| 1 | #!/usr/bin/env python3 |
| 2 | """ |
| 3 | MiniMax image generation backend. |
| 4 | |
| 5 | Configuration keys: |
| 6 | MINIMAX_API_KEY (required) |
| 7 | MINIMAX_BASE_URL (optional) |
| 8 | MINIMAX_MODEL (optional) |
| 9 | """ |
| 10 | |
| 11 | import sys |
| 12 | from pathlib import Path |
| 13 | |
| 14 | _SCRIPTS_DIR = Path(__file__).resolve().parents[1] |
| 15 | if str(_SCRIPTS_DIR) not in sys.path: |
| 16 | sys.path.insert(0, str(_SCRIPTS_DIR)) |
| 17 | |
| 18 | from console_encoding import configure_utf8_stdio # noqa: E402 |
| 19 | |
| 20 | configure_utf8_stdio() |
| 21 | |
| 22 | if __name__ == "__main__": |
| 23 | print(__doc__) |
| 24 | print("Use via: python3 skills/ppt-master/scripts/image_gen.py \"prompt\" --backend minimax") |
| 25 | raise SystemExit(0 if any(arg in {"-h", "--help", "help"} for arg in sys.argv[1:]) else 1) |
| 26 | |
| 27 | import base64 |
| 28 | import os |
| 29 | import time |
| 30 | |
| 31 | import requests |
| 32 | |
| 33 | from image_backends.backend_common import ( |
| 34 | MAX_RETRIES, |
| 35 | detect_image_extension, |
| 36 | http_error, |
| 37 | is_rate_limit_error, |
| 38 | normalize_image_size, |
| 39 | require_api_key, |
| 40 | resolve_output_path, |
| 41 | retry_delay, |
| 42 | save_image_bytes, |
| 43 | ) |
| 44 | |
| 45 | |
| 46 | DEFAULT_ENDPOINT = "https://api.minimaxi.com/v1/image_generation" |
| 47 | DEFAULT_MODEL = "image-01" |
| 48 | |
| 49 | # International fallback: set MINIMAX_BASE_URL=https://api.minimax.io if needed |
| 50 | |
| 51 | ASPECT_RATIO_SIZE_MAP = { |
| 52 | "512px": { |
| 53 | "1:1": (512, 512), |
| 54 | "16:9": (640, 360), |
| 55 | "4:3": (576, 432), |
| 56 | "3:2": (624, 416), |
| 57 | "2:3": (416, 624), |
| 58 | "3:4": (432, 576), |
| 59 | "9:16": (360, 640), |
| 60 | "21:9": (672, 288), |
| 61 | }, |
| 62 | "1K": { |
| 63 | "1:1": (1024, 1024), |
| 64 | "16:9": (1280, 720), |
| 65 | "4:3": (1152, 864), |
| 66 | "3:2": (1248, 832), |
| 67 | "2:3": (832, 1248), |
| 68 | "3:4": (864, 1152), |
| 69 | "9:16": (720, 1280), |
| 70 | "21:9": (1344, 576), |
| 71 | }, |
| 72 | "2K": { |
| 73 | "1:1": (2048, 2048), |
| 74 | "16:9": (2048, 1152), |
| 75 | "4:3": (2048, 1536), |
| 76 | "3:2": (2048, 1368), |
| 77 | "2:3": (1368, 2048), |
| 78 | "3:4": (1536, 2048), |
| 79 | "9:16": (1152, 2048), |
| 80 | "21:9": (2048, 880), |
| 81 | }, |
| 82 | "4K": { |
| 83 | "1:1": (2048, 2048), |
| 84 | "16:9": (2048, 1152), |
| 85 | "4:3": (2048, 1536), |
| 86 | "3:2": (2048, 1368), |
| 87 | "2:3": (1368, 2048), |
| 88 | "3:4": (1536, 2048), |
| 89 | "9:16": (1152, 2048), |
| 90 | "21:9": (2048, 880), |
| 91 | }, |
| 92 | } |
| 93 | |
| 94 | |
| 95 | def _resolve_url(base_url: str) -> str: |
| 96 | """Resolve the MiniMax image generation endpoint. |
| 97 | |
| 98 | Accepts three forms of MINIMAX_BASE_URL: |
| 99 | - Full endpoint: https://api.minimax.io/v1/image_generation → used as-is |
| 100 | - Versioned base: https://api.minimax.io/v1 → appends /image_generation |
| 101 | - Root base: https://api.minimax.io → appends /v1/image_generation |
| 102 | """ |
| 103 | base = base_url.rstrip("/") |
| 104 | if base.endswith("/image_generation"): |
| 105 | return base |
| 106 | if base.endswith("/v1"): |
| 107 | return base + "/image_generation" |
| 108 | return base + "/v1/image_generation" |
| 109 | |
| 110 | |
| 111 | def _resolve_dimensions(aspect_ratio: str, image_size: str) -> tuple[int, int]: |
| 112 | """Resolve width and height from the unified aspect_ratio/image_size pair.""" |
| 113 | normalized = normalize_image_size(image_size) |
| 114 | dimensions = (ASPECT_RATIO_SIZE_MAP.get(normalized) or {}).get(aspect_ratio) |
| 115 | if not dimensions: |
| 116 | supported = sorted(ASPECT_RATIO_SIZE_MAP["1K"]) |
| 117 | raise ValueError( |
| 118 | f"Unsupported aspect ratio '{aspect_ratio}' for MiniMax backend. " |
| 119 | f"Supported: {supported}" |
| 120 | ) |
| 121 | return dimensions |
| 122 | |
| 123 | |
| 124 | def _extract_image_bytes(payload: dict) -> bytes | None: |
