo, and the AI composites the person into the scene.
Remember: the scene image (Image 2) for the dual-image type is mandatory — without it, generation cannot proceed.
IV. How to Fill Each Item (Item-by-Item)
1. Style Name
Just give the style an easy-to-recognize name, e.g.:
- Lion Photo
- Hangzhou West Lake Check-in
- Beijing Tiananmen Check-in
2. Type
Fill in "Single Image" or "Dual Image". If unsure, see the explanation above.
3. Upload Image (Image 1)
Place the portrait photo. The person in this photo is the person who appears in the final image.
- The photo must be clear, with the face unobstructed
- Label it "Image 1" when filling the form, so it maps to the description in the prompt
4. Scene Image (Image 2)
Place the background scene photo. Only required for the "Dual Image" type; for "Single Image" fill in "None" or leave it blank.
- This photo is the background of the final image
- Label it "Image 2" when filling the form
- Mandatory for dual-image type; cannot be left empty
5. Prompt (Important)
This is the most critical item. Describe clearly in text "what image you want to generate."
How well the prompt is written directly determines the generation quality.
6. Composite Ratio
Choose an image ratio:
| Ratio | Suited For |
|---|---|
| 1:1 | Square — suited to social-media avatars, WeChat Moments |
| 9:16 | Portrait — suited to phone wallpapers, Douyin/Kuaishou covers |
| 16:9 | Landscape — suited to desktop wallpapers, landscape posters |
V. Reference Cases
The two cases below come from the reference sheet; you can study them directly.
Case 1: Lion Photo (Single Image)
| Item | Content |
|---|---|
| Style Name | Lion Photo |
| Type | Single Image |
| Upload Image (Image 1) | Portrait photo |
| Scene Image (Image 2) | None |
| Composite Ratio | 1:1 |
Prompt:
Hyper-realistic photographic style, HD quality, natural lighting. The main subject is the person from Image 1 as the protagonist, with the protagonist's facial light moderately brightened, the protagonist lying prone on a large rock shoulder-to-shoulder with an adult male lion, the lion's fur fluffy and expression gentle. The background is the lion enclosure at Jinniu Lake Wildlife Kingdom, with grass beside the rock, and behind it a white wall, metal fence, and green vegetation — the overall scene real and natural, with clear and delicate details of the person and lion, fur and skin texture, and do not show the text "Jinniu Lake".
Analysis: This is a single-image type, providing only the portrait. The prompt describes in detail what the lion looks like, what the background is, how the lighting is, and which details must be clear. Because no specific scene photo is needed, the scene image is filled as "None".
Case 2: Hangzhou West Lake Check-in (Dual Image)
| Item | Content |
|---|---|
| Style Name | Hangzhou West Lake Check-in |
| Type | Dual Image |
| Upload Image (Image 1) | Portrait photo |
| Scene Image (Image 2) | West Lake scene photo |
| Composite Ratio | 9:16 |
Prompt:
Based on Image 2, strictly replace the person in Image 2 with the person from Image 1, then beautify the person.
Analysis: This is a dual-image type, providing both the portrait and the West Lake scene photo. The prompt is short because the scene is already in Image 2 — you only need to tell the AI "place the person from Image 1 into the scene of Image 2, then beautify".
VI. How to Write the Prompt
The logic of writing a prompt has four key points:
| No. | Key Point | Explanation |
|---|---|---|
| 1 | Upload image = Image 1 | "Image 1" in the prompt refers to the uploaded portrait |
| 2 | Scene image = Image 2 | "Image 2" in the prompt refers to the scene photo (dual-image type) |
| 3 | State clearly what to generate | Describe the final image in text: what the person is doing, what the background is, what style |
| 4 | Support person description | You may describe the person's style, gestures, sitting posture, standing posture, etc. |
How to Write a Single-Image Prompt
A single image has no scene photo, so the prompt must describe the scene too. Refer to the Lion Photo example:
- First state the overall style (hyper-realistic photographic style, HD quality)
- Then the person (the person from Image 1 as protagonist, facial light brightened)
- Then the scene (lying prone with a lion on a rock; background is the enclosure, grass, wall, etc.)
- Finally the detail requirements (clear fur and skin texture, do not show a certain text)
How to Write a Dual-Image Prompt
A dual image has a scene photo, so the prompt can be short — focus on how to composite. Refer to the West Lake example:
- Base it on Image 2
- Replace the person in Image 2 with the person from Image 1
- Add effect requirements (beautify, etc.)
Prompt Tips
- Be as specific as possible: don't say "a nice photo", say "hyper-realistic photographic style, natural lighting, HD quality"
- You can specify what to exclude: e.g. "do not show a certain text"
- You can describe the person's pose: e.g. "person standing smiling" or "person sitting on a bench"
- You can describe lighting and mood: e.g. "soft sunlight" or "night-scene lighting"
VII. Filling Notes
- Image 1 and Image 2 must be clearly labeled: the "Image 1"/"Image 2" referenced in the prompt must correspond to the images uploaded in the form
- Single-image type scene image fill in "None": don't leave it blank; write "None" explicitly
- Dual-image type scene image is mandatory: this is a hard requirement
- The prompt is the top priority: the more detailed and specific, the better the generation
- Photo quality must be high: portrait clear, face unobstructed, normal lighting
- Decide the ratio in advance: choose 1:1, 9:16, or 16:9 based on the final use
VIII. Material Checklist Template
When adding a style, fill in according to the following table:
| Style Name | Type (Single/Dual) | Upload Image (Image 1) | Scene Image (Image 2) | Prompt (Important) | Composite Ratio |
|---|---|---|---|---|---|
| (fill style name) | (Single/Dual) | (portrait photo) | (scene photo / None) | (describe the effect to generate in detail) | (1:1 / 9:16 / 16:9) |
| (fill style name) | (Single/Dual) | (portrait photo) | (scene photo / None) | (describe the effect to generate in detail) | (1:1 / 9:16 / 16:9) |
| (fill style name) | (Single/Dual) | (portrait photo) | (scene photo / None) | (describe the effect to generate in detail) | (1:1 / 9:16 / 16:9) |
Copy this table into Excel to use it. One row is one style; as many rows as you fill, that many styles get configured.
Appendix: FAQ
Q1: How do I tell single from dual image?
It depends on whether you need to provide a specific background photo:
- No specific background needed; let the AI generate the scene → Single Image
- A specific background photo needed (e.g. a certain attraction) → Dual Image
Q2: How long should the prompt be?
No hard limit, but recommended:
- Single Image: write in more detail, 100–300 characters, describing person, scene, style, and details clearly
- Dual Image: can be short, just state how to composite, 30–100 characters is fine
Q3: What are the requirements for the Image 1 portrait?
- High clarity, face unobstructed
- Preferably a front or semi-side view
- Normal lighting, not too dark or overexposed
- Less cluttered background works better
Q4: Where does the Image 2 scene photo come from?
You can take it yourself or find it online. E.g. the West Lake check-in scene photo can be a photo you took at West Lake or an online West Lake scenery shot. Whether the scene photo has people in it doesn't matter — the AI replaces it with the person from Image 1.
Q5: Which composite ratio to choose?
Depends on where the image is used:
- WeChat Moments, avatar, product display → 1:1 (square)
- Phone wallpaper, short-video cover → 9:16 (portrait)
- Desktop wallpaper, landscape poster → 16:9 (landscape)
Q6: Can one style have multiple Image 1 photos?
Yes. After the style is configured, each user uploads their own photo as Image 1 and can generate different-person versions of the same style. This is also the core value of the feature — configure once, use by many.
Q7: Can the prompt say "do not include certain content"?
Yes. E.g. the Lion Photo case wrote "do not show the text 'Jinniu Lake'". Use the phrasing "do not show xxx".
Q8: Who do I hand the filled form to?
After filling, send the entire Excel file to the administrator, who will configure the style in the system backend based on your input.
Q9: Can the style name be anything?
Yes, but it's recommended to use a name that's clear at a glance, for easier later management. E.g. "Lion Photo" is much clearer than "Style 1".
Q10: Must the dual-image prompt write "based on Image 2"?
Not necessarily, but it's recommended to clarify the relationship between Image 1 and Image 2. E.g. "based on Image 2, replace the person in the scene with the person from Image 1" is very clear. Without it, the AI may not know how to handle the relationship between the two images.
If you encounter unsolvable issues while debugging the device, contact the cloud-printing representative for troubleshooting. Contact details: Tel: 137-5177-4454 [WeChat same number] Email: info@ColorPark.cn