Video Models

All video models share the same Banana Flow credit pool — switch between them without juggling provider accounts. Generation typically takes 30 seconds to 3 minutes depending on model and length.

Quick reference

Veo 3

Google’s flagship video model. Strong camera control (dolly, pan, tracking) and physically plausible motion. Great pick when you need the camera itself to do something specific.

Aspect ratios: 16:9, 9:16

Seedance 2.0

ByteDance’s multimodal video director. Accepts text, images, video clips, and audio as direction. Generates native synchronized audio (including lip-sync) without a separate pass.

• Multi-shot character consistency across cuts and camera angles

• Aspect ratios: 1:1, 4:3, 3:4, 16:9, 9:16, 21:9 (cinematic ultra-wide)

• Cost (per second, scales with duration):

  • 480p → 19 credits/s
  • 720p → 41 credits/s
  • 1080p → 102 credits/s

  At 12s, 720p that’s 492 credits; at 4s, 480p just 76 credits.

Pick this when: you need dialogue, multi-shot scenes, or cinematic ratios.

Kling 2.1

Best-in-class character motion. Faces, hands, and full-body action all hold up better than most competitors. Use it for clips where a person or creature is the focus.

Runway Gen-3

The original cinematic-style generator. Reliable for atmospheric short clips and stylized motion.

Hailuo 2.3

MiniMax’s video model. Lightweight, fast, and good at stylized rather than photoreal output.

Choosing a model

• Need a moving camera through a real-feeling scene? → Veo 3
• Need dialogue or multi-shot continuity? → Seedance 2.0
• Character/creature is the star? → Kling 2.1
• Stylized or atmospheric short? → Runway Gen-3 or Hailuo 2.3

You can also generate the same prompt across multiple Video nodes side by side to compare before committing more credits to your favorite.