Spaces:
Running
Running
File size: 20,038 Bytes
f7598da |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 |
import os
import uuid
import statistics
def create_premiere_xml(project_name, video_path, overlay_segments, duration_frames, width=1080, height=1920, timebase=30, video_file_id=None, audio_file_id=None, scale_value=100.0, face_data=None, source_width=1920, source_height=1080):
"""
Generates a Premiere Pro XML with segmented cuts, supporting Dual-Track (Split Screen) for multi-face scenarios.
"""
def get_uid(): return str(uuid.uuid4())[:12]
if not video_file_id: video_file_id = f"file-video-{get_uid()}"
if not audio_file_id: audio_file_id = f"file-audio-{get_uid()}"
sequence_uuid = str(uuid.uuid4())
# helper for file blocks
def get_file_block(fid, fpath, is_audio_only=False):
audio_blk = "" if is_audio_only else "<audio><samplecharacteristics><depth>16</depth><samplerate>48000</samplerate></samplecharacteristics><channelcount>2</channelcount></audio>"
width_f = int(source_width)
height_f = int(source_height)
return f"""<file id="{fid}"><name>{os.path.basename(fpath)}</name><pathurl>{fpath}</pathurl><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate><duration>{duration_frames}</duration><media><video><samplecharacteristics><width>{width_f}</width><height>{height_f}</height><alpha>straight</alpha></samplecharacteristics></video>{audio_blk}</media></file>"""
# --- PROCESS FACE DATA (Per Frame) ---
# We store raw faces per frame to decide clustering later
faces_per_frame = {}
# Dimensions for Coordinate Normalization (Default to source if not in JSON)
coords_w = source_width
coords_h = source_height
if face_data:
# Check for Metadata in first entry to determine Coordinate System Scale
if len(face_data) > 0:
first_entry = face_data[0]
if "src_size" in first_entry:
try:
w_json, h_json = first_entry["src_size"]
if w_json > 0 and h_json > 0:
coords_w = w_json
coords_h = h_json
print(f"Coordinate System Reference: {coords_w}x{coords_h}")
# DO NOT overwrite source_width/source_height (Actual Media Dims)
except: pass
print(f"Processing {len(face_data)} face entries for Dual-Track logic...")
for entry in face_data:
f_idx = entry.get('frame')
faces = entry.get('faces', [])
if not faces: continue
processed_faces = []
for f in faces:
cx = (f[0] + f[2]) / 2.0
cy = (f[1] + f[3]) / 2.0
area = (f[2]-f[0]) * (f[3]-f[1])
# Calculate Normalized Center using COORDS Dimensions
# nx, ny are 0..1 relative to the original detection frame
nx = cx / max(1.0, float(coords_w))
ny = cy / max(1.0, float(coords_h))
# rh uses coords_h
rh_val = 0.1
if len(f) > 4:
rh_val = float(f[4])
else:
rh_val = (f[3] - f[1]) / max(1.0, float(coords_h))
processed_faces.append({
'cx': cx,
'cy': cy,
'nx': nx,
'ny': ny,
'area': area,
'rh': rh_val
})
faces_per_frame[f_idx] = processed_faces
# Ensure source_width/height are floats for calculation later
source_width = float(source_width)
source_height = float(source_height)
# --- SEGMENTATION LOGIC ---
cuts_v1 = [] # Track 1 (Main / Left)
cuts_v2 = [] # Track 2 (Secondary / Right)
fps_float = float(timebase)
# Store dynamic scale suggestion per cut if possible
# (Not fully implemented per-cut yet, but we can compute a global or per-segment average if we stored it)
if overlay_segments:
current_frame = 0
# Defaults (Normalized Centers)
last_center_v1 = (0.5, 0.5)
last_center_v2 = (0.5, 0.5)
# We also want to track optimal scale for the segment
last_opt_scale = None
sorted_segs = sorted(overlay_segments, key=lambda x: x['start'])
is_last_dual = False # Initialize is_last_dual
