diff --git a/wayd/scripts/img2ascii.py b/wayd/scripts/img2ascii.py index 712acc2..284d90d 100644 --- a/wayd/scripts/img2ascii.py +++ b/wayd/scripts/img2ascii.py @@ -1,25 +1,15 @@ #!/usr/bin/env python3 """Convert an image to high-quality ASCII art for WAYD posts. -Produces results comparable to asciiart.eu: edge-enhanced, sharpened, -contrast-boosted, with a dense 70-char gradient. +Uses Jarvis-Judice-Ninke error diffusion dithering with a sparse character +set and brightness threshold — matching the asciiart.eu aesthetic: +light/background areas become spaces, only edges and shadows get chars. Usage: - img2ascii.py --image PATH [--width N] [--invert] [--caption TEXT] - [--edge-weight F] [--sharpen] [--contrast F] + img2ascii.py --image PATH [--width N] [--invert] [--threshold N] + [--contrast F] [--caption TEXT] Prints JSON to stdout: {"ok": true, "art": "...", "chars": N} - -Options: - --image PATH Image file (JPEG, PNG, GIF, WebP, DNG, …) - --width N Width in chars (default: 100). Height auto-calculated. - --invert Invert brightness (for light-background images). - --caption TEXT Text appended below the art (2 blank lines separator). - --edge-weight F How much edge detection to blend in, 0.0–1.0 (default 0.4). - Higher = more defined outlines like asciiart.eu. - --sharpen Apply sharpening before conversion (default: on). - --no-sharpen Disable sharpening. - --contrast F Contrast multiplier (default: 1.3). 1.0 = no change. """ from __future__ import annotations @@ -31,29 +21,23 @@ sys.path.insert(0, str(Path(__file__).resolve().parent)) import shared # noqa: E402 -# Dense 70-char gradient from dark to light (dark terminal). -# Derived from the classic Paulm gradient used by most quality converters. -_RAMP_DARK = r'$@B%8&WM#*oahkbdpqwmZO0QLCJUYXzcvunxrjft/\|()1{}[]?-_+~<>i!lI;:,"^`\'. ' +# Sparse ramp dark→light. Last entry is space (bright pixels). +_RAMP_DARK = "@%#*+:. " _RAMP_LIGHT = _RAMP_DARK[::-1] -def _px(luminance: int, ramp: str) -> str: - return ramp[int(luminance / 255 * (len(ramp) - 1))] - - def image_to_ascii( image_path: str, - width: int = 100, + width: int = 120, invert: bool = False, - edge_weight: float = 0.4, - sharpen: bool = True, - contrast: float = 1.3, + contrast: float = 1.5, + threshold: int = 200, caption: str = "", max_chars: int | None = None, ) -> str: - """Return ASCII art string. Raises ValueError on failure.""" + """Return ASCII art string using JJN dithering + threshold. Raises ValueError on failure.""" try: - from PIL import Image, ImageEnhance, ImageFilter # type: ignore[import] + from PIL import Image, ImageEnhance, ImageOps # type: ignore[import] except ImportError: raise ValueError("Pillow is required: pip install Pillow") @@ -64,25 +48,19 @@ def image_to_ascii( except Exception as exc: raise ValueError(f"Cannot open image: {exc}") - # Convert to RGB then grayscale (handles RGBA, P, CMYK, DNG/TIFF, etc.) img = img.convert("RGB") - # --- Preprocessing (mimics asciiart.eu quality pipeline) --- - - if sharpen: - img = ImageEnhance.Sharpness(img).enhance(2.0) - if contrast != 1.0: img = ImageEnhance.Contrast(img).enhance(contrast) gray = img.convert("L") - # Compute target height preserving aspect ratio. - # Terminal chars are ~2:1 tall:wide; 0.45 corrects for that. + # Equalize histogram so the full 0-255 range is used regardless of image tone + gray = ImageOps.equalize(gray) + aspect = img.height / img.width height = max(1, int(width * aspect * 0.45)) - # Shrink to fit max_chars budget if given (art-only chars, no caption). if max_chars is not None: caption_cost = len(caption) + 2 if caption else 0 while width >= 20: @@ -91,32 +69,41 @@ def image_to_ascii( width = int(width * 0.9) height = max(1, int(width * aspect * 0.45)) - # Resize both base image and edge map to the target size. - base = gray.resize((width, height), Image.LANCZOS) + gray = gray.resize((width, height), Image.LANCZOS) - # Edge detection: find edges on a slightly blurred version for clean lines. - edge_src = gray.filter(ImageFilter.GaussianBlur(1)).filter(ImageFilter.FIND_EDGES) - edges = edge_src.resize((width, height), Image.LANCZOS) + ramp = _RAMP_LIGHT if invert else _RAMP_DARK + n_chars = len(ramp) - 1 # last slot = space, reserved for threshold + step = threshold / n_chars - # Blend: final_lum = base * (1 - w) + edges * w - # Edge pixels push luminance toward dark (dense chars = outlines). - base_px = base.tobytes() - edge_px = edges.tobytes() + px = [[float(gray.getpixel((x, y))) for x in range(width)] for y in range(height)] - ramp = _RAMP_LIGHT if invert else _RAMP_DARK lines: list[str] = [] - - for row in range(height): - chars = [] - for col in range(width): - idx = row * width + col - b = base_px[idx] - e = edge_px[idx] - # Invert edge contribution so edges → darker chars (denser). - blended = int(b * (1 - edge_weight) + (255 - e) * edge_weight) - blended = max(0, min(255, blended)) - chars.append(_px(blended, ramp)) - lines.append("".join(chars)) + for y in range(height): + row: list[str] = [] + for x in range(width): + old = max(0.0, min(255.0, px[y][x])) + + if old >= threshold: + # Bright pixel → space, distribute error + err = old - 255.0 + char = " " + else: + level = min(n_chars - 1, int(old / step)) + err = old - level * step + char = ramp[level] + + # Jarvis-Judice-Ninke error diffusion kernel (denominator 48) + for dy, dx, w in ( + (0, 1, 7), (0, 2, 5), + (1, -2, 3), (1, -1, 5), (1, 0, 7), (1, 1, 5), (1, 2, 3), + (2, -2, 1), (2, -1, 3), (2, 0, 5), (2, 1, 3), (2, 2, 1), + ): + ny, nx = y + dy, x + dx + if 0 <= ny < height and 0 <= nx < width: + px[ny][nx] += err * w / 48.0 + + row.append(char) + lines.append("".join(row).rstrip()) art = "\n".join(lines) if caption: @@ -125,15 +112,13 @@ def image_to_ascii( def main() -> None: - parser = argparse.ArgumentParser(description="High-quality image → ASCII art.") + parser = argparse.ArgumentParser(description="High-quality image → ASCII art (JJN dithering).") parser.add_argument("--image", required=True) - parser.add_argument("--width", type=int, default=100) + parser.add_argument("--width", type=int, default=120) parser.add_argument("--invert", action="store_true") parser.add_argument("--caption", default="") - parser.add_argument("--edge-weight", type=float, default=0.4) - parser.add_argument("--sharpen", dest="sharpen", action="store_true", default=True) - parser.add_argument("--no-sharpen", dest="sharpen", action="store_false") - parser.add_argument("--contrast", type=float, default=1.3) + parser.add_argument("--contrast", type=float, default=1.5) + parser.add_argument("--threshold", type=int, default=200) parser.add_argument("--max-chars", type=int, default=None) args = parser.parse_args() @@ -142,9 +127,8 @@ def main() -> None: image_path=args.image, width=args.width, invert=args.invert, - edge_weight=args.edge_weight, - sharpen=args.sharpen, contrast=args.contrast, + threshold=args.threshold, caption=args.caption, max_chars=args.max_chars, ) diff --git a/wayd/scripts/post.py b/wayd/scripts/post.py index 5b9e8df..1473dc1 100644 --- a/wayd/scripts/post.py +++ b/wayd/scripts/post.py @@ -85,7 +85,7 @@ def cmd_check_rate_limit(_: argparse.Namespace) -> None: def _convert_image_to_art(image_path: str, text_len: int, max_chars: int) -> str | None: - """Convert image to ASCII art sized to fit within the remaining char budget. + """Convert image to ASCII art. Art is exempt from max_chars, so no budget cap. Returns the art string, or None if conversion fails (caller logs the error and continues without art so the post still goes through). @@ -97,14 +97,8 @@ def _convert_image_to_art(image_path: str, text_len: int, max_chars: int) -> str shared.log_error("img2ascii not found; skipping image conversion") return None - # Reserve chars for: text + 2 newlines + art block markers (~20 chars overhead) - budget = max_chars - text_len - 20 - if budget < 50: - shared.log_error("not enough char budget for ASCII art after text") - return None - try: - return image_to_ascii(image_path=image_path, max_chars=budget) + return image_to_ascii(image_path=image_path, width=120) except Exception as exc: shared.log_error(f"img2ascii failed for {image_path!r}: {exc}") return None diff --git a/wayd/scripts/shared.py b/wayd/scripts/shared.py index 7f1ee73..4872774 100644 --- a/wayd/scripts/shared.py +++ b/wayd/scripts/shared.py @@ -266,13 +266,17 @@ def remove_blocked(username: str) -> bool: MARKER_RE = re.compile(r"") -# Matches the ASCII art block: +# Matches the ASCII art block in
format: +#
...
...
...
# Group 1 = the art content (may be multi-line). ART_MARKER_RE = re.compile( - r"", + r"
.*?
(.*?)
.*?
", re.DOTALL, ) +# Also match the old HTML-comment format for backward compat with old posts. +_ART_COMMENT_RE = re.compile(r"", re.DOTALL) + def build_post_title(vibe_slug: str, vibe_emoji: str, body: str) -> str: """Build the issue title: '[ ] '.""" @@ -290,30 +294,43 @@ def build_post_body(vibe_slug: str, text: str, marker_version: str = "v1") -> st def build_post_body_with_art( vibe_slug: str, text: str, art: str, marker_version: str = "v1" ) -> str: - """Build issue body with embedded ASCII art in an HTML comment block. + """Build issue body with ASCII art in a collapsible
block. - The art block is invisible when GitHub renders the issue but parseable - by scroll.py. Format: + Visible on GitHub as '📎 ASCII image' (click to expand), and parseable + by scroll.py via ART_MARKER_RE. Uses 
 for monospace rendering.
+    Format:
         
 
-        
+
+ +
+ """ - art_block = f"" + art_block = f"
\n📎 ASCII image\n\n
\n{art}\n
\n\n
" return f"{text.strip()}\n\n{art_block}\n\n" def extract_art(body: str) -> str | None: """Return the ASCII art embedded in body, or None if not present.""" m = ART_MARKER_RE.search(body) - return m.group(1) if m else None + if m: + return m.group(1) + # Backward compat: old posts used HTML comment format + m2 = _ART_COMMENT_RE.search(body) + return m2.group(1) if m2 else None def strip_art(body: str) -> str: """Return body with the ASCII art block removed (for display purposes).""" - return ART_MARKER_RE.sub("", body).strip() + body = ART_MARKER_RE.sub("", body) + body = _ART_COMMENT_RE.sub("", body) + return body.strip() def parse_post_body(body: str) -> dict[str, Any]: