AI Models Reference

Influencer Studio integrates the most advanced AI models in the industry. Each model is optimized for specific tasks to deliver the best possible results.

Video Generation Models

Text-to-Video Models

Kling v2.6 Pro

The industry-leading model for photorealistic video generation with native audio generation support.

Best For:

High-fidelity video content
Smooth motion and realistic physics
Character animations
Cinematic sequences
Videos with native audio/dialogue

Key Features:

Photorealistic output quality
Advanced motion control
Consistent character appearance
Multiple duration options (5s, 10s)
Aspect ratio support (16:9, 9:16, 1:1)
Native audio generation (English/Chinese with auto-translation)

Credits: 70 credits for 5s (audio off), 140 credits for 5s (audio on)

Veo 3.1 Fast

Google's latest state-of-the-art video generation model. Fast and cost-effective.

Best For:

Complex scene compositions
Precise camera movements
Controlled scene transitions
Professional video production

Key Features:

Superior scene understanding
Precise control over video elements
Excellent temporal consistency
Aspect ratio support (16:9, 9:16)

Credits: 80 credits per 8-second video

MiniMax Hailuo-02 Pro

Advanced video generation model with 1080p resolution.

Best For:

High-resolution video content
Professional productions
Marketing videos

Key Features:

1080p output
Optional prompt optimization
High-quality motion

Credits: 125 credits per video

ByteDance Seedance v1.0 Pro Fast

Upgraded fast video generation model from ByteDance.

Best For:

Quick video generation
Social media content
Rapid iterations

Credits: Variable based on duration

ByteDance Seedance v1.0 Lite

Lightweight, fast video generation for quick iterations.

Credits: Lower cost option for faster generation

Wan 2.6

Wan 2.6 text-to-video model with enhanced quality and motion generation.

Best For:

High-quality video generation
Creative projects
NSFW content support

Credits: Variable based on settings

Wan 2.2 5B

Wan 2.2 5B text-to-video model with NSFW support.

Credits: ~1 credit per second of video

Mochi v1

Specialized video generation model.

Credits: Variable

Hunyuan Video

Advanced video generation from Tencent.

Credits: Variable

CogVideoX

Video generation with advanced cognitive understanding.

Credits: Variable

Image-to-Video Models

Kling v2.6 Pro (Image-to-Video)

Animate static images with photorealistic motion and native audio support.

Best For:

Bringing photos to life
Product animations
Character animations from photos
Talking head videos with native audio

Credits: 70 credits for 5s (audio off), 140 credits for 5s (audio on)

VEED Fabric 1.0

Image to talking video with built-in lip sync capabilities.

Best For:

Talking head videos
Animated portraits
Marketing videos with speech

Credits: Variable

OmniHuman (ByteDance)

Human image to video with audio support.

Best For:

Human animations
Talking avatars
Influencer content

Credits: Variable

Veo 3.1 Fast (Image-to-Video)

Convert static images to video with Google's advanced model.

Best For:

Scene animation
Product videos
Creative transitions

Credits: Similar to text-to-video pricing

Veo 3.1 First-Last Frame

Create videos from first and last frame keyframes.

Best For:

Precise video control
Animated transitions
Storyboard-based videos

Credits: Variable

ByteDance Seedance v1.0 (Image-to-Video)

Fast image-to-video generation from ByteDance.

Credits: Variable based on duration

Wan 2.6 / 2.2 5B (Image-to-Video)

Wan models for image-to-video generation with NSFW support.

Credits: ~1 credit per second of video

MiniMax Hailuo-02 Pro (Image-to-Video)

High-resolution image-to-video conversion.

Credits: 125 credits per video

Hunyuan Video (Image-to-Video)

Tencent's image-to-video model.

Credits: Variable

Luma Dream Machine Ray 2

Advanced image-to-video with ray-traced quality.

Credits: Variable

CogVideoX (Image-to-Video)

Cognitive image-to-video generation.

Credits: Variable

Framepack

Specialized frame animation model.

Credits: Variable

Magi Image-to-Video

Creative image animation model.

Credits: Variable

Image Generation Models

Text-to-Image Models

Flux-Krea

State-of-the-art photorealistic image generation model with 12 billion parameters.