for idx, seg in enumerate(sorted_segs):
start_f = int(seg['start'] * fps_float)
end_f = int(seg['end'] * fps_float)
# Fill Gaps
if start_f > current_frame:
cuts_v1.append({"start": current_frame, "end": start_f, "center": last_center_v1, "opt_scale": last_opt_scale})
if is_last_dual:
cuts_v2.append({"start": current_frame, "end": start_f, "center": last_center_v2, "opt_scale": last_opt_scale})
pass
# Analyze Faces
segment_faces = []
frame_count = 0
dual_face_frames = 0
for f_idx in range(start_f, end_f):
if f_idx in faces_per_frame:
fs = faces_per_frame[f_idx]
segment_faces.append(fs)
if len(fs) >= 2:
dual_face_frames += 1
frame_count += 1
is_dual_track = False
if frame_count > 0:
dual_ratio = dual_face_frames / frame_count
if dual_ratio > 0.3:
is_dual_track = True
elif frame_count < 15 and dual_face_frames > 0:
is_dual_track = True
center_v1 = last_center_v1
center_v2 = last_center_v2
# Coordinate lists for mode calculation
cand_v1_x, cand_v1_y = [], []
cand_v2_x, cand_v2_y = [], []
cand_rh = [] # Relative heights
if segment_faces:
for fs in segment_faces:
# Filter Top 2 by Area
top_faces = sorted(fs, key=lambda x: x['area'], reverse=True)[:2]
# Sort by X (Left to Right)
fs_sorted = sorted(top_faces, key=lambda x: x['nx'])
if is_dual_track and len(fs_sorted) >= 2:
# Left -> V2 (Top Track, Upper Screen)
# Right -> V1 (Bottom Track, Lower Screen)
f_left = fs_sorted[0]
f_right = fs_sorted[-1]
cand_rh.append(f_left.get('rh', 0.1))
cand_rh.append(f_right.get('rh', 0.1))
if abs(f_left['nx'] - f_right['nx']) < 0.20:
# Fallback to single
f_main = max(fs, key=lambda x: x['area'])
cand_v1_x.append(f_main['nx'])
cand_v1_y.append(f_main['ny'])
if 'rh' in f_main: cand_rh[-2:] = [f_main['rh']]
else:
# Swap Assignment Here:
# Left Face -> V2 (Top)
cand_v2_x.append(f_left['nx'])
cand_v2_y.append(f_left['ny'])
# Right Face -> V1 (Bottom)
cand_v1_x.append(f_right['nx'])
cand_v1_y.append(f_right['ny'])
elif fs_sorted:
# Single -> V1
f1 = max(fs_sorted, key=lambda x: x['area'])
cand_v1_x.append(f1['nx'])
cand_v1_y.append(f1['ny'])
cand_rh.append(f1.get('rh', 0.1))
# Smart Scale Logic REMOVED per user request
# We will rely on strict "Fill Split Pane Height" logic in make_video_track
opt_scale = None
last_opt_scale = None
# Apply Mode (Robust avg)
def get_mode_avg(vals):
if not vals: return 0.5
try: return statistics.mean(vals)
except: return vals[0]
# If after filtering we have no valid V2 candidates, revert to Single Track
if is_dual_track and not cand_v2_x:
is_dual_track = False
if cand_v1_x:
center_v1 = (get_mode_avg(cand_v1_x), get_mode_avg(cand_v1_y))
if is_dual_track:
if cand_v2_x:
center_v2 = (get_mode_avg(cand_v2_x), get_mode_avg(cand_v2_y))
else:
# This branch should rarely be hit now due to check above
if last_center_v2 != (0.5, 0.5): center_v2 = last_center_v2
else: center_v2 = (center_v1[0] + 0.25, center_v1[1])
# Append Cuts
cuts_v1.append({"start": start_f, "end": end_f, "center": center_v1, "opt_scale": opt_scale})
if is_dual_track:
cuts_v2.append({"start": start_f, "end": end_f, "center": center_v2, "opt_scale": opt_scale})
last_center_v2 = center_v2
is_last_dual = True
else:
is_last_dual = False
last_center_v1 = center_v1
current_frame = end_f
# Final gap
if current_frame < duration_frames:
cuts_v1.append({"start": current_frame, "end": duration_frames, "center": last_center_v1, "opt_scale": last_opt_scale})
else:
cuts_v1.append({"start": 0, "end": duration_frames, "center": (0.5, 0.5), "opt_scale": None})
print(f"Generated {len(cuts_v1)} V1 cuts and {len(cuts_v2)} V2 cuts.")
