@lumy-pack/scene-sieve 0.0.8 → 0.0.9
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +93 -0
- package/dist/cli.mjs +411 -7
- package/dist/commands/Sieve.d.ts +2 -0
- package/dist/constants.d.ts +2 -0
- package/dist/core/extractor.d.ts +23 -0
- package/dist/core/index.d.ts +2 -1
- package/dist/core/segmenter.d.ts +38 -0
- package/dist/core/workspace.d.ts +1 -0
- package/dist/index.cjs +382 -4
- package/dist/index.mjs +382 -4
- package/dist/pipeline-worker.mjs +382 -4
- package/dist/types/index.d.ts +22 -0
- package/dist/utils/concurrency.d.ts +5 -0
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -99,6 +99,99 @@ scene-sieve <input> [options]
|
|
|
99
99
|
| Video | `.mp4`, `.mov`, `.avi`, `.mkv`, `.webm` |
|
|
100
100
|
| Animation | `.gif` |
|
|
101
101
|
|
|
102
|
+
### Parameter Tuning Guide
|
|
103
|
+
|
|
104
|
+
Not sure where to start? Here's how each parameter affects the output, based on real benchmarks with a ~19s screen recording (MOV) and a GIF animation.
|
|
105
|
+
|
|
106
|
+
#### `--count` — How many frames to keep
|
|
107
|
+
|
|
108
|
+
| Setting | Extracted | Selected | Notes |
|
|
109
|
+
|---------|-----------|----------|-------|
|
|
110
|
+
| `-n 3` | 90 | 3 | First and last frames are always preserved (boundary protection) |
|
|
111
|
+
| `-n 10` | 90 | 10 | Good for short summaries |
|
|
112
|
+
| `-n 20` (default) | 90 | 20 | Balanced for most videos |
|
|
113
|
+
| `-n 50` | 90 | 22 | Only 22 frames passed the score threshold — count above actual scenes has no effect |
|
|
114
|
+
|
|
115
|
+
#### `--threshold` — Minimum score to keep a frame
|
|
116
|
+
|
|
117
|
+
Higher values = stricter filtering = fewer frames.
|
|
118
|
+
|
|
119
|
+
| Setting | Selected | Notes |
|
|
120
|
+
|---------|----------|-------|
|
|
121
|
+
| `-t 0.1` | 20 | Very permissive — most scene changes pass |
|
|
122
|
+
| `-t 0.3` | 20 | Still permissive for screen recordings |
|
|
123
|
+
| `-t 0.5` (default) | 20 | Capped by the default count of 20 |
|
|
124
|
+
| `-t 0.7` | 19 | Starts filtering subtle changes |
|
|
125
|
+
| `-t 0.9` | 12 | Only major scene transitions survive |
|
|
126
|
+
|
|
127
|
+
> **Tip**: Use `-t` alone for "give me everything important". Combine with `-n` to set an upper bound (e.g., `-t 0.3 -n 10`).
|
|
128
|
+
|
|
129
|
+
#### `--fps` and `--max-frames` — Extraction density
|
|
130
|
+
|
|
131
|
+
These control how many frames are pulled from the video before analysis. More frames = more precision but longer processing.
|
|
132
|
+
|
|
133
|
+
| Setting | Extracted | Selected | Time |
|
|
134
|
+
|---------|-----------|----------|------|
|
|
135
|
+
| `--fps 1` | 18 | 6 | ~5s |
|
|
136
|
+
| `--fps 5` (default) | 90 | 20 | ~25s |
|
|
137
|
+
| `--fps 10` | 180 | 20 | ~47s |
|
|
138
|
+
| `-mf 50` | 47 | 13 | ~13s |
|
|
139
|
+
|
|
140
|
+
> **Tip**: For quick previews, `--fps 1` is 5x faster. For frame-accurate analysis, `--fps 10` captures finer transitions.
|
|
141
|
+
|
|
142
|
+
#### `--scale` — Analysis resolution
|
|
143
|
+
|
|
144
|
+
Controls the resolution used for vision analysis (not output resolution). Lower = faster but less sensitive.
|
|
145
|
+
|
|
146
|
+
| Setting | Selected | Time | Output Size |
|
|
147
|
+
|---------|----------|------|-------------|
|
|
148
|
+
| `-s 360` | 7 | ~6s | 72 KB |
|
|
149
|
+
| `-s 720` (default) | 20 | ~25s | 634 KB |
|
|
150
|
+
| `-s 1080` | 20 | ~54s | 1,172 KB |
|
|
151
|
+
|
|
152
|
+
> **Tip**: `360` is good for quick scans. `720` provides the best speed/quality balance. `1080` is only needed when detecting very subtle UI changes.