| 125 | """Extract image bytes from a MiniMax response payload.""" |
| 126 | data = payload.get("data") or {} |
| 127 | image_base64 = data.get("image_base64") or [] |
| 128 | if image_base64: |
| 129 | return base64.b64decode(image_base64[0]) |
| 130 | return None |
| 131 | |
| 132 | |
| 133 | def _generate_image(api_key: str, prompt: str, |
| 134 | aspect_ratio: str = "1:1", image_size: str = "1K", |
| 135 | output_dir: str = None, filename: str = None, |
| 136 | model: str = DEFAULT_MODEL, base_url: str = DEFAULT_ENDPOINT) -> str: |
| 137 | """Generate one image with the MiniMax backend.""" |
| 138 | width, height = _resolve_dimensions(aspect_ratio, image_size) |
| 139 | url = _resolve_url(base_url) |
| 140 | |
| 141 | headers = { |
| 142 | "Authorization": f"Bearer {api_key}", |
| 143 | "Content-Type": "application/json", |
| 144 | } |
| 145 | payload = { |
| 146 | "model": model, |
| 147 | "prompt": prompt, |
| 148 | "width": width, |
| 149 | "height": height, |
| 150 | "response_format": "base64", |
| 151 | "n": 1, |
| 152 | } |
| 153 | |
| 154 | print("[MiniMax Image]") |
| 155 | print(f" Model: {model}") |
| 156 | print(f" Prompt: {prompt[:120]}{'...' if len(prompt) > 120 else ''}") |
| 157 | print(f" Aspect Ratio: {aspect_ratio}") |
| 158 | print(f" Resolution: {width}x{height} (from image_size={image_size})") |
| 159 | print() |
| 160 | print(" [..] Generating...", end="", flush=True) |
| 161 | start = time.time() |
| 162 | response = requests.post(url, headers=headers, json=payload, timeout=300) |
| 163 | elapsed = time.time() - start |
| 164 | print(f"\n [DONE] Response received ({elapsed:.1f}s)") |
| 165 | |
| 166 | if response.status_code != 200: |
| 167 | raise http_error(response, "MiniMax image generation") |
| 168 | |
| 169 | data = response.json() |
| 170 | base_resp = data.get("base_resp") or {} |
| 171 | if base_resp.get("status_code") not in (None, 0, "0"): |
| 172 | raise RuntimeError(f"MiniMax image generation failed: {data}") |
| 173 | |
| 174 | image_bytes = _extract_image_bytes(data) |
| 175 | if not image_bytes: |
| 176 | raise RuntimeError(f"MiniMax response missing image data: {data}") |
| 177 | |
| 178 | ext = detect_image_extension(image_bytes) or ".jpeg" |
| 179 | path = resolve_output_path(prompt, output_dir, filename, ext) |
| 180 | return save_image_bytes(image_bytes, path) |
| 181 | |
| 182 | |
| 183 | def generate(prompt: str, |
| 184 | aspect_ratio: str = "1:1", image_size: str = "1K", |
| 185 | output_dir: str = None, filename: str = None, |
| 186 | model: str = None, max_retries: int = MAX_RETRIES) -> str: |
| 187 | """Generate an image with retries using the MiniMax backend.""" |
| 188 | api_key = require_api_key( |
| 189 | "MINIMAX_API_KEY", |
| 190 | message="No API key found. Set MINIMAX_API_KEY in the current environment or a .env file.", |
| 191 | ) |
| 192 | base_url = os.environ.get("MINIMAX_BASE_URL") or DEFAULT_ENDPOINT |
| 193 | resolved_model = model or os.environ.get("MINIMAX_MODEL") or DEFAULT_MODEL |
| 194 | normalized_size = normalize_image_size(image_size) |
| 195 | |
| 196 | last_error = None |
| 197 | for attempt in range(max_retries + 1): |
| 198 | try: |
| 199 | return _generate_image( |
| 200 | api_key=api_key, |
| 201 | prompt=prompt, |
| 202 | aspect_ratio=aspect_ratio, |
| 203 | image_size=normalized_size, |
| 204 | output_dir=output_dir, |
| 205 | filename=filename, |
| 206 | model=resolved_model, |
| 207 | base_url=base_url, |
| 208 | ) |
| 209 | except Exception as exc: |
| 210 | last_error = exc |
| 211 | if attempt >= max_retries: |
| 212 | break |
| 213 | limited = is_rate_limit_error(exc) |
| 214 | delay = retry_delay(attempt, rate_limited=limited) |
| 215 | label = "Rate limit hit" if limited else f"Error: {exc}" |
| 216 | print(f"\n [WARN] {label}. Retrying in {delay}s...") |
| 217 | time.sleep(delay) |
| 218 | |
| 219 | raise RuntimeError(f"Failed after {max_retries + 1} attempts. Last error: {last_error}") |
| 220 |