Usage Guide
Learn how to get the best results from LTX Video Generator.
Table of contents
- Basic Workflow
- Gemma Prompt Enhancement
- Image-to-Video
- Adding Audio
- Writing Effective Prompts
- Using the Queue
- History Features
- Tips for Best Results
Basic Workflow
1. Write Your Prompt
Enter a descriptive text prompt in the main text field. The more detail you provide, the better the results.
Good prompt:
“The camera slowly pans across a misty forest at dawn, with rays of golden sunlight filtering through the trees”
Less effective:
“forest”
2. Choose a Preset
Select a preset from the dropdown to quickly configure parameters:
| Preset | Resolution | Frames | Steps | Best For |
|---|---|---|---|---|
| Quick Preview | 512×320 | 49 | 15 | Quick tests |
| Standard | 768×512 | 121 | 30 | Balanced quality |
| High Quality | 768×512 | 121 | 40 | Best results |
| Portrait | 512×768 | 97 | 30 | Vertical videos |
| Square | 512×512 | 97 | 30 | Social media |
The LTX-2 Unified model uses a 2-stage generation pipeline with built-in audio.
3. Generate
Click the Generate button. Progress shows:
- Model loading status
- Stage 1 and Stage 2 denoising progress
- Video encoding and saving
4. View Results
- Queue sidebar shows real-time progress
- History tab displays all generated videos with thumbnails
- Videos save to your configured output directory
Gemma Prompt Enhancement
Improve results by having Gemma rewrite your prompt with vivid details:
- Go to Settings > Generation and enable Enable Gemma Prompt Enhancement
- In the prompt view, expand Prompt Enhancement (Gemma)
- Optionally enable Use uncensored enhancer to avoid content filters (first run downloads ~7GB)
- Click Preview enhanced prompt to see the rewritten prompt before generating
- Generate as usual—the enhanced prompt is used automatically
If enhancement returns empty (e.g. safety filter), the app auto-retries with filtered words replaced, then merges originals back.
Image-to-Video
You can animate images into videos:
- Expand the Image to Video section
- Click Select Image to choose a source image
- Adjust Image Strength (1.0 = full influence, 0.0 = ignore image)
- The first frame will be conditioned on your image
Adding Audio
Voiceover / Narration
Add text-to-speech voiceover to your videos:
- Expand the Voiceover / Narration section
- Choose your source:
- MLX Audio (Local) - Free, runs on-device, good quality
- ElevenLabs (Cloud) - High quality, requires API key
- Select a voice from the dropdown (10 voices for MLX, 9 for ElevenLabs)
- Enter your narration text
- Generate your video - audio will be added automatically
You can also add audio later by right-clicking any video thumbnail in the Video Archive.
Background Music
Add AI-generated instrumental music (requires ElevenLabs API key):
- Expand the Background Music section
- Toggle Generate background music on
- Choose a genre from the dropdown:
- Organized into 9 categories with 54 total presets
- Preview the prompt by hovering over the selection
- Music automatically matches your video length
Music Genre Categories
| Category | Genres |
|---|---|
| Electronic | EDM, House, Techno, Ambient, Chillwave, Synthwave, D&B, Trance |
| Hip-Hop/R&B | Trap, Lo-Fi, Boom Bap, Slow R&B, Modern R&B, Soul |
| Rock | Classic, Alternative, Indie, Metal, Punk, Acoustic |
| Pop | Modern, Indie, Dance, Acoustic |
| Jazz/Blues | Smooth, Bebop, Lounge, Electric Blues, Acoustic Blues |
| Classical/Cinematic | Orchestral, Piano, Chamber, Epic, Tense, Uplifting |
| World | Latin, Reggae, Afrobeat, Middle Eastern, Asian |
| Country/Folk | Modern, Classic, Acoustic Folk, Indie Folk |
| Functional | Corporate, Motivational, Relaxing, Suspense, Action, Romantic, Happy, Sad, Dramatic, Mystery |
Audio from History
Add audio to previously generated videos:
- Go to Video Archive
- Right-click any video thumbnail
- Select Add Audio (or Replace Audio if it already has audio)
- Choose from three tabs:
- Voiceover - Add narration only
- Music - Add background music only
- Both - Add voiceover and music together
When combining voiceover and music, the music is automatically ducked to 20% volume so the voice remains clear.
Writing Effective Prompts
Include Camera Movement
"The camera slowly pans left revealing..."
"A drone shot flying over..."
"Close-up tracking shot of..."
"The camera pushes in toward..."
Describe Motion
"waves gently crashing on the shore"
"leaves falling in slow motion"
"clouds drifting across the sky"
"a person walking through..."
Specify Lighting
"at golden hour with warm lighting"
"under dramatic storm clouds"
"illuminated by neon city lights"
"in soft diffused morning light"
Add Atmosphere
"with fog rolling through the valley"
"rain drops falling on the window"
"dust particles floating in sunbeams"
"snow gently falling"
Using the Queue
Add Multiple Generations
- Write your prompt
- Click Add to Queue (instead of Generate)
- Modify the prompt or parameters
- Add more to the queue
- Videos generate one after another
Batch Variations
Click the batch menu (stack icon) to:
- Generate 3 variations
- Generate 5 variations
- Each uses a random seed for different results
Queue Management
- Cancel the current generation with the X button
- Remove pending items from the queue
- Clear the entire queue with the Clear button
History Features
Browse Videos
- Thumbnails show a frame from each video
- Sort by newest, oldest, or prompt alphabetically
- Search prompts to find specific videos
Video Details
Click a video to see:
- Full video preview (loops automatically)
- Original prompt
- All generation parameters
- Timestamp and generation duration
- Seed value for reproducibility
Actions
- Show in Finder - Reveal the video file
- Share - Share via macOS share sheet
- Reuse Prompt - Copy prompt back to input
- Delete - Remove video
Tips for Best Results
Start Small
- Use Quick Preview preset first
- Iterate on prompts quickly
- Only increase quality for final renders
Use Negative Prompts
Click the disclosure arrow to add negative prompts:
worst quality, blurry, jittery, distorted, watermark
Reproducible Results
- Note the seed value of good generations
- Enter the same seed to reproduce results
- Useful for making variations with slight prompt changes
Memory Management
- Higher resolutions use more memory
- Close other apps if you encounter issues
- 32GB RAM minimum, 64GB recommended