|
|
153
|
+
|
|
154
|
+
#### `--iou-threshold` and `--anim-threshold` — Animation sensitivity
|
|
155
|
+
|
|
156
|
+
These control how aggressively repeating animations (spinners, blinking cursors) are detected and suppressed.
|
|
157
|
+
|
|
158
|
+
| Setting | Animations Detected | Notes |
|
|
159
|
+
|---------|-------------------|-------|
|
|
160
|
+
| `-it 0.5 -at 3` | 7 (MOV), 4 (GIF) | Sensitive — catches most repeating motion |
|
|
161
|
+
| `-it 0.9 -at 5` (default) | 0 | Conservative — only obvious loops |
|
|
162
|
+
| `-it 0.95 -at 10` | 0 | Very conservative |
|
|
163
|
+
|
|
164
|
+
> **Tip**: If your video has loading spinners or repeated UI animations, try `-it 0.5 -at 3` to suppress them.
|
|
165
|
+
|
|
166
|
+
#### `--quality` — Output JPEG quality
|
|
167
|
+
|
|
168
|
+
Only affects file size, **not** scene detection. The same frames are selected regardless of quality.
|
|
169
|
+
|
|
170
|
+
| Setting | File Size (5 frames) |
|
|
171
|
+
|---------|---------------------|
|
|
172
|
+
| `-q 30` | 62 KB |
|
|
173
|
+
| `-q 80` (default) | 151 KB |
|
|
174
|
+
| `-q 100` | 407 KB |
|
|
175
|
+
|
|
176
|
+
### Recommended Presets
|
|
177
|
+
|
|
178
|
+
```bash
|
|
179
|
+
# Quick preview — fast, rough selection
|
|
180
|
+
scene-sieve input.mp4 --fps 1 -s 360 -n 10
|
|
181
|
+
|
|
182
|
+
# Balanced (default) — good for most use cases
|
|
183
|
+
scene-sieve input.mp4
|
|
184
|
+
|
|
185
|
+
# High precision — catches subtle transitions
|
|
186
|
+
scene-sieve input.mp4 --fps 10 -s 1080 -t 0.3
|
|
187
|
+
|
|
188
|
+
# UI recording — suppress animations, keep key states
|
|
189
|
+
scene-sieve recording.mov -it 0.5 -at 3 -t 0.3 -n 15
|
|
190
|
+
|
|
191
|
+
# Minimal summary — just the major scenes
|
|
192
|
+
scene-sieve input.mp4 -t 0.9 -n 5
|
|
193
|
+
```
|
|
194
|
+
|
|
102
195
|
### Examples
|
|
103
196
|
|
|
104
197
|
```bash
|
package/dist/cli.mjs
CHANGED
|
@@ -51,7 +51,7 @@ import { join } from "path";
|
|
|
51
51
|
function getTempWorkspaceDir(sessionId) {
|
|
52
52
|
return join(TEMP_BASE_DIR, `${WORKSPACE_PREFIX}${sessionId}`);
|
|
53
53
|
}
|
|
54
|
-
var APP_NAME, DEFAULT_COUNT, DEFAULT_THRESHOLD, DEFAULT_FPS, DEFAULT_SCALE, DEFAULT_QUALITY, DEFAULT_MAX_FRAMES, NORMALIZATION_LOGISTIC_K, NORMALIZATION_ALPHA, NORMALIZATION_MAD_COEFFICIENT, NORMALIZATION_MIN_SAMPLE_SIZE, WORKSPACE_PREFIX, TEMP_BASE_DIR, FRAME_OUTPUT_EXTENSION, FRAME_FILENAME_PATTERN, OPENCV_BATCH_SIZE, DBSCAN_ALPHA, DBSCAN_MIN_PTS, IOU_THRESHOLD, DECAY_LAMBDA, ANIMATION_FRAME_THRESHOLD, MATCH_DISTANCE_THRESHOLD, PIXELDIFF_GAUSSIAN_KERNEL, PIXELDIFF_BINARY_THRESHOLD, PIXELDIFF_CONTOUR_MIN_AREA, PIXELDIFF_SAMPLE_SPACING;
|
|
54
|
+
var APP_NAME, DEFAULT_COUNT, DEFAULT_THRESHOLD, DEFAULT_FPS, DEFAULT_SCALE, DEFAULT_QUALITY, DEFAULT_MAX_FRAMES, NORMALIZATION_LOGISTIC_K, NORMALIZATION_ALPHA, NORMALIZATION_MAD_COEFFICIENT, NORMALIZATION_MIN_SAMPLE_SIZE, WORKSPACE_PREFIX, TEMP_BASE_DIR, FRAME_OUTPUT_EXTENSION, FRAME_FILENAME_PATTERN, OPENCV_BATCH_SIZE, DBSCAN_ALPHA, DBSCAN_MIN_PTS, IOU_THRESHOLD, DECAY_LAMBDA, ANIMATION_FRAME_THRESHOLD, MATCH_DISTANCE_THRESHOLD, PIXELDIFF_GAUSSIAN_KERNEL, PIXELDIFF_BINARY_THRESHOLD, PIXELDIFF_CONTOUR_MIN_AREA, PIXELDIFF_SAMPLE_SPACING, DEFAULT_MAX_SEGMENT_DURATION, DEFAULT_SEGMENT_CONCURRENCY;
