diff --git a/clients/gpt_image_client.py b/clients/gpt_image_client.py index de9012f..8fe738f 100644 --- a/clients/gpt_image_client.py +++ b/clients/gpt_image_client.py @@ -231,11 +231,11 @@ class GptImageClient: size: str, n: int, seed: int, - image_tensor: Optional[torch.Tensor] = None, + image_list: Optional[List[torch.Tensor]] = None, ) -> List[Image.Image]: """ 调用 /v1/images/generations/ 接口。 - 当传入 image_tensor 时，以 data URI 格式内联图像（图生图）。 + 当传入 image_list 时，以 data URI 格式内联图像（图生图）。 """ # 模型名映射：UI 显示名 → API 参数名 api_model = _MODEL_NAME_MAP.get(model, model) @@ -251,22 +251,23 @@ class GptImageClient: body["size"] = size if size else "auto" - # 图生图：将 tensor 转成 data URI 内联 - if image_tensor is not None: - pil_images = tensor_to_pil(image_tensor) + # 图生图：将 tensor 列表转成 data URI 内联 + if image_list is not None: data_urls = [] - for idx_img, img in enumerate(pil_images): + for idx_img, img_tensor in enumerate(image_list): + pil_images = tensor_to_pil(img_tensor) + img = pil_images[0] buf = BytesIO() img.save(buf, format="PNG") png_bytes = buf.getvalue() # 单张图像预算：20MB 按图数平摊，至少保留 1MB 给其他字段 per_image_budget = max( 1024 * 1024, - (self._MAX_BODY_BYTES - 1024 * 1024) // len(pil_images), + (self._MAX_BODY_BYTES - 1024 * 1024) // len(image_list), ) # base64 膨胀约 4/3，所以 PNG 目标上限 = budget * 3/4 png_budget = int(per_image_budget * 3 / 4) - label = f"第{idx_img + 1}张" if len(pil_images) > 1 else "" + label = f"第{idx_img + 1}张" if len(image_list) > 1 else "" png_bytes = self._shrink_png_to_limit(png_bytes, png_budget, label) b64 = base64.b64encode(png_bytes).decode("utf-8") data_urls.append(f"data:image/png;base64,{b64}") @@ -322,7 +323,7 @@ class GptImageClient: size: str, n: int, seed: int, - image_tensor: torch.Tensor, + image_list: List[torch.Tensor], mask_tensor: Optional[torch.Tensor] = None, ) -> List[Image.Image]: """ @@ -333,10 +334,13 @@ class GptImageClient: # 模型名映射：UI 显示名 → API 参数名 api_model = _MODEL_NAME_MAP.get(model, model) - # 将 batch tensor 拆成逐帧列表 - if image_tensor.dim() == 3: - image_tensor = image_tensor.unsqueeze(0) # [H,W,C] → [1,H,W,C] - num_images = image_tensor.shape[0] + # 统一 tensors 为 [1,H,W,C] 格式，支持不同尺寸 + normalized_tensors = [] + for t in image_list: + if t.dim() == 3: + t = t.unsqueeze(0) # [H,W,C] → [1,H,W,C] + normalized_tensors.append(t) + num_images = len(normalized_tensors) # o1key 中转服务的 edits 接口暂不支持 background / moderation / seed， # 待服务方更新后可重新加入。 @@ -355,8 +359,7 @@ class GptImageClient: 1024 * 1024, (self._MAX_BODY_BYTES - mask_reserve) // num_images, ) - for i in range(num_images): - frame = image_tensor[i:i+1] # [1,H,W,C] + for i, frame in enumerate(normalized_tensors): img_bytes = self._tensor_to_png_bytes(frame) label = f"第{i + 1}张" if num_images > 1 else "" img_bytes = self._shrink_png_to_limit(img_bytes, per_image_budget, label) @@ -367,7 +370,9 @@ class GptImageClient: content_type="image/png", ) - ih, iw = image_tensor.shape[1], image_tensor.shape[2] + # 蒙版尺寸校验以第一张图为基准 + first_tensor = normalized_tensors[0] + ih, iw = first_tensor.shape[1], first_tensor.shape[2] if mask_tensor is not None: mask_png = self._mask_tensor_to_rgba_png_bytes(mask_tensor, (ih, iw)) @@ -429,7 +434,7 @@ class GptImageClient: size: str, n: int, seed: int, - image_tensor: Optional[torch.Tensor] = None, + image_tensor: Optional[List[torch.Tensor]] = None, mask_tensor: Optional[torch.Tensor] = None, ) -> List[Image.Image]: """ @@ -447,13 +452,13 @@ class GptImageClient: coro = self._edit_async( prompt=prompt, model=model, quality=quality, background=background, size=size, n=n, seed=seed, - image_tensor=image_tensor, mask_tensor=mask_tensor, + image_list=image_tensor, mask_tensor=mask_tensor, ) else: coro = self._generate_async( prompt=prompt, model=model, quality=quality, background=background, size=size, n=n, seed=seed, - image_tensor=image_tensor, + image_list=image_tensor, ) def _run(): diff --git a/nodes/gpt_image.py b/nodes/gpt_image.py index 4d2e4a6..acbe0b4 100644 --- a/nodes/gpt_image.py +++ b/nodes/gpt_image.py @@ -4,8 +4,6 @@ o1key GPT Image 节点 """ import time -import torch - from ..clients.gpt_image_client import GptImageClient from ..utils.image_utils import parse_batch_prompts @@ -132,10 +130,7 @@ class O1keyGPTImage: if key in kwargs and kwargs[key] is not None: reference_tensors.append(kwargs[key]) - if reference_tensors: - 图片 = torch.cat(reference_tensors, dim=0) - else: - 图片 = None + 图片 = reference_tensors if reference_tensors else None # ── 1. 参数校验 ─────────────────────────────────────────────────────── if 遮罩 is not None and 图片 is None: