{
  "keyframe": {
    "model": "openai/gpt-image-2/edit",
    "model_first_aerial_without_photos": "openai/gpt-image-2",
    "input": {
      "quality": "high",
      "num_images": 1,
      "output_format": "jpeg"
    },
    "image_sizes": {
      "9:16": {
        "width": 864,
        "height": 1536
      },
      "16:9": {
        "width": 1536,
        "height": 864
      },
      "9:16-1080p": {
        "width": 1152,
        "height": 2048
      },
      "16:9-1080p": {
        "width": 2048,
        "height": 1152
      }
    },
    "max_references": 3,
    "approx_cost_per_image": {
      "without_references": 0.25,
      "with_references": 0.27
    }
  },
  "video": {
    "draft": {
      "model": "bytedance/seedance-2.0/mini/image-to-video",
      "input": {
        "resolution": "480p",
        "duration": "4",
        "aspect_ratio": "auto",
        "generate_audio": false
      },
      "approx_cost_per_second": 0.0721
    },
    "final": {
      "model": "bytedance/seedance-2.5/image-to-video",
      "input": {
        "resolution": "720p",
        "duration": "4",
        "aspect_ratio": "auto",
        "generate_audio": true,
        "bitrate_mode": "high",
        "codec": "H264"
      },
      "approx_cost_per_second": {
        "720p": 0.473,
        "1080p": 1.164
      }
    },
    "alternatives_for_faces_or_failures": {
      "note": "Seedance on fal refuses clearly visible realistic faces. Tested 2026-10-16 on one face-forward synthetic close-up: Kling accepted it and held the face for 3 s; Veo accepted it but the face drifted within 4 s. Test one draft first; the look differs from Seedance.",
      "kling": {
        "model": "fal-ai/kling-video/v3/turbo/pro/image-to-video",
        "input": {
          "duration": "4"
        },
        "fields": "prompt, image_url, duration \"3\"-\"15\", optional multi_prompt; no resolution, end_image_url, generate_audio, bitrate_mode or codec; returned 1920x1080 with a sound track",
        "approx_cost_per_second": 0.14
      },
      "veo": {
        "model": "fal-ai/veo3.1/image-to-video",
        "input": {
          "duration": "4s",
          "resolution": "720p",
          "generate_audio": false
        },
        "fields": "prompt, image_url (16:9 or 9:16, 720p or more), duration \"4s\"/\"6s\"/\"8s\", resolution 720p/1080p, generate_audio, aspect_ratio, optional negative_prompt",
        "approx_cost_per_second": {
          "silent": 0.2,
          "with_audio": 0.4
        }
      }
    }
  },
  "defaults": {
    "aspect": "9:16",
    "shots": 5,
    "shot_seconds": 4,
    "final_resolution": "720p",
    "assembly": {
      "grain": 6,
      "bloom": "0.12 for blue-hour and night, 0.08 for daylight grades",
      "vignette": 0.3,
      "dissolve_seconds": 0.5,
      "fade_out_seconds": 0.5,
      "crf": 20,
      "preview_max_mb": 7.5,
      "fps": 24
    }
  },
  "cost_note": "fal list prices checked 2026-10-16. A 5-shot 720p teaser is about $12-16 including drafts and a re-roll: keyframes ~$1.35, drafts ~$1.45, finals ~$9.46. Seedance bills per video token (width x height x 24 x seconds / 1024). Ask before every paid step."
}