|
|
55
55
|
var init_constants = __esm({
|
|
56
56
|
"src/constants.ts"() {
|
|
57
57
|
"use strict";
|
|
@@ -81,6 +81,8 @@ var init_constants = __esm({
|
|
|
81
81
|
PIXELDIFF_BINARY_THRESHOLD = 30;
|
|
82
82
|
PIXELDIFF_CONTOUR_MIN_AREA = 100;
|
|
83
83
|
PIXELDIFF_SAMPLE_SPACING = 8;
|
|
84
|
+
DEFAULT_MAX_SEGMENT_DURATION = 300;
|
|
85
|
+
DEFAULT_SEGMENT_CONCURRENCY = 2;
|
|
84
86
|
}
|
|
85
87
|
});
|
|
86
88
|
|
|
@@ -693,6 +695,23 @@ async function buildFrameList(framesDir, duration) {
|
|
|
693
695
|
extractPath: join2(framesDir, file)
|
|
694
696
|
}));
|
|
695
697
|
}
|
|
698
|
+
async function extractFramesForRange(inputPath, outputDir, fps, scale, startTime, duration) {
|
|
699
|
+
const outputPattern = join2(outputDir, FRAME_FILENAME_PATTERN);
|
|
700
|
+
await execa(ffmpegPath, [
|
|
701
|
+
"-ss",
|
|
702
|
+
String(startTime),
|
|
703
|
+
"-i",
|
|
704
|
+
inputPath,
|
|
705
|
+
"-t",
|
|
706
|
+
String(duration),
|
|
707
|
+
"-vf",
|
|
708
|
+
`fps=${fps},scale=-1:${scale}`,
|
|
709
|
+
"-q:v",
|
|
710
|
+
"2",
|
|
711
|
+
outputPattern
|
|
712
|
+
]);
|
|
713
|
+
return buildFrameList(outputDir, duration);
|
|
714
|
+
}
|
|
696
715
|
var init_extractor = __esm({
|
|
697
716
|
"src/core/extractor.ts"() {
|
|
698
717
|
"use strict";
|
|
@@ -760,6 +779,15 @@ async function finalizeOutput(ctx, selectedFrames) {
|
|
|
760
779
|
await rename(stagingDir, outputPath);
|
|
761
780
|
return outputFiles;
|
|
762
781
|
}
|
|
782
|
+
async function createSegmentWorkspace(parentWorkspacePath, segmentIndex) {
|
|
783
|
+
const segmentPath = join3(
|
|
784
|
+
parentWorkspacePath,
|
|
785
|
+
"segments",
|
|
786
|
+
String(segmentIndex)
|
|
787
|
+
);
|
|
788
|
+
await ensureDir(join3(segmentPath, "frames"));
|
|
789
|
+
return segmentPath;
|
|
790
|
+
}
|
|
763
791
|
async function cleanupWorkspace(workspacePath) {
|
|
764
792
|
if (!workspacePath) return;
|
|
765
793
|
try {
|
|
@@ -844,7 +872,9 @@ function resolveOptions(options) {
|
|
|
844
872
|
quality: options.quality ?? DEFAULT_QUALITY,
|
|
845
873
|
iouThreshold: options.iouThreshold ?? IOU_THRESHOLD,
|
|
846
874
|
animationThreshold: options.animationThreshold ?? ANIMATION_FRAME_THRESHOLD,
|
|
847
|
-
debug: options.debug ?? false
|
|
875
|
+
debug: options.debug ?? false,
|
|
876
|
+
maxSegmentDuration: options.maxSegmentDuration ?? DEFAULT_MAX_SEGMENT_DURATION,
|
|
877
|
+
concurrency: options.concurrency ?? DEFAULT_SEGMENT_CONCURRENCY
|
|
848
878
|
};
|
|
849
879
|
}
|
|
850
880
|
async function resolveInput(options, workspacePath) {
|
|
@@ -1123,18 +1153,381 @@ var init_pruner = __esm({
|
|
|
1123
1153
|
}
|
|
1124
1154
|
});
|
|
1125
1155
|
|
|
1156
|
+
// src/utils/concurrency.ts
|
|
1157
|
+
function concurrencyLimit(limit) {
|
|
1158
|
+
limit = Math.max(1, limit);
|
|
1159
|
+
let active = 0;
|
|
1160
|
+
const queue = [];
|
|
1161
|
+
return async (fn) => {
|
|
1162
|
+
while (active >= limit) {
|
|
1163
|
