Image model picker
Ask one question: “Is this draft or delivery?”
- Draft / explore / many variations → Nano Banana 2
- Hero asset / product accuracy / precise edit → GPT Image 2
- Heavy reference editing with strict file types → GPT Image 2 (JPEG/PNG/WebP, 10 MB each)
Text-only video
If you only have words:
- Cinematic 4–8s spot, 16:9 or 9:16, up to 4K → VEO 3.1
- Longer clip (up to 15s) or more aspect choices → Seedance 2.0
Reference-driven video
If you have structured references:
- Start + end stills → Image to Video (try VEO for filmic motion, Seedance for flexible duration)
- Mood board of many stills → Multi-images to Video (Seedance)
- Motion reference clips → Video to Video (Seedance only)
Credit-aware modeling
Prototype with the cheaper/faster configuration (shorter duration, 480p/720p, sound off), then rerun once with premium settings.
The Generate button always shows the cost before you commit.
Category: AI models