Best For:

Ultra-realistic photographs
Portrait photography
Product photography
Marketing materials
High-detail imagery

Key Features:

Unparalleled photorealism
Exceptional detail rendering
Natural lighting understanding
Consistent character generation with LoRAs
Multiple aspect ratios
40 inference steps for quality

Credits: ~25 credits per megapixel (e.g., 1024x1024 = 25 credits, 2048x2048 = 100 credits)

Usage Example:

{
  "prompt": "Professional headshot of a woman in business attire, studio lighting, high detail, 8k",
  "model": "flux-krea",
  "settings": {
    "aspect_ratio": "1:1",
    "num_inference_steps": 40
  }
}

Flux Ultra v1.1

The most high-resolution and best quality image generation model available. Premium quality with 2x credit cost.

Best For:

Ultra-high resolution images
Maximum quality requirements
Professional photography
Print materials

Key Features:

Best-in-class image quality
Highest resolution output
Professional color accuracy

Credits: 40 credits per image

Flux Schnell

Fast, high-quality image generation. Best value for quality.

Best For:

Quick generations
Batch processing
Cost-effective high-quality images

Key Features:

Fully uncensored
Fast 4-step inference
High quality output
Best value proposition

Credits: 24 credits per image

Flux SRPO

High aesthetic quality image generation with superior prompt adherence.

Best For:

Artistic content
Creative projects
High aesthetic standards

Credits: Variable based on settings

ByteDance Seedream v4

ByteDance's advanced text-to-image model with improved quality and prompt adherence.

Best For:

High-quality image generation
Complex prompts
Professional content

Key Features:

Multiple aspect ratios
Safety checker option
Configurable guidance scale
Seed control for reproducibility

Credits: 40 credits per image

Qwen Image

Intelligent image generation with natural language understanding.

Best For:

Complex prompt interpretation
Natural language descriptions
Intelligent scene composition

Credits: Variable

HiDream I1 Fast

Fast, high-quality image generation.

Best For:

Quick turnaround
Batch processing
Social media content

Credits: Variable

Flux Pro Kontext Max

Advanced image generation with superior context understanding.

Credits: Variable

Juggernaut Flux (Lightning & LoRA)

Fast flux-based models optimized for speed and LoRA support.

Credits: Variable

Ideogram v2 Turbo / v3

Advanced text rendering and image generation.

Best For:

Images with text
Logos and graphics
Marketing materials

Credits: Variable

ICLight v2

Advanced lighting control for image generation.

Best For:

Studio photography simulation
Lighting experiments
Product photography

Credits: Variable

Hunyuan Image v3

Tencent's advanced image generation model with excellent prompt adherence.

Credits: Variable

Nano Banana Pro (Text-to-Image)

Google's Gemini 3 Pro Image architecture for production-quality text-to-image generation with industry-leading text rendering.

Best For:

Marketing campaign generation
Product visualization workflows
Creative content requiring text accuracy
Infographic and diagram creation
Content with typography and text elements

Key Features:

Multimodal understanding through Gemini 3 Pro architecture
Industry-leading text rendering in multiple languages
Advanced semantic interpretation without prompt engineering
Natural language creative direction
Character consistency for up to 5 people
Resolution options: 1K, 2K, 4K
Multiple aspect ratios (21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16)

Credits: 22 credits per image (44 credits for 4K resolution)

Usage Example:

{
  "prompt": "Professional marketing poster with text 'Grand Opening' in elegant typography, modern design, vibrant colors",
  "model": "fal-ai/nano-banana-pro",
  "settings": {
    "aspect_ratio": "16:9",
    "resolution": "2K"
  }
}

Best Practices:

Use natural language descriptions for best results
Leverage text rendering for signs, posters, and graphics
Specify typography style and mood for text elements
Ideal for batch A/B testing with consistent quality

Image-to-Image & Editing Models

All editing models support both text-to-image and image-to-image workflows.

Seedream 4.0 Edit

Advanced AI editing with powerful manipulation capabilities.

Best For:

Complex image editing
Image modifications
Background replacement
Style transfer
Creative image alterations

Key Features:

Intelligent content understanding
Seamless blending
Context-aware editing
High-quality output
Support for multiple reference images

Credits: 12 credits per image

Usage Example:

{
  "image_url": "https://example.com/image.jpg",
  "prompt": "Replace the background with a tropical beach",
  "model": "seedream-4.0-edit"
}

Best Practices:

Provide clear, specific editing instructions
Can use multiple reference images
Describe desired changes in detail
Specify style matching requirements

Nano Banana Pro (Image-to-Image)

Google's state-of-the-art Nano Banana 2 image editing model for precise modifications using multimodal understanding.

