AI Models Guide

Understand the AI models available in Genvid and what you can control when generating images, videos, and audio.

Overview

Genvid integrates with multiple AI providers to offer a diverse selection of models for content generation. Each model has different strengths, costs, and capabilities. This guide helps you understand which model to choose for your creative needs and what parameters you can adjust.

Key Concepts

• Model: An AI system trained to generate specific types of content (images, videos, or audio)
• Render Type: The category of generation task (T2I for text-to-image, I2V for image-to-video, etc.)
• Aspect Ratio: The width-to-height proportion of your output (16:9, 9:16, 1:1, etc.)
• Resolution: The pixel dimensions of your output (720p, 1080p, 4K, etc.)
• Duration: For video models, how long the generated video will be

How Models Are Organized

Models in Genvid are organized by media type:

Within each type, models are further categorized by their specific task (render type).

---

Quick Reference: All Models at a Glance

This table provides a quick overview of all 58 active models. Refer to detailed sections below for full information.

Audio Models (2 models)

Image Models - Text-to-Image (13 models)

Image Models - Image-to-Image Edit (6 models)

Image Models - Special (2 models)

Video Models (35+ models)

---

What You Can Control vs. What Is Fixed

Understanding what parameters you can adjust is crucial for getting the results you want.

Parameters You CAN Control

These settings are available in the generation interface:

Parameters Fixed by the Model

These are determined by the model and cannot be changed:

---

Audio Models

Audio models generate voice content from text, enabling realistic character dialogue.

ElevenLabs v2 (Default)

The default text-to-speech model, providing high-quality voice synthesis.

What you control:

• Voice selection (from your cast members)
• The text/dialogue to speak
• Pacing through punctuation

What is fixed:

• Audio sample rate and quality
• Output format

Best for: General voice synthesis needs, shorter dialogue segments.

---

FAL ElevenLabs Dialogue v3

A cost-effective option for longer dialogue, charged per second of generated audio.

What you control:

• Voice selection
• Dialogue text
• Natural pacing through script writing

What is fixed:

• Audio quality parameters
• Output format

Best for: Longer dialogue sequences where per-second pricing is more economical.

---

Image Models: Text-to-Image (T2I)

These models create images from text descriptions. Use them to generate keyframes, asset images, and visual concepts.

Understanding Size Configuration

Different models handle image dimensions in different ways:

---

FLUX1.1 [dev]

A versatile development model offering excellent quality at reasonable cost.

Supported Aspect Ratios:

• 21:9 (Ultra-wide)
• 16:9 (Widescreen)
• 4:3 (Standard)
• 1:1 (Square)
• 3:4 (Portrait standard)
• 9:16 (Vertical/mobile)
• 9:21 (Ultra-tall)

What you control:

• Prompt description
• Aspect ratio
• Resolution/image size

Best for: General-purpose image generation with flexible size options.

---

Flux Pro v1.1 Ultra

Premium quality model for high-end image generation.

Supported Aspect Ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, 9:21

Special Feature: Supports "raw" mode for unprocessed aesthetic.

What you control:

• Prompt description
• Aspect ratio
• Raw mode (true/false)

What is fixed:

• Output resolution (model-determined based on aspect ratio)

Best for: Premium quality images where you need top-tier results.

---

Ideogram v2 / v2 Turbo

Excellent for images containing text, logos, or typography.

Supported Aspect Ratios: 16:9, 4:3, 1:1, 3:4, 9:16, 10:16, 16:10

What you control:

• Prompt description (including text to render)
• Aspect ratio

What is fixed:

• Output resolution

Best for: Any image that needs legible text, signs, logos, or typography.

---

Ideogram v3 / v3 Turbo

The latest Ideogram models with expanded aspect ratio support and improved text rendering.

Supported Aspect Ratios: Includes all v2 ratios plus 3:1 and 1:3 for panoramic/banner formats.

Best for: Highest quality text rendering in images, banner and panoramic formats.

---

Imagen 4 Preview / Ultra / Ultra [Fast]

Google's photorealistic image generation models.

Size Control: Full control with separate width and height parameters. The system provides presets for each aspect ratio and resolution combination.

What you control:

• Prompt description
• Width and height (via presets)
• Aspect ratio

Best for: Photorealistic images, realistic human subjects, natural environments.

---

Nano-Banana Pro T2I

Budget-friendly option with full size control.

What you control:

• Prompt, width, height

Best for: Budget-conscious generation where you need size control.

---

Qwen2.5 VL Image

The most affordable text-to-image option.

Best for: Rapid prototyping, concept exploration, high-volume generation.

---

Seedream 4.0

Balanced quality and cost for general use.

Best for: Everyday image generation with good quality-to-cost ratio.

---

Leonardo Origin

Known for artistic and stylized outputs.

Best for: Artistic styles, illustrations, stylized content.

---

Image Models: Image-to-Image Edit (I2I)

These models modify existing images based on your instructions. Use them for edits, touch-ups, and transformations.

---

FLUX Kontext Pro

A smart image editing model that understands context and makes intelligent modifications.

Supported Aspect Ratios: Yes, but only affects composition, not output resolution.

What you control:

• Edit prompt (what changes to make)
• Aspect ratio preference

What is fixed:

• Output resolution (~1344x786)
• Cannot output at input resolution

Best for: Quick conceptual edits where final resolution is not critical. Use for previewing changes before applying with a different tool.

Not recommended for: Final production images, high-resolution requirements.

---

Nano-Banana Pro I2I / Nano-Banana I2I

Budget editing options with full size control.

What you control:

• Edit prompt
• Output width and height

Best for: Budget editing where you need to preserve or control resolution.