+
await new Promise((resolve2) => queue.push(resolve2));
|
|
1164
|
+
}
|
|
1165
|
+
active++;
|
|
1166
|
+
try {
|
|
1167
|
+
return await fn();
|
|
1168
|
+
} finally {
|
|
1169
|
+
active--;
|
|
1170
|
+
queue.shift()?.();
|
|
1171
|
+
}
|
|
1172
|
+
};
|
|
1173
|
+
}
|
|
1174
|
+
var init_concurrency = __esm({
|
|
1175
|
+
"src/utils/concurrency.ts"() {
|
|
1176
|
+
"use strict";
|
|
1177
|
+
}
|
|
1178
|
+
});
|
|
1179
|
+
|
|
1180
|
+
// src/core/segmenter.ts
|
|
1181
|
+
import { randomUUID } from "crypto";
|
|
1182
|
+
import { join as join5 } from "path";
|
|
1183
|
+
function shouldSegment(resolvedOptions, originalOptions) {
|
|
1184
|
+
if (resolvedOptions.mode === "frames") return false;
|
|
1185
|
+
if (originalOptions.mode === "file") {
|
|
1186
|
+
if (originalOptions.inputPath.toLowerCase().endsWith(".gif")) return false;
|
|
1187
|
+
}
|
|
1188
|
+
return true;
|
|
1189
|
+
}
|
|
1190
|
+
function computeSegmentPlan(totalDuration, maxSegmentDuration, maxFrames, fps) {
|
|
1191
|
+
const effectiveFps = Math.max(0.5, Math.min(fps, maxFrames / totalDuration));
|
|
1192
|
+
if (totalDuration <= maxSegmentDuration) {
|
|
1193
|
+
return [
|
|
1194
|
+
{
|
|
1195
|
+
index: 0,
|
|
1196
|
+
startTime: 0,
|
|
1197
|
+
endTime: totalDuration,
|
|
1198
|
+
duration: totalDuration,
|
|
1199
|
+
allocatedFrames: Math.min(
|
|
1200
|
+
Math.ceil(effectiveFps * totalDuration),
|
|
1201
|
+
maxFrames
|
|
1202
|
+
),
|
|
1203
|
+
effectiveFps,
|
|
1204
|
+
overlapBefore: 0,
|
|
1205
|
+
overlapAfter: 0,
|
|
1206
|
+
extractStartTime: 0,
|
|
1207
|
+
extractDuration: totalDuration
|
|
1208
|
+
}
|
|
1209
|
+
];
|
|
1210
|
+
}
|
|
1211
|
+
const segmentCount = Math.ceil(totalDuration / maxSegmentDuration);
|
|
1212
|
+
const overlapTime = 1 / effectiveFps;
|
|
1213
|
+
const segments = [];
|
|
1214
|
+
for (let i = 0; i < segmentCount; i++) {
|
|
1215
|
+
const startTime = i * maxSegmentDuration;
|
|
1216
|
+
const endTime = Math.min((i + 1) * maxSegmentDuration, totalDuration);
|
|
1217
|
+
const duration = endTime - startTime;
|
|
1218
|
+
const overlapBefore = i > 0 ? 1 : 0;
|
|
1219
|
+
const overlapAfter = i < segmentCount - 1 ? 1 : 0;
|
|
1220
|
+
const extractStartTime = Math.max(
|
|
1221
|
+
0,
|
|
1222
|
+
startTime - overlapBefore * overlapTime
|
|
1223
|
+
);
|
|
1224
|
+
const extractEndTime = Math.min(
|
|
1225
|
+
totalDuration,
|
|
1226
|
+
endTime + overlapAfter * overlapTime
|
|
1227
|
+
);
|
|
1228
|
+
const extractDuration = extractEndTime - extractStartTime;
|
|
1229
|
+
segments.push({
|
|
1230
|
+
index: i,
|
|
1231
|
+
startTime,
|
|
1232
|
+
endTime,
|
|
1233
|
+
duration,
|
|
1234
|
+
allocatedFrames: Math.ceil(effectiveFps * duration),
|
|
1235
|
+
effectiveFps,
|
|
1236
|
+
overlapBefore,
|
|
1237
|
+
overlapAfter,
|
|
1238
|
+
extractStartTime,
|
|
1239
|
+
extractDuration
|
|
1240
|
+
});
|
|
1241
|
+
}
|
|
1242
|
+
const totalAllocated = segments.reduce(
|
|
1243
|
+
(sum, s) => sum + s.allocatedFrames,
|
|
1244
|
+
0
|
|
1245
|
+
);
|
|
1246
|
+
if (totalAllocated > maxFrames) {
|
|
1247
|
+