# --- GENERATE XML TRACKS ---
dual_starts = set(c['start'] for c in cuts_v2)
def make_video_track(cuts_list, track_type="main"):
items = ""
for cut in cuts_list:
seg_start, seg_end = cut['start'], cut['end']
nx, ny = cut['center'] # These are Normalized Source Coords (0..1)
if seg_end - seg_start <= 0: continue
is_dual = (seg_start in dual_starts)
# --- DIMENSION CHECKS ---
src_w = float(source_width)
src_h = float(source_height)
if src_h < 100: src_h = 1080.0 # Safety default
# --- SCALE LOGIC ---
# Fill Sequence Height (Matches User's Request for correct scaling)
# Use the actual Sequence Height passed to create_premiere_xml
target_h = float(height)
# ALWAYS scale to fill the sequence height
final_scale = (target_h / src_h) * 100.0
# Boost scale for split screen to frame faces tighter (User request: "zoom is larger when split")
if track_type == "secondary" or is_dual:
final_scale *= 1.2
if final_scale < 10.0: final_scale = 100.0
s_val = final_scale / 100.0
# --- POSITIONING LOGIC (Shift-Based) ---
# We assume Anchor Point is (0,0) -> CENTER of Clip.
# We want to move the Face (nx, ny) to the Target Screen Position.
# 1. Face Offset from Clip Center (in Source Pixels)
# Center of Source is 0.5, 0.5
off_x_src = (nx - 0.5) * src_w
off_y_src = (ny - 0.5) * src_h
# 2. Face Offset in Screen Pixels (after Scale)
off_x_seq = off_x_src * s_val
off_y_seq = off_y_src * s_val
# 3. Target Screen Position (Pixels)
# Sequence Dimensions: width, height (e.g. 1080, 1920)
target_screen_x = 0.5 * width # Center X
target_screen_y = 0.5 * height # Center Y (Default)
if track_type == "secondary":
target_screen_y = 0.25 * height # Top Quarter
elif track_type == "main" and is_dual:
target_screen_y = 0.75 * height # Bottom Quarter
# 4. Required Clip Center Position
# To place Face at Target, we shift Clip Center by -Offset
req_center_x = target_screen_x - off_x_seq
req_center_y = target_screen_y - off_y_seq
# 5. Normalize for XML (0..1 relative to Sequence)
# XML Coordinate System is Relative to Center (0,0 is Center).
# Absolute 0..1 maps to -0.5..0.5 in XML.
pos_h = (req_center_x / float(width)) - 0.5
pos_v = (req_center_y / float(height)) - 0.5
seg_id = f"clipitem-video-{get_uid()}"
# EXPLICITLY REMOVE Anchor Point (centerOffset) to use Default (Center of Clip).
# We calculate pos_h/pos_v assuming we are placing the Clip Center.
basic_motion = f"""<filter><effect><name>Basic Motion</name><effectid>basic</effectid><effectcategory>motion</effectcategory><effecttype>motion</effecttype><mediatype>video</mediatype><parameter authoringApp="PremierePro"><parameterid>scale</parameterid><name>Scale</name><value>{final_scale:.2f}</value></parameter><parameter authoringApp="PremierePro"><parameterid>center</parameterid><name>Center</name><value><horiz>{pos_h:.5f}</horiz><vert>{pos_v:.5f}</vert></value></parameter></effect></filter>"""
