Hlo
PixAI Mio.2
Hello there! I'm Mio, PixAI's mascot girl~ Nice to meet you! ✨
How can I help you today? Are you looking to generate some images, or do you have questions about characters or prompts? I'm here to help bring your ideas to life~
Prompt
Plz tell me why pixai don’t following camera angle
PixAI Mio.2
Oh, that's a really common frustration! Let me explain what's going on~
**Why camera angles can be tricky:**
The AI model behind PixAI (Stable Diffusion-based) learns from millions of images, but camera angle is actually one of the harder things for it to understand precisely. Here's why:
1. **Angle tags are inconsistent in training data** — not every image is labeled with "low angle shot" or "bird's eye view", so the model doesn't always associate those words with the right perspective
2. **It prioritizes the subject over the composition** — the model really wants to show faces and bodies clearly, so if you ask for a dramatic high angle, it might "cheat" and show the face anyway
3. **Complex 3D understanding is hard** — truly understanding perspective requires 3D spatial reasoning, which is tough for 2D image diffusion models
**Tips to get better camera angles:**
- **Use specific photography terms**: "low angle shot from below", "bird's eye view", "dutch angle", "extreme close-up"
- **Emphasize with repetition**: "from above, high angle, looking down, top-down view"
- **Describe what you should see**: "looking up at her from ground level, sky visible behind her"
- **Mention camera lens**: "shot with wide-angle lens", "fisheye perspective"
- **Add weight to the angle**: Some people use (from below:1.3) style emphasis
Do you have a specific scene you're struggling with? I can help craft a prompt that might work better! What angle are you trying to achieve?
Prompt
Far click image and character show small in frame but pixai don’t following camera shoot it always ignore 😭😭
PixAI Mio.2
Oh no, that's such a frustrating one! 😭 The AI really loves to zoom in on faces and ignore "far shot" requests. It's like it has a mind of its own!
**Why this happens:**
The model is *obsessed* with showing faces clearly — if the character is tiny in the frame, the face becomes blurry/unrecognizable, so the AI "cheats" and makes them bigger!
**Here are tricks that actually work:**
**1. Describe the ENVIRONMENT first, character second**
```
a vast misty mountain landscape, a small lone figure standing on a cliff edge in the distance, tiny silhouette, atmospheric perspective, wide angle landscape shot, the character is very far away and small
```
**2. Use "silhouette" or "distant figure"**
- "silhouette of a girl" / "distant figure" / "tiny person"
- These imply "small in frame" better than character names
**3. Emphasize the scale difference**
- "character is only 10% of frame"
- "character visible as a small dot in the distance"
- "vast empty space, tiny figure in corner"
**4. Scene-first composition**
```
epic fantasy battlefield, smoke and ruins stretching for miles, a small knight visible far in the background, aerial establishing shot, cinematic wide shot
```
**5. Try landscape aspect ratio**
- Use `width: 1536, height: 768` (landscape) — this physically forces more empty space
Want me to generate an example with a character small in a big scene? What kind of setting did you have in mind? I can show you how the prompt changes things! ✨
Prompt
Plz tell me how can I solve this problem permanently plz
PixAI Mio.2
Oh, I wish I could give you a magic permanent fix, but I need to be honest with you... 😢
**There's no permanent solution because this is baked into the AI itself.** The model is fundamentally trained to prioritize faces and clear subjects — it's not a bug, it's how diffusion models work. They "want" to show recognizable features.
**BUT here are the closest things to "permanent workarounds":**
**1. Two-Step Method (Most Reliable)**
- Generate the BACKGROUND/SCENE first without the character
- Then use inpainting to add a tiny figure afterward
- This bypasses the AI's "face obsession" entirely
**2. Use Very Wide Aspect Ratios**
- `width: 2048, height: 768` (super wide)
- Forces empty space physically
- Character name goes LAST in prompt with low emphasis
**3. Avoid Character Names Entirely**
- Use "silhouette," "distant figure," "tiny person"
- The moment you use a specific character name, the AI zooms in
**4. ControlNet / Pose Reference**
- If you have a specific composition in mind, a sketch/scribble helps lock the camera angle
**5. Generate Background Only**
- Make your wide landscape first
- Add character in Photoshop/other tools (honestly faster sometimes!)
I'm sorry I can't promise a perfect fix — even pro AI artists battle this! Would you like me to try generating a super-wide scene with a tiny figure to show you what works best? What scene did you have in mind? 🥺
PixAI Mio.2