Model Selection

Chat Video Pro has two different model selection systems, each serving a specific purpose. Understanding the difference helps you choose the right workflow for your needs.

How to Use the Different Model Selectors

Two Model Selectors

1. Top-Right Model Selector (Chat Header)

Location: Top-right corner of the chat header, next to the sidebar toggle

What it controls:

• LLM Model - The language model used for text generation and chat responses. Official lineup: GPT-5.3 Instant (fast default), GPT-5.2 Thinking (deeper reasoning), Claude Sonnet 4.6, and Gemini 3.1 Pro.

• Default Image Model - The image model used for natural language requests like "generate me an image of..."

When to use:

• Text conversations - Choosing which AI handles your chat

• Quick image generation - When you say "create an image of a sunset" in chat

• NLP image requests - Natural language image generation without opening Generate Media mode

Characteristics:

• ✅ Fast path - Quick and simple

• ✅ Natural language - Just describe what you want

• ❌ Limited control - No aspect ratio, resolution, or advanced settings

• ❌ No video generation - Cannot generate videos from this selector

Example workflow:

1. Type: "Generate me an image of a coffee shop"

2. AI uses your default image model (set in top-right selector)

3. Image appears in chat

4. No settings to configure - uses defaults

2. Generate Media Model Selector (Composer)

Location: Appears in the composer when you enable the Generate Media toggle

What it controls:

• Video Models - Sora, Veo, Kling, Hailuo, WAN, etc.

• Image Models - Flux, GPT Image, Nano Banana, etc.

• Video-to-Video Models - SAM 3, Kling VFX, LTX Reshoot, Upscale

When to use:

• Video generation - The only way to generate videos

• Custom image settings - When you need specific aspect ratio, resolution, or advanced options

• Video editing - When you want to edit existing videos

• Full control - When you need to customize all parameters

Characteristics:

• ✅ Full control - Aspect ratio, resolution, duration, audio, etc.

• ✅ Video generation - Only way to create videos

• ✅ Advanced settings - Model-specific parameters

• ✅ Adaptive - Model list changes based on what you attach

Example workflow:

1. Enable Generate Media toggle

2. Select video model (e.g., Veo 3.1 or Veo 3.1 Lite for a lower-cost Veo option)

3. Set aspect ratio (16:9), duration (8s), enable audio

4. Enter detailed prompt

5. Generate with full control

How the Generate Media Model List Adapts

The Generate Media model selector intelligently adapts based on what you attach to the composer. This ensures you only see relevant models for your current workflow.

No Attachments (Text-to-Media)

Available Models:

• Text-to-Video models - Sora 2, Veo 3.1, Veo 3.1 Lite, Kling, Hailuo, WAN

• Text-to-Image models - Flux 2 Max, GPT Image 2, Nano Banana Pro, etc.

Use case: Creating content from scratch with text prompts

One Image Attached

Available Models:

• Image-to-Video models - Animate your image into motion

• Image-to-Image models - Edit or modify the image

• Text-to-Video models - Still available (image becomes reference)

Automatic behavior:

• Model selector may auto-switch to image-to-video mode

• You can manually switch between image and video modes

• Image editing models appear when in image mode

Use case:

• Animate a still image into a video

• Edit an existing image

• Use image as reference for video generation

Two Images Attached

Available Models:

• Transition models - Kling O3 Transition, Kling 3.0 Pro/Standard, Veo 3.1 First/Last, Veo 3.1 Lite

• Text-to-Video models - Still available (images become references)

Automatic behavior:

• Transition Mode activates automatically

• Transition-specific models appear in the selector

• Duration and aspect ratio may be locked based on source images

Use case:

• Create smooth transitions between two keyframes

• Morph from one scene to another

• Generate transition footage for edits

One Video Attached

Available Models:

• Video-to-Video models - SAM 3, Kling O3 VFX, LTX Reshoot

• Upscale models - Topaz, Flash VSR 2, Bria

Automatic behavior:

• Text-to-video and image models disappear

• Only video editing models are shown

• Model selector may auto-select SAM 3 (default for video editing)

Use case:

• Remove background from video (SAM 3)

• Add visual effects (Kling VFX)

• Upscale video resolution

Model Selection Workflow

Quick Image Generation

Best for: Fast, simple image generation

1. Set default image model in top-right selector

2. Type in chat: "Generate me an image of a mountain landscape"

3. Image appears - No settings to configure

4. Done - Quick and simple

Full Control Image Generation (Generate Media)

Best for: When you need specific settings

1. Enable Generate Media toggle

2. Select image model from dropdown

3. Set aspect ratio (16:9, 9:16, 1:1, etc.)

4. Set resolution (if available for model)

5. Enter prompt with details

6. Generate with full control

Video Generation (Generate Media Only)

Best for: Creating videos (only way to do it)

1. Enable Generate Media toggle

2. Select video model (Sora, Veo, Kling, etc.)

3. Set all parameters:

   • Aspect ratio

   • Resolution

   • Duration

   • Audio (if supported)

4. Enter detailed prompt

5. Generate video

Video Editing (Generate Media with Video Attached)

Best for: Modifying existing videos

1. Attach video using Import Clip button or drag & drop

2. Enable Generate Media (if not already enabled)

3. Model selector shows only video-to-video models

4. Select editing mode:

   • SAM 3 for background removal

   • Kling VFX for effects

   • LTX Reshoot for regeneration

   • Upscale for resolution increase

5. Configure settings for selected model

6. Generate

Understanding Model Modes

Text-to-Video Mode

• No images attached

• Text prompt only

• All video models available

• Full parameter control

Image-to-Video Mode

• 1 image attached

• Describe motion in prompt

• Image-to-video models available

• Image becomes the starting frame

Transition Mode

• 2 images attached

• Automatic activation

• Transition models only

• Creates smooth morph between frames

Video-to-Video Mode

• 1+ videos attached

• Editing existing footage

• Video editing models only

• No text-to-video models shown

Tips for Choosing the Right Selector

Use Top-Right Selector When:

• ✅ You want quick image generation

• ✅ You don't need custom settings

• ✅ You're having a conversation and want images

• ✅ You want the fastest workflow

Use Generate Media Selector When:

• ✅ You need to generate videos

• ✅ You want custom aspect ratio/resolution

• ✅ You need advanced model settings

• ✅ You're editing existing videos

• ✅ You want full control over generation

Model Selection Persistence

Top-Right Selector

• LLM model - Persists across conversations

• Default image model - Persists across conversations

Generate Media Selector

• Last selected model - Remembers your choice

• Mode (image/video) - Remembers last used mode

• Settings - Per-model settings are saved

• Adaptive - Changes based on attachments

Troubleshooting

"I can't generate videos"

Solution: Videos can only be generated using the Generate Media model selector. Enable the Generate Media toggle first.

"Wrong models showing"

Check:

• What's attached to the composer? (images, videos, nothing)

• Is Generate Media enabled?

• Are you in the right mode? (image vs video)

"Settings don't match the model"

Solution: Some settings are model-specific. The interface automatically shows/hides options based on the selected model.

"Model selector is empty"

Solutions:

• Check your Fal.ai account has credits

• Verify API keys are configured

• Some models may have temporary availability issues

• Check you're in the correct mode (image vs video)

---

Next: Learn about Conversation Starters to get started quickly with pre-configured assistants.