# --- CROP LOGIC (Pane Masking) ---
# We calculate crops based on the Screen Boundaries of the Pane.
# This ensures the split line is perfectly respected.
crop_xml = ""
pane_top_y = 0.0
pane_bottom_y = float(height) # Default Full Screen
should_crop = False
if track_type == "secondary":
# Top Pane (0.0 to 0.5)
pane_bottom_y = height / 2.0
should_crop = True
elif track_type == "main" and is_dual:
# Bottom Pane (0.5 to 1.0)
pane_top_y = height / 2.0
should_crop = True
if should_crop:
# 1. Calculate Clip's Screen Coordinates
# req_center_y is the Screen Y of the Clip Center
clip_screen_h = src_h * s_val
clip_top_screen_y = req_center_y - (clip_screen_h / 2.0)
clip_bottom_screen_y = req_center_y + (clip_screen_h / 2.0)
# 2. Calculate Required Crop in Screen Pixels
# Pixels to remove from Top: Distance from ClipTop to PaneTop
# max(0, PaneTop - ClipTop)
crop_top_px = max(0.0, pane_top_y - clip_top_screen_y)
# Pixels to remove from Bottom: Distance from PaneBottom to ClipBottom
# max(0, ClipBottom - PaneBottom)
crop_bottom_px = max(0.0, clip_bottom_screen_y - pane_bottom_y)
# 3. Convert to Source Percentage
# CropPx / Scale = SourcePx
# SourcePx / SourceHeight * 100 = %
pct_top = (crop_top_px / s_val) / src_h * 100.0
pct_bottom = (crop_bottom_px / s_val) / src_h * 100.0
# Clamp 0-100
pct_top = max(0.0, min(100.0, pct_top))
pct_bottom = max(0.0, min(100.0, pct_bottom))
crop_parameters = ""
crop_parameters += f"""<parameter authoringApp="PremierePro"><parameterid>top</parameterid><name>Top</name><value>{pct_top:.2f}</value></parameter>"""
crop_parameters += f"""<parameter authoringApp="PremierePro"><parameterid>bottom</parameterid><name>Bottom</name><value>{pct_bottom:.2f}</value></parameter>"""
crop_xml = f"""<filter><effect><name>Crop</name><effectid>crop</effectid><effectcategory>transform</effectcategory><effecttype>video</effecttype><mediatype>video</mediatype>{crop_parameters}</effect></filter>"""
items += f"""<clipitem id="{seg_id}"><name>{os.path.basename(video_path)}</name><duration>{duration_frames}</duration><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate><start>{seg_start}</start><end>{seg_end}</end><in>{seg_start}</in><out>{seg_end}</out>{get_file_block(video_file_id, video_path)}{basic_motion}{crop_xml}</clipitem>"""
return f"<track>{items}</track>"
track_v1 = make_video_track(cuts_v1, "main")
track_v2 = make_video_track(cuts_v2, "secondary")
# --- OVERLAY TRACK ---
track_overlay_block = ""
if overlay_segments:
overlay_clips = ""
for seg in overlay_segments:
# ... (overlay logic same as before)
# Re-implement simple loop here to ensure variable scope
start_f = int(seg['start'] * fps_float)
end_f = int(seg['end'] * fps_float)
clip_dur = end_f - start_f
if clip_dur <= 0: continue
ov_fid = f"file-ov-{seg['index']}-{get_uid()}"
ov_cid = f"clip-ov-{seg['index']}-{get_uid()}"
file_blk = f"""<file id="{ov_fid}"><name>{os.path.basename(seg['path'])}</name><pathurl>{seg['path']}</pathurl><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate><duration>{clip_dur}</duration><media><video><samplecharacteristics><width>{width}</width><height>{height}</height><alpha>straight</alpha></samplecharacteristics></video></media></file>"""
overlay_clips += f"""<clipitem id="{ov_cid}"><name>{os.path.basename(seg['path'])}</name><duration>{clip_dur}</duration><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate><start>{start_f}</start><end>{end_f}</end><in>0</in><out>{clip_dur}</out>{file_blk}<compositemode>normal</compositemode></clipitem>"""
track_overlay_block = f"<track>{overlay_clips}</track>"
else:
track_overlay_block = "<track></track>"
# --- ASSEMBLE ---
timecode_block = f"""<timecode><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate><string>00:00:00:00</string><frame>0</frame><displayformat>NDF</displayformat></timecode>"""
audio_blk = f"""<track><clipitem id="{audio_file_id}"><name>{os.path.basename(video_path)}</name><duration>{duration_frames}</duration><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate><start>0</start><end>{duration_frames}</end>{get_file_block(video_file_id, video_path)}<sourcetrack><mediatype>audio</mediatype><trackindex>1</trackindex></sourcetrack></clipitem></track>"""
return f"""<?xml version="1.0" encoding="UTF-8"?><xmeml version="4"><sequence id="{sequence_uuid}"><name>{project_name}_CutRef</name><duration>{duration_frames}</duration><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate>{timecode_block}<media><video><format><samplecharacteristics><rate><timebase>{timebase}</timebase><ntsc>FALSE</ntsc></rate><width>{width}</width><height>{height}</height><pixelaspectratio>square</pixelaspectratio></samplecharacteristics></format>{track_v1}{track_v2}{track_overlay_block}</video><audio>{audio_blk}</audio></media></sequence></xmeml>"""
|