segments[segments.length - 1].allocatedFrames -= totalAllocated - maxFrames;
|
|
1248
|
+
}
|
|
1249
|
+
return segments;
|
|
1250
|
+
}
|
|
1251
|
+
function mergeSegmentFrames(segmentResults) {
|
|
1252
|
+
if (segmentResults.length === 0) {
|
|
1253
|
+
return { frames: [], edges: [], animations: [] };
|
|
1254
|
+
}
|
|
1255
|
+
const allFrames = [];
|
|
1256
|
+
for (const result of segmentResults) {
|
|
1257
|
+
for (const frame of result.frames) {
|
|
1258
|
+
allFrames.push({
|
|
1259
|
+
frame: {
|
|
1260
|
+
...frame,
|
|
1261
|
+
// Use extractStartTime for timestamp correction (Section 18 note 1)
|
|
1262
|
+
timestamp: frame.timestamp + result.segment.extractStartTime
|
|
1263
|
+
},
|
|
1264
|
+
segmentIndex: result.segment.index,
|
|
1265
|
+
localId: frame.id
|
|
1266
|
+
});
|
|
1267
|
+
}
|
|
1268
|
+
}
|
|
1269
|
+
allFrames.sort((a, b) => a.frame.timestamp - b.frame.timestamp);
|
|
1270
|
+
const effectiveFps = segmentResults[0].segment.effectiveFps;
|
|
1271
|
+
const dupThreshold = 1 / (effectiveFps * 2);
|
|
1272
|
+
const uniqueFrames = [];
|
|
1273
|
+
for (const entry of allFrames) {
|
|
1274
|
+
if (uniqueFrames.length > 0) {
|
|
1275
|
+
const last = uniqueFrames[uniqueFrames.length - 1];
|
|
1276
|
+
if (Math.abs(entry.frame.timestamp - last.frame.timestamp) < dupThreshold) {
|
|
1277
|
+
continue;
|
|
1278
|
+
}
|
|
1279
|
+
}
|
|
1280
|
+
uniqueFrames.push(entry);
|
|
1281
|
+
}
|
|
1282
|
+
const globalIdMap = /* @__PURE__ */ new Map();
|
|
1283
|
+
const frames = uniqueFrames.map((entry, globalId) => {
|
|
1284
|
+
globalIdMap.set(`${entry.segmentIndex}:${entry.localId}`, globalId);
|
|
1285
|
+
return {
|
|
1286
|
+
id: globalId,
|
|
1287
|
+
timestamp: entry.frame.timestamp,
|
|
1288
|
+
extractPath: entry.frame.extractPath
|
|
1289
|
+
};
|
|
1290
|
+
});
|
|
1291
|
+
const edges = [];
|
|
1292
|
+
const edgeMap = /* @__PURE__ */ new Map();
|
|
1293
|
+
for (const result of segmentResults) {
|
|
1294
|
+
for (const edge of result.edges) {
|
|
1295
|
+
const newSourceId = globalIdMap.get(
|
|
1296
|
+
`${result.segment.index}:${edge.sourceId}`
|
|
1297
|
+
);
|
|
1298
|
+
const newTargetId = globalIdMap.get(
|
|
1299
|
+
`${result.segment.index}:${edge.targetId}`
|
|
1300
|
+
);
|
|
1301
|
+
if (newSourceId === void 0 || newTargetId === void 0) continue;
|
|
1302
|
+
const edgeKey = `${newSourceId}-${newTargetId}`;
|
|
1303
|
+
const existingIdx = edgeMap.get(edgeKey);
|
|
1304
|
+
if (existingIdx !== void 0) {
|
|
1305
|
+
if (edges[existingIdx].score < edge.score) {
|
|
1306
|
+
edges[existingIdx] = {
|
|
1307
|
+
sourceId: newSourceId,
|
|
1308
|
+
targetId: newTargetId,
|
|
1309
|
+
score: edge.score
|
|
1310
|
+
};
|
|
1311
|
+
}
|
|
1312
|
+
} else {
|
|
1313
|
+
edgeMap.set(edgeKey, edges.length);
|
|
1314
|
+
edges.push({
|
|
1315
|
+
sourceId: newSourceId,
|
|
1316
|
+
targetId: newTargetId,
|
|
1317
|
+
score: edge.score
|
|
1318
|
+
});
|
|
1319
|
+
}
|
|
1320
|
+
}
|
|
1321
|
+
}
|
|
1322
|
+
const animations = [];
|
|
1323
|
+
for (const result of segmentResults) {
|
|
1324
|
+
for (const anim of result.animations) {
|
|
1325
