estimateImageTokens method

int estimateImageTokens(int width, int height, { String detail = 'auto', })

Estimates the token cost of an image based on its dimensions and detail level.

Uses the tile-based formula from OpenAI / Anthropic documentation:

• low detail: fixed 85 tokens
• high detail: 170 tokens per 512x512 tile + 85 base

Implementation

int estimateImageTokens(int width, int height, {String detail = 'auto'}) {
  if (detail == 'low') return 85;
  // Scale down to fit within 2048x2048 then tile at 512x512.
  var w = width.toDouble();
  var h = height.toDouble();
  if (w > 2048 || h > 2048) {
    final scale = 2048 / math.max(w, h);
    w = (w * scale).ceilToDouble();
    h = (h * scale).ceilToDouble();
  }
  // Scale shortest side to 768.
  final minSide = math.min(w, h);
  if (minSide > 768) {
    final scale = 768 / minSide;
    w = (w * scale).ceilToDouble();
    h = (h * scale).ceilToDouble();
  }
  final tilesX = (w / 512).ceil();
  final tilesY = (h / 512).ceil();
  return 170 * tilesX * tilesY + 85;
}