Best For:

High-quality edits and adjustments
Multi-image editing (up to 14 images)
Professional image modifications
Advanced image transformations
Context-aware image blending

Key Features:

Multimodal semantic understanding
Multi-image support (up to 14 images)
Precise control
Professional results
Natural language editing instructions

Credits: 22 credits per image

Usage Example:

{
  "image_url": "https://example.com/image.jpg",
  "prompt": "Adjust lighting to golden hour",
  "model": "nano-banana-edit"
}

Qwen Image Edit

AI editing with natural language control and intelligence.

Best For:

Conversational editing commands
Complex multi-step edits
Intelligent scene understanding
Context-aware modifications

Key Features:

Natural language processing
Intelligent interpretation
Smart object recognition

Credits: Variable

Usage Example:

{
  "image_url": "https://example.com/image.jpg",
  "prompt": "Make the person smile more and add warmer lighting",
  "model": "qwen-edit"
}

Flux Krea Image-to-Image

Use Flux-Krea's powerful model for image-to-image transformations.

Credits: Same as text-to-image (~25 credits per megapixel)

Flux SRPO Image-to-Image

High aesthetic quality image editing and transformation.

Credits: Variable

3D & Text Models

Meshy

3D model generation from text or images.

Best For:

Text-to-3D generation
Image-to-3D conversion
3D asset creation
Product visualization
Game asset generation

Key Features:

Text-to-3D synthesis
Image-to-3D conversion
Multiple export formats
Optimized geometry
Texture generation

Usage Example:

{
  "prompt": "A modern wooden chair with metal legs",
  "model": "meshy",
  "settings": {
    "output_format": "obj",
    "texture_resolution": "2k"
  }
}

Supported Formats:

OBJ
FBX
GLB/GLTF
STL

Audio Generation Models

Sonic-3

State-of-the-art voice synthesis and audio generation.

Best For:

Voiceovers and narration
Character voices
Podcast audio
Marketing videos
Talking head videos (with lip sync)

Key Features:

Natural-sounding speech
Multiple voice options
Emotion and tone control
Multiple languages
Professional audio quality
Pronunciation control

Usage Example:

{
  "text": "Welcome to Influencer Studio, where AI meets creativity.",
  "model": "sonic-3",
  "settings": {
    "voice": "professional-female",
    "emotion": "enthusiastic",
    "speed": 1.0
  }
}

Available Voice Profiles:

Professional Male/Female
Casual/Conversational
Energetic/Enthusiastic
Calm/Soothing
Character voices
Multiple accents and languages

Best Practices:

Use proper punctuation for natural pauses
Specify emotion and tone for better results
Choose appropriate voice profile for your content
Adjust speed for different content types

Model Selection Guide

Choosing the Right Model

For Maximum Quality:

Images: Flux-Krea
Videos: Kling or Veo 3.1
Audio: Sonic-3
3D: Meshy

For Speed & Efficiency:

Quick edits: Nano Banana Pro
Fast iterations: Finetuned Flux models
Video editing: Veed

For Creative Control:

Complex edits: Seedream 4.0
Natural language edits: Qwen
Precise video control: Veo 3.1

For Character Consistency:

Train a custom LoRA (influencer)
Use Flux-Krea with your LoRA
Generate thousands of consistent images

Influencer Training (LoRA Models)

Train custom character models (influencers) with consistent appearance across all generations.

What is Influencer Training?

Influencer training creates a custom LoRA (Low-Rank Adaptation) model that learns the unique characteristics of a person or character from a set of reference photos. Once trained, this model can generate unlimited consistent images of that character in any pose, outfit, location, or scenario.

Training Process

Requirements:

8-20 high-quality photos of the same person
Clear, well-lit images
Variety of angles and expressions
Consistent lighting preferred
Photos should show the face clearly

Training Time:

Typically 15-45 minutes depending on dataset size
Training happens in the background
You'll be notified when complete

Best Practices:

Photo Quality: Use high-resolution, clear photos (minimum 512x512, preferably 1024x1024 or higher)
Variety: Include different angles: front, side, 3/4 view
Expressions: Mix of neutral and smiling expressions
Consistency: Same person throughout all training images
Backgrounds: Variety is good, but person should be the focus
Clothing: Different outfits help the model learn the person, not just specific clothes
Avoid: Blurry images, heavy filters, multiple people in frame, sunglasses/hats covering face

Using Trained Influencers

Once trained, you can use your influencer in any image generation:

{
  "prompt": "A professional headshot in business attire, studio lighting",
  "model": "flux-krea",
  "influencer_id": "your_influencer_id",
  "settings": {
    "aspect_ratio": "1:1"
  }
}

The influencer will appear in the generated image while following your prompt's instructions for pose, clothing, location, and style.

