import os import subprocess import folder_paths import torch import numpy as np from PIL import Image import shutil import datetime import platform class VideoCompressorUnified: """ Node de compression vidéo unifié : - Accepte images ou vidéo - Trois modes : fast_crf, 2pass, gpu """ @classmethod def INPUT_TYPES(s): return { "required": { "mode": (["fast_crf", "2pass", "gpu"], {"default": "2pass"}), "target_mb": ("FLOAT", {"min": 0.5, "max": 2000.0, "step": 0.5, "default": 8.0}), "codec": (["libx264", "libx265", "h264_nvenc", "hevc_nvenc"], {"default": "libx264"}), }, "optional": { "images": ("IMAGE",), "video_path_in": ("STRING", {"forceInput": True}), "fps": ("INT", {"default": 24, "min": 1, "max": 60}), "crf": ("INT", {"default": 23, "min": 10, "max": 40}), "preset": ([ "ultrafast","superfast","veryfast","faster", "fast","medium","slow","slower","veryslow" ], {"default": "fast"}), } } RETURN_TYPES = ("STRING", "IMAGE") RETURN_NAMES = ("video_path_out", "images") FUNCTION = "process" CATEGORY = "🎥Video Utils" def _get_video_info(self, video_path): command = [ "ffprobe", "-v", "error", "-select_streams", "v:0", "-show_entries", "stream=r_frame_rate,duration", "-of", "csv=p=0", video_path ] result = subprocess.run(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, check=True) parts = result.stdout.strip().split(',') if len(parts) < 2: raise ValueError("Impossible d'analyser la vidéo.") fps_fraction = parts[0].split('/') fps = float(fps_fraction[0]) / float(fps_fraction[1]) if len(fps_fraction) == 2 else 30.0 duration = float(parts[1]) return {"duration": duration, "fps": fps} def _load_video_to_tensor(self, video_path, temp_dir): frames_dir = os.path.join(temp_dir, "loaded_frames") os.makedirs(frames_dir, exist_ok=True) subprocess.run( ["ffmpeg", "-i", video_path, os.path.join(frames_dir, "frame_%05d.png")], check=True, capture_output=True ) frame_files = sorted([os.path.join(frames_dir, f) for f in os.listdir(frames_dir) if f.endswith('.png')]) images = [torch.from_numpy(np.array(Image.open(f).convert("RGB")).astype(np.float32) / 255.0).unsqueeze(0) for f in frame_files] if not images: return torch.empty(0) return torch.cat(images, 0) def process(self, mode, target_mb, codec, images=None, video_path_in=None, fps=24, crf=23, preset="fast"): if images is None and video_path_in is None: return ("", torch.empty(0)) temp_dir = os.path.join(folder_paths.get_temp_directory(), "compressor_unified") shutil.rmtree(temp_dir, ignore_errors=True) os.makedirs(temp_dir, exist_ok=True) pass_log_file = os.path.join(temp_dir, "ffmpeg_passlog") try: if images is not None: output_images = images for i, img_tensor in enumerate(images): Image.fromarray((255. * img_tensor.cpu().numpy()).astype(np.uint8)).save(os.path.join(temp_dir, f"frame_{i:05d}.png")) initial_video_path = os.path.join(temp_dir, "initial_from_images.mp4") subprocess.run( ["ffmpeg", "-y", "-framerate", str(fps), "-i", os.path.join(temp_dir, "frame_%05d.png"), "-c:v", "libx264", "-pix_fmt", "yuv420p", initial_video_path], check=True, capture_output=True ) else: if not os.path.exists(video_path_in): raise FileNotFoundError(f"Fichier vidéo non trouvé: {video_path_in}") initial_video_path = video_path_in output_images = self._load_video_to_tensor(video_path_in, temp_dir) output_dir = os.path.join(folder_paths.get_output_directory(), "compressed_videos") os.makedirs(output_dir, exist_ok=True) timestamp = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") filename = f"compressed_{timestamp}.mp4" final_video_path = os.path.join(output_dir, filename) info = self._get_video_info(initial_video_path) duration_s = info["duration"] if duration_s <= 0: raise ValueError("Durée de la vidéo nulle ou invalide.") # ----- Choix du mode ----- if mode == "2pass": # Calcule le bitrate cible target_size_bits = target_mb * 1024 * 1024 * 8 audio_bitrate_bits = 128 * 1000 video_bitrate_bits = (target_size_bits / duration_s) - audio_bitrate_bits if video_bitrate_bits < 100 * 1000: video_bitrate_bits = 100 * 1000 video_bitrate_k = int(video_bitrate_bits / 1000) null_output = "NUL" if platform.system() == "Windows" else "/dev/null" common_args = ["-y", "-i", initial_video_path, "-c:v", codec, "-b:v", f"{video_bitrate_k}k", "-preset", preset, "-threads", "0"] # Première passe subprocess.run(["ffmpeg"] + common_args + ["-pass", "1", "-an", "-f", "mp4", "-passlogfile", pass_log_file, null_output], check=True, capture_output=True) # Deuxième passe subprocess.run(["ffmpeg"] + common_args + ["-pass", "2", "-c:a", "aac", "-b:a", "128k", "-passlogfile", pass_log_file, final_video_path], check=True, capture_output=True) elif mode == "fast_crf": subprocess.run([ "ffmpeg", "-y", "-i", initial_video_path, "-c:v", codec, "-preset", preset, "-crf", str(crf), "-c:a", "aac", "-b:a", "128k", "-threads", "0", final_video_path ], check=True, capture_output=True) elif mode == "gpu": # Ici on force un CRF rapide avec encodeur GPU subprocess.run([ "ffmpeg", "-y", "-i", initial_video_path, "-c:v", codec, "-preset", preset, "-cq", str(crf), "-c:a", "aac", "-b:a", "128k", "-threads", "0", final_video_path ], check=True, capture_output=True) else: raise ValueError(f"Mode inconnu: {mode}") return (final_video_path, output_images) finally: shutil.rmtree(temp_dir, ignore_errors=True)