---

Qwen2.5 VL Image Edit Plus

Affordable image editing with good results.

Best for: High-volume editing tasks, affordable touch-ups.

---

Seedream v4 Edit

Budget-friendly editing with full control.

Best for: General editing at reasonable cost.

---

Ideogram Edit

Specialized for editing images with text content.

Best for: Editing images that contain text, modifying typography, fixing text in images.

---

Image Models: Special Purpose

FLUX Kontext Multi (Merge)

Combines multiple images into a single coherent output.

What you control:

• Multiple input images
• Merge prompt describing how to combine them

Best for: Character turnarounds, composite scenes, combining elements from multiple sources.

---

Remove Background

Automatically removes image backgrounds.

What you control:

• Input image only

Output:

• Transparent PNG with background removed

Best for: Extracting subjects from images, creating assets for compositing.

---

Video Models

Video models are categorized by their input type and generation approach.

Understanding Video Render Types

---

Video Duration: What You Can Control

Video models handle duration differently:

---

Image-to-Video Models (I2V)

These models animate a single image into a video. The most common workflow is to create a keyframe and then generate a video from it.

Kling v2.1 Master I2V (Default)

The default and recommended I2V model, offering excellent quality and motion.

What you control:

• First frame (keyframe image)
• Video prompt
• Duration (5s or 10s)
• Aspect ratio

Best for: High-quality character animation, professional production.

---

Kling Model Family (I2V)

All Kling I2V models support: 5 or 10 second durations, aspect ratio control.

---

Veo Models (I2V)

Google's video generation models.

Best for: Photorealistic motion, natural movement.

---

LTX Video Models (I2V)

Ultra-affordable video generation.

Best for: Prototyping, concept testing, high-volume generation.

---

Other I2V Models

---

Text-to-Video Models (T2V)

Generate videos from text descriptions without any input images.

When to Use T2V

• Abstract concepts without visual reference
• Establishing shots and backgrounds
• Exploring ideas before creating keyframes
• When no keyframe exists yet

---

Kling T2V Models

---

Other T2V Models

---

Keyframe-to-Video Models (KF2V)

Generate videos that smoothly transition between a starting image and ending image.

When to Use KF2V

• Cinematic transitions
• Controlled camera movements
• Character repositioning
• When you know exactly how a shot should start and end

---

KF2V Models

What you control:

• First frame (start image)
• Last frame (end image)
• Transition prompt
• Duration (where applicable)

---

Video-to-Video Models (V2V)

Transform existing videos with style changes or enhancements.

Kling v2.1 Master V2V

What you control:

• Source video
• Transformation prompt
• Optional dialog sync

Best for: Style transfer, restyling existing videos, syncing to dialog.

---

Reference-to-Video Models (R2V)

Generate videos using reference images for character/scene consistency.

Kling v2.1 Master R2V

What you control:

• 1-N reference images (cast, locations, props)
• Video prompt
• Duration

Best for: Maintaining character consistency across shots, using specific asset appearances.

---

Upscale Models

Enhance video resolution and quality.

Topaz Video Upscale

What you control:

• Source video
• Scale factor
• Target resolution
• Target FPS

What is fixed:

• Upscaling algorithm
• Output format

Best for: Final polish before delivery, enhancing resolution of selected videos.

---

Choosing the Right Model

For Images

For Video

For Audio

---

Cost Optimization Tips

Image Generation

1. Use Qwen2.5 for prototyping ($0.003) then switch to premium for finals
2. Ideogram Turbo vs Standard - Turbo is half the price with good quality
3. Megapixel pricing - FLUX1.1 dev charges by megapixel; smaller images cost less

Video Generation

1. Use LTX for concepts ($0.002/second) before committing to expensive models
2. 5 seconds vs 10 seconds - Choose shorter durations when possible
3. Upscale last - Only upscale your selected, final videos
4. Per-frame vs per-second - Calculate which is cheaper for your video length

General Tips

1. Generate multiple variations with budget models first
2. Refine your prompt before using premium models
3. Track credit usage in Budget Management
4. Delete rejected content to declutter your workspace

---

Common Questions

Why can't I control the resolution on some models?

Different AI models have different architectures. Some are trained to output specific resolutions and cannot change them. Models like FLUX Kontext Pro produce fixed-size outputs regardless of input. Always check the model's "Size Control" specification before generating.

Why is my edited image smaller than my original?

If you used FLUX Kontext Pro for image editing, it outputs at approximately 1344x786 regardless of input size. For edits that preserve resolution, use Nano-Banana I2I, Qwen2.5 VL Edit, or Seedream v4 Edit.

Why can Veo 3 only make 8-second videos?

This is a fundamental characteristic of how Veo 3 was trained. The model optimizes for quality at a specific duration and does not support other lengths. If you need different durations, use Veo 2 or Kling models.

What's the difference between I2V and R2V?

Why are some models more expensive?

Price generally correlates with:

• Output quality - Higher fidelity, better details
• Speed - Faster generation often costs more
• Features - More controllable parameters
• Training cost - Models trained on larger datasets

How do I know which models support my aspect ratio?

The interface filters models by compatibility. When you select a generation mode and aspect ratio, only compatible models appear in the dropdown. You can also refer to the detailed model information in this guide.

Can I use multiple models together?

Yes! A common workflow is:

1. Generate keyframe with an image model
2. Create video with an I2V model
3. Upscale with the upscale model
4. Add dialog with an audio model

Each step can use a different model optimized for that task.

---

Next Steps

• Video Editor - Generate videos using these models
• Image Editor - Create and edit images
• Keyframe Editor - Compose keyframes for video generation
• Subscriptions & Credits - Understand credit costs
• Budget Management - Track spending across models