|
+
const newStartId = globalIdMap.get(
|
|
1326
|
+
`${result.segment.index}:${anim.startFrameId}`
|
|
1327
|
+
);
|
|
1328
|
+
const newEndId = globalIdMap.get(
|
|
1329
|
+
`${result.segment.index}:${anim.endFrameId}`
|
|
1330
|
+
);
|
|
1331
|
+
if (newStartId === void 0 || newEndId === void 0) continue;
|
|
1332
|
+
animations.push({
|
|
1333
|
+
...anim,
|
|
1334
|
+
startFrameId: newStartId,
|
|
1335
|
+
endFrameId: newEndId
|
|
1336
|
+
});
|
|
1337
|
+
}
|
|
1338
|
+
}
|
|
1339
|
+
return { frames, edges, animations };
|
|
1340
|
+
}
|
|
1341
|
+
function buildSegmentContext(segment, frames, segmentWorkspacePath, resolvedOptions, onProgress) {
|
|
1342
|
+
return {
|
|
1343
|
+
options: {
|
|
1344
|
+
...resolvedOptions,
|
|
1345
|
+
fps: segment.effectiveFps,
|
|
1346
|
+
maxFrames: segment.allocatedFrames
|
|
1347
|
+
},
|
|
1348
|
+
workspacePath: segmentWorkspacePath,
|
|
1349
|
+
frames,
|
|
1350
|
+
graph: [],
|
|
1351
|
+
status: "ANALYZING",
|
|
1352
|
+
emitProgress: onProgress
|
|
1353
|
+
};
|
|
1354
|
+
}
|
|
1355
|
+
async function processSegment(inputPath, segment, workspacePath, resolvedOptions, onProgress) {
|
|
1356
|
+
const framesDir = join5(workspacePath, "frames");
|
|
1357
|
+
const frames = await extractFramesForRange(
|
|
1358
|
+
inputPath,
|
|
1359
|
+
framesDir,
|
|
1360
|
+
segment.effectiveFps,
|
|
1361
|
+
resolvedOptions.scale,
|
|
1362
|
+
segment.extractStartTime,
|
|
1363
|
+
segment.extractDuration
|
|
1364
|
+
);
|
|
1365
|
+
if (frames.length < 2) {
|
|
1366
|
+
return { segment, frames, edges: [], animations: [] };
|
|
1367
|
+
}
|
|
1368
|
+
const ctx = buildSegmentContext(
|
|
1369
|
+
segment,
|
|
1370
|
+
frames,
|
|
1371
|
+
workspacePath,
|
|
1372
|
+
resolvedOptions,
|
|
1373
|
+
onProgress
|
|
1374
|
+
);
|
|
1375
|
+
const { edges, animations } = await analyzeFrames(ctx);
|
|
1376
|
+
return { segment, frames, edges, animations };
|
|
1377
|
+
}
|
|
1378
|
+
async function runSegmentedPipeline(options, resolvedOptions) {
|
|
1379
|
+
const pipelineStart = Date.now();
|
|
1380
|
+
const sessionId = randomUUID();
|
|
1381
|
+
let mainWorkspace = "";
|
|
1382
|
+
try {
|
|
1383
|
+
mainWorkspace = await createWorkspace(sessionId);
|
|
1384
|
+
logger.debug(`Segmented pipeline: workspace at ${mainWorkspace}`);
|
|
1385
|
+
const { resolvedInputPath } = await resolveInput(options, mainWorkspace);
|
|
1386
|
+
const inputPath = resolvedInputPath ?? resolvedOptions.inputPath;
|
|
1387
|
+
if (!inputPath) {
|
|
1388
|
+
throw new Error("No input path available for segmented pipeline");
|
|
1389
|
+
}
|
|
1390
|
+
const metadata = await getVideoMetadata(inputPath);
|
|
1391
|
+
const totalDuration = parseFloat(metadata.format?.duration ?? "0");
|
|
1392
|
+
if (totalDuration <= 0) {
|
|
1393
|
+
throw new Error(`Invalid video duration: ${totalDuration}`);
|
|
1394
|
+
}
|
|
1395
|
+
logger.debug(`Video duration: ${totalDuration}s`);
|
|
1396
|
+
const segments = computeSegmentPlan(
|
|
1397
|
+
totalDuration,
|
|
1398
|
+
resolvedOptions.maxSegmentDuration,
|
|
1399
|
+
resolvedOptions.maxFrames,
|
|
