Creating Consistent AI Characters with WaveSpeed.ai
WaveSpeed.ai is a fast, cost-effective platform that gives access to advanced AI models for generating and editing images/videos. Most steps happen directly on wavespeed.ai (sign up at wavespeed.ai), with occasional help from free chatbots (like ChatGPT) for prompt crafting.
Step 1
Decide on Your Character Type
I recommend two simple methods for creating your main reference image:
Method 1 — Use an Existing Photo
Find a photo of a person you like. Search on platforms like Pinterest or Instagram or anywhere else you prefer. Choose a face or person that matches the overall vibe you're going for and download the image.
Use ChatGPT (or any other chatbot) to generate a prompt based on that image → upload to ChatGPT → ask: "Generate a highly detailed prompt describing only this person's face, hair, body type, and features from the uploaded image."
If you want you can easily edit the prompt asking ChatGPT to add scars or skin pigmentation, change the hairstyle or hair length or color, adjust the face shape, modify expressions, or refine specific features to better fit your character.
Method 2 — Build from Scratch
Create your own custom character using the template below. If you don't want to base your character on an existing photo, you can build one from scratch using the template below. This gives you full creative control over facial structure, details, mood, and style from the beginning.
You don't have to write prompts completely on your own — AI tools can always help you analyze, generate, or refine them. You can use this template to build a character of your own imagination, or copy it and have ChatGPT generate a detailed prompt describing a random person:

"Generate a detailed prompt that describes 3 random persons, following the structure provided in this template — [Copy & paste template here]."
Template: CLICK HERE
Step 2
Write a Detailed Prompt
Prompting guide: CLICK HERE
Example: "Ultra-realistic cinematic photography of [YOUR CHARACTERS DESCRIPTION], portrait view showing upper body and face, shot on a 35mm lens, soft natural daylight with realistic shadows, shallow depth of field, true-to-life textures, visible imperfections, subtle film grain, natural color grading, rule of thirds composition, slight imperfections in skin and clothing, realistic focal length."
Refine further with ChatGPT: "Make this prompt significantly more detailed and enhance realism - [ADD YOUR PROMPT HERE]"
Realism isn’t about “stronger AI” - it’s about stronger instructions.
Step 3
Generate Your Main Reference Image
Best model for realistic avatars: NanoBanana Pro
  1. Go to wavespeed.ai → browse the right model: google/nano-banana-pro/text-to-image → enter your prompt and photo settings → generate.
  1. Regenerate until you get a perfect base image you like. Save/download this as your main reference image.

This is your character's "identity anchor" — you will use this image in every future generation to maintain consistency.
Next Steps
Create Custom Scenes & Poses
01
Find Inspiration
Search Pinterest (e.g., "man gym lifting weights", "woman yoga pose", "woman drinking coffee"). Download the image with the desired pose/scene.
02
Get a Scene-Only Prompt
Upload the Pinterest image to ChatGPT. Ask: "Generate a detailed prompt describing ONLY the pose, body position, scene, environment, lighting, and accessories. Do NOT include face details, hair color/length, or body textures."
For female characters: Optionally include hair style/pose flow (e.g., "long hair flowing behind"), but skip color/length.
03
Generate on WaveSpeed.ai
Go to model: google/nano-banana-pro/edit-ultra. Upload your main reference image. Paste the scene/pose prompt and adjust photo settings. Generate → regenerate if needed.
04
Result
Your character in the new pose/scene, face consistent.
Video Creation
Create Videos for TikTok/Reels
For short videos (e.g., dance, gym routine, makeup routine, talking head):
Step 1 — Choose a Reference Video
Download a TikTok/Reel with the movements you want.
Step 2 — Prepare Outfit (Optional)
If changing clothes from reference video: Use Pinterest → find outfit image → get outfit-only prompt from ChatGPT (e.g., "Describe ONLY the clothing, fabric, fit, and accessories").
On wavespeed.ai: Upload main reference → use model: google/nano-banana-pro/edit-ultra with outfit prompt → generate new base image in desired outfit.
Step 3 — Animate
On wavespeed.ai use model: kwaivgi/kling-v2.6-pro/motion-control
  • Upload your reference image (main base or new outfit version)
  • Upload the TikTok/Reel as reference video
  • Generate video
Advanced
NSFW Content Creation
WaveSpeed.ai supports editing/generation suitable for adult content (use responsibly, follow platform rules).
For Static NSFW Images
1
Find Inspiration
Find inspiration on Pinterest (e.g., "girl in bikini pose").
2
Get Prompt from Grok
Upload to Grok.com chatbot (because it's uncensored) → ask for prompt: "Generate detailed prompt for ONLY scene, pose, body position, accessories, and clothing. Exclude face, hair color, skin texture."
3
Generate on WaveSpeed.ai
Use model: x-ai/grok-imagine-image/edit → upload main reference → use with that prompt (e.g., in bikini/lingerie) → generate.
Use a prompt like: "Remove bikini top and bottom, keep pose, detailed breasts [size], hands holding breasts, seductive expression, finger on lip" (be specific about changes). Generate variations with different poses and facial expressions.
For NSFW Videos
1
Start with a Static Image
Use your NSFW static image as a starting point or generate a new one.
2
Animate on WaveSpeed.ai
Use model: x-ai/grok-imagine-video/image-to-video → upload your image → add prompt for actions/movements (e.g., "sensual dance, touching body, explicit movements").

Always use responsibly and follow platform rules.
Made with