Skip to main content
Multiple Image-to-Image models use reference images combined with text prompts to generate refined or stylized versions. In Automat Studio, these models power the “Generate First Frame” workflow in Shot Studio, allowing you to iterate on shot compositions using previous frames or reference images.
These models are used in the Generate First Frame tab of Shot Studio to iterate on shot compositions.

When to Use

  • First Frame Generation — Create the hero still for shots using shot descriptions and visual references
  • Style Consistency — Maintain visual coherence across shots by referencing previous frames
  • Iterative Refinement — Build upon successful generations to perfect compositions
  • Shot Matching — Align new shots with existing visual style from the project
These models consistently deliver the best results for shot frame generation:

Google Nano Banana 2.1

  • Credits: 12
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: Google
  • Best for: Consistent characters and objects composed from up to 14 references (Default model)

Bytedance Seedream V5 Lite Edit

  • Credits: 6
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: fal.ai
  • Best for: Cost-effective, high-quality refinement

Seedream 5.5 Edit

  • Credits: 12
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: fal.ai
  • Best for: Multi-reference compositions from up to 10 images

FLUX.2 [pro]

  • Credits: 5
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: fal.ai
  • Avg. Duration: ~33 seconds
  • Best for: Fast, high-quality editing with excellent prompt adherence

FLUX 3 Image Edit

  • Credits: 8
  • Rating: ⭐⭐⭐⭐ (4/5)
  • Provider: fal.ai
  • Best for: Balanced compositions from up to 10 references

Google Nano Banana Pro

  • Credits: 21
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: Google
  • Avg. Duration: ~35 seconds
  • Best for: Premium quality image editing with exceptional detail

OpenAI GPT Image 2

  • Credits: 29
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: OpenAI
  • Best for: Complex compositions with excellent prompt comprehension; superseded by GPT Image 2.6 for speed and cost

OpenAI GPT Image 2.6 Flare

  • Credits: 12
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: OpenAI
  • Best for: Composes characters, wardrobe and locations into one frame, at speed

OpenAI GPT Image 2.6 Sunburst

  • Credits: 12
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: OpenAI
  • Best for: The most controlled multi-reference compositions; slower than Flare

Seedream 4.5

  • Credits: 6
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: fal.ai
  • Avg. Duration: ~53 seconds
  • Best for: High-quality editing with excellent detail

Qwen Image Edit Plus

  • Credits: 5
  • Rating: ⭐⭐⭐⭐ (4/5)
  • Provider: fal.ai
  • Avg. Duration: ~20 seconds
  • Best for: Cost-effective editing and refinement

Reve Remix

  • Credits: 6
  • Rating: ⭐⭐⭐⭐ (4/5)
  • Provider: fal.ai
  • Avg. Duration: ~38 seconds
  • Best for: Creative reinterpretations with high quality

Vidu Reference-to-Image

  • Credits: 15
  • Rating: ⭐⭐⭐⭐ (4/5)
  • Provider: fal.ai
  • Avg. Duration: ~49 seconds
  • Best for: Working with video reference frames

Supported Models

These models are available but may have varying quality or processing times:

OmniGen v2

  • Credits: 45
  • Rating: ⭐⭐⭐ (3/5)
  • Provider: fal.ai
  • Avg. Duration: ~116 seconds
  • Use when: Exploring alternative generation methods

Tips for Best Results

  1. Use Clear References — Provide high-quality reference images that clearly show what you want
  2. Combine with Prompts — Text descriptions enhance reference image understanding
  3. Iterate Systematically — Build on successful generations rather than starting from scratch
  4. Maintain Shot History — Save multiple variations to compare and choose the best
  5. Match Camera Settings — Reference frames work best when shot attributes (lens, angle) are similar

Generate First Frame

Learn how to use these models in Shot Studio

Shot Studio

Complete guide to shot creation and generation