1400
|
+
resolvedOptions.fps
|
|
1401
|
+
);
|
|
1402
|
+
logger.debug(`Segment plan: ${segments.length} segments`);
|
|
1403
|
+
const limit = concurrencyLimit(resolvedOptions.concurrency);
|
|
1404
|
+
const segmentProgresses = new Array(segments.length).fill(0);
|
|
1405
|
+
const weights = segments.map((s) => s.duration / totalDuration);
|
|
1406
|
+
const emitOverallProgress = (phase) => {
|
|
1407
|
+
if (!options.onProgress) return;
|
|
1408
|
+
const overall = weights.reduce(
|
|
1409
|
+
(sum, w, i) => sum + w * (segmentProgresses[i] ?? 0),
|
|
1410
|
+
0
|
|
1411
|
+
);
|
|
1412
|
+
options.onProgress(phase, Math.min(100, overall));
|
|
1413
|
+
};
|
|
1414
|
+
options.onProgress?.("EXTRACTING", 0);
|
|
1415
|
+
const results = await Promise.all(
|
|
1416
|
+
segments.map(
|
|
1417
|
+
(segment) => limit(async () => {
|
|
1418
|
+
const segWorkspace = await createSegmentWorkspace(
|
|
1419
|
+
mainWorkspace,
|
|
1420
|
+
segment.index
|
|
1421
|
+
);
|
|
1422
|
+
const result = await processSegment(
|
|
1423
|
+
inputPath,
|
|
1424
|
+
segment,
|
|
1425
|
+
segWorkspace,
|
|
1426
|
+
resolvedOptions,
|
|
1427
|
+
(percent) => {
|
|
1428
|
+
segmentProgresses[segment.index] = percent;
|
|
1429
|
+
emitOverallProgress("ANALYZING");
|
|
1430
|
+
}
|
|
1431
|
+
);
|
|
1432
|
+
return result;
|
|
1433
|
+
})
|
|
1434
|
+
)
|
|
1435
|
+
);
|
|
1436
|
+
options.onProgress?.("ANALYZING", 100);
|
|
1437
|
+
const { frames, edges, animations } = mergeSegmentFrames(results);
|
|
1438
|
+
logger.debug(
|
|
1439
|
+
`Merged: ${frames.length} frames, ${edges.length} edges, ${animations.length} animations`
|
|
1440
|
+
);
|
|
1441
|
+
options.onProgress?.("PRUNING", 0);
|
|
1442
|
+
const survivingIds = pruneByThresholdWithCap(
|
|
1443
|
+
edges,
|
|
1444
|
+
frames,
|
|
1445
|
+
resolvedOptions.threshold,
|
|
1446
|
+
resolvedOptions.count
|
|
1447
|
+
);
|
|
1448
|
+
const prunedFrames = frames.filter((f) => survivingIds.has(f.id));
|
|
1449
|
+
options.onProgress?.("PRUNING", 100);
|
|
1450
|
+
options.onProgress?.("FINALIZING", 0);
|
|
1451
|
+
const ctx = {
|
|
1452
|
+
options: resolvedOptions,
|
|
1453
|
+
workspacePath: mainWorkspace,
|
|
1454
|
+
frames,
|
|
1455
|
+
graph: edges,
|
|
1456
|
+
animations,
|
|
1457
|
+
status: "FINALIZING",
|
|
1458
|
+
emitProgress: (percent) => options.onProgress?.("FINALIZING", percent)
|
|
1459
|
+
};
|
|
1460
|
+
let outputFiles = [];
|
|
1461
|
+
let outputBuffers;
|
|
1462
|
+
if (resolvedOptions.mode === "buffer" || resolvedOptions.mode === "frames") {
|
|
1463
|
+
outputBuffers = await readFramesAsBuffers(
|
|
1464
|
+
prunedFrames,
|
|
1465
|
+
resolvedOptions.quality
|
|
1466
|
+
);
|
|
1467
|
+
} else {
|
|
1468
|
+
outputFiles = await finalizeOutput(ctx, prunedFrames);
|
|
1469
|
+
}
|
|
1470
|
+
options.onProgress?.("FINALIZING", 100);
|
|
1471
|
+
logger.success(
|
|
1472
|
+
`Segmented pipeline: ${prunedFrames.length} scenes from ${frames.length} frames (${segments.length} segments)`
|
|
1473
|
+
);
|
|
1474
|
+
return {
|
|
1475
|
+
success: true,
|
|
1476
|
+
originalFramesCount: frames.length,
