Text-to-Image Generation

Text-to-image models create still images from a text description — no source media required. Use them for concept art, storyboard frames, thumbnail options, social media assets, or any visual that does not exist yet.

Step-by-Step

1. Open the Panel

In Premiere Pro, go to Window > Extensions > modelBridge.ai.

2. Search for a Text-to-Image Model

Open the model selector and search for a text-to-image model. Some options to start with:

Model	Strengths	Typical Cost
FLUX 2 Pro	Photorealistic, prompt-accurate	~$0.04–0.06 per image
FLUX 2 Dev	Fast, lower cost	~$0.02–0.04 per image
Recraft v3	Strong typography, design-oriented	~$0.04 per image

Use the Image Gen filter chip to show only text-to-image models.

3. Write Your Prompt

Describe the image you want. Text-to-image models respond well to specific, layered prompts.

Good prompt example:

“Close-up portrait of a weathered fisherman, golden hour side lighting, shallow depth of field, film grain, Kodak Portra 400 look”

Weak prompt example:

“Man outside”

Tips for effective prompts:

Describe composition — “close-up”, “wide establishing shot”, “overhead flat lay”
Specify lighting — “harsh midday sun”, “neon backlight”, “soft window light”
Include style references — “35mm film”, “editorial photography”, “oil painting”
Mention mood — “melancholic”, “vibrant”, “clinical”
Be specific about subjects — “a black Labrador retriever on a mossy forest trail” rather than “a dog in nature”

4. Configure Parameters

Adjust the settings below the prompt field:

Resolution / aspect ratio — match your project dimensions or intended use (16:9 for timeline, 1:1 for social, 9:16 for vertical)
Number of images — some models can generate multiple variations in a single request
Guidance scale — controls how closely the model follows your prompt (higher = more literal, lower = more creative)

5. Use the Prompt Optimizer

Click the sparkle button below the prompt field to have AI rewrite your prompt for better results. This is useful when you know what you want but are not sure how to describe it in terms the model responds to. The optimization costs approximately $0.01.

6. Check the Cost Estimate

The cost badge next to the Generate button updates live as you change parameters. Most text-to-image generations cost under $0.10. Resolution and number of images are the main cost drivers.

7. Generate

Click Generate. Text-to-image models are typically fast — most results arrive in 5–30 seconds.

The progress stages:

Submitting your request
Queued (waiting for GPU)
Generating
Downloading the result
Importing to Premiere Pro

8. Preview and Import

The generated image appears in the Source Monitor for review. From there:

Import to Timeline — places the image on the timeline at the playhead position
Save to Project Bin — imports without timeline placement
Generate again — try a different prompt or different model for comparison

No Source Clip Required

Text-to-image models do not need a clip selected on the timeline. The media card will show that no selection is needed, and the Generate button will be active as long as you have entered a prompt.

Working with Generated Images

Once imported, a generated image is a standard Premiere Pro asset. You can:

Set its duration on the timeline
Apply transitions and effects
Use it as a source for image-to-video generation (animate it with AI)
Export it as a still frame

Common Issues

Result does not match the prompt

AI models interpret prompts differently. Try rephrasing, adding more detail, or switching to a different model. FLUX models tend to follow prompts more literally; other models may take more creative liberty.

Image quality is low

Check the resolution setting. Some models default to lower resolutions for speed. Increasing resolution improves quality but may also increase cost.

Wrong aspect ratio

Make sure the aspect ratio parameter matches your intended use. Generating a 1:1 image when you need 16:9 wastes a generation. Set this before clicking Generate.