Influencer Consistency

Trained influencers maintain:

Facial features and structure
Skin tone and texture
Eye color and shape
Hair color and style (though you can change these via prompts)
Overall appearance and identity

You can still customize:

Poses and expressions
Clothing and outfits
Locations and backgrounds
Lighting and mood
Artistic style

API Availability

Note: Influencer training is currently not available through the REST API. You can only train influencers through the web interface at app.influencerstudio.com.

However, once trained, you can use your influencers via the API by including the influencer_id parameter in image generation requests.

To get your influencer IDs:

Train influencers via the web interface
Find your influencer ID in the influencer management panel
Use the ID in API requests for consistent character generation

Training Credits

Training an influencer consumes credits based on:

Number of training images
Training duration
Model complexity

Check the web interface for current training costs.

Model Pricing

Credits are consumed per generation. Pricing is fixed per operation:

Image Generation (Text-to-Image):

Flux-Krea: ~25 credits per megapixel (1024x1024 = 25 credits, 2048x2048 = 100 credits)
Flux Ultra v1.1: 40 credits per image
Flux Schnell: 24 credits per image
ByteDance Seedream v4: 40 credits per image
Other models: 24-40 credits per image typically

Image Editing (Image-to-Image):

Seedream 4.0 Edit: 12 credits per image
Nano Banana Pro: 22 credits per image (44 for 4K)
Qwen Edit: Variable
Flux-based editing: Similar to text-to-image pricing

Video Generation (Text-to-Video):

Kling v2.6 Pro: 70 credits for 5s (audio off), 140 credits for 5s (audio on)
Veo 3.1 Fast: 80 credits per 8s video
MiniMax Hailuo-02 Pro: 125 credits per video
Wan 2.2 5B: ~1 credit per second
Other models: Variable based on duration and quality

Video Generation (Image-to-Video):

Kling v2.6 Pro: 70 credits for 5s (audio off), 140 credits for 5s (audio on)
Similar pricing structure to text-to-video for most models

3D Generation:

Meshy: Variable based on complexity

Audio Generation:

Sonic-3: Variable based on length

LoRA Training:

Custom model training: Variable based on dataset size

API Access

All models are available through our REST API:

Model Updates

We continuously update and improve our models. Check our changelog for:

New model releases
Performance improvements
Feature additions
Deprecated models

Support

Need help choosing the right model?

Check the API documentation for detailed parameters
Try different models in the web interface
Contact support for recommendations
Review example use cases in our gallery

AI Models Reference

On this page

Video Generation Models

Text-to-Video Models

Kling v2.6 Pro

Veo 3.1 Fast

MiniMax Hailuo-02 Pro

ByteDance Seedance v1.0 Pro Fast

ByteDance Seedance v1.0 Lite

Wan 2.6

Wan 2.2 5B

Mochi v1

Hunyuan Video

CogVideoX

Image-to-Video Models

Kling v2.6 Pro (Image-to-Video)

VEED Fabric 1.0

OmniHuman (ByteDance)

Veo 3.1 Fast (Image-to-Video)

Veo 3.1 First-Last Frame

ByteDance Seedance v1.0 (Image-to-Video)

Wan 2.6 / 2.2 5B (Image-to-Video)

MiniMax Hailuo-02 Pro (Image-to-Video)

Hunyuan Video (Image-to-Video)

Luma Dream Machine Ray 2

CogVideoX (Image-to-Video)

Framepack

Magi Image-to-Video

Image Generation Models

Text-to-Image Models

Flux-Krea

Flux Ultra v1.1

Flux Schnell

Flux SRPO

ByteDance Seedream v4

Qwen Image

HiDream I1 Fast

Flux Pro Kontext Max

Juggernaut Flux (Lightning & LoRA)

Ideogram v2 Turbo / v3

ICLight v2

Hunyuan Image v3

Nano Banana Pro (Text-to-Image)

Image-to-Image & Editing Models

Seedream 4.0 Edit

Nano Banana Pro (Image-to-Image)

Qwen Image Edit

Flux Krea Image-to-Image

Flux SRPO Image-to-Image

3D & Text Models

Meshy

Audio Generation Models

Sonic-3

Model Selection Guide

Choosing the Right Model

Influencer Training (LoRA Models)

What is Influencer Training?

Training Process

Using Trained Influencers

Influencer Consistency

API Availability

Training Credits

Model Pricing

API Access

Model Updates

Support