|
|
1477
|
+
prunedFramesCount: prunedFrames.length,
|
|
1478
|
+
outputFiles,
|
|
1479
|
+
outputBuffers,
|
|
1480
|
+
animations,
|
|
1481
|
+
video: {
|
|
1482
|
+
originalDurationMs: totalDuration * 1e3,
|
|
1483
|
+
fps: resolvedOptions.fps,
|
|
1484
|
+
resolution: {
|
|
1485
|
+
width: resolvedOptions.scale,
|
|
1486
|
+
height: Math.round(resolvedOptions.scale * 9 / 16)
|
|
1487
|
+
}
|
|
1488
|
+
},
|
|
1489
|
+
executionTimeMs: Date.now() - pipelineStart
|
|
1490
|
+
};
|
|
1491
|
+
} catch (error) {
|
|
1492
|
+
const err = error instanceof Error ? error : new Error(String(error));
|
|
1493
|
+
logger.error(`Segmented pipeline failed: ${err.message}`);
|
|
1494
|
+
throw err;
|
|
1495
|
+
} finally {
|
|
1496
|
+
if (!resolvedOptions.debug) {
|
|
1497
|
+
await cleanupWorkspace(mainWorkspace);
|
|
1498
|
+
} else {
|
|
1499
|
+
logger.debug(`Debug mode: workspace preserved at ${mainWorkspace}`);
|
|
1500
|
+
}
|
|
1501
|
+
}
|
|
1502
|
+
}
|
|
1503
|
+
var init_segmenter = __esm({
|
|
1504
|
+
"src/core/segmenter.ts"() {
|
|
1505
|
+
"use strict";
|
|
1506
|
+
init_concurrency();
|
|
1507
|
+
init_logger();
|
|
1508
|
+
init_analyzer();
|
|
1509
|
+
init_extractor();
|
|
1510
|
+
init_input_resolver();
|
|
1511
|
+
init_pruner();
|
|
1512
|
+
init_workspace();
|
|
1513
|
+
}
|
|
1514
|
+
});
|
|
1515
|
+
|
|
1126
1516
|
// src/core/orchestrator.ts
|
|
1127
1517
|
var orchestrator_exports = {};
|
|
1128
1518
|
__export(orchestrator_exports, {
|
|
1129
1519
|
runPipeline: () => runPipeline
|
|
1130
1520
|
});
|
|
1131
|
-
import { randomUUID } from "crypto";
|
|
1521
|
+
import { randomUUID as randomUUID2 } from "crypto";
|
|
1132
1522
|
async function runPipeline(options) {
|
|
1133
|
-
const startTime = Date.now();
|
|
1134
|
-
const sessionId = randomUUID();
|
|
1135
1523
|
const debug = options.debug ?? false;
|
|
1136
1524
|
if (debug) setDebugMode(true);
|
|
1137
1525
|
const resolvedOptions = resolveOptions(options);
|
|
1526
|
+
if (shouldSegment(resolvedOptions, options)) {
|
|
1527
|
+
return runSegmentedPipeline(options, resolvedOptions);
|
|
1528
|
+
}
|
|
1529
|
+
const startTime = Date.now();
|
|
1530
|
+
const sessionId = randomUUID2();
|
|
1138
1531
|
const ctx = {
|
|
1139
1532
|
options: resolvedOptions,
|
|
1140
1533
|
workspacePath: "",
|
|
@@ -1236,6 +1629,7 @@ var init_orchestrator = __esm({
|
|
|
1236
1629
|
init_extractor();
|
|
1237
1630
|
init_input_resolver();
|
|
1238
1631
|
init_pruner();
|
|
1632
|
+
init_segmenter();
|
|
1239
1633
|
init_workspace();
|
|
1240
1634
|
}
|
|
1241
1635
|
});
|
|
@@ -1310,7 +1704,7 @@ var PhaseStep = ({ phase }) => {
|
|
|
1310
1704
|
};
|
|
1311
1705
|
|
|
1312
1706
|
// src/core/run-in-worker.ts
|
|
1313
|
-
import { dirname, join as
|
|
1707
|
+
import { dirname, join as join6 } from "path";
|
|
1314
1708
|
import { fileURLToPath } from "url";
|
|
1315
1709
|
import { Worker } from "worker_threads";
|
|
1316
1710
|
async function runPipelineInWorker(options, onProgress) {
|
|
@@ -1319,7 +1713,7 @@ async function runPipelineInWorker(options, onProgress) {
|
|
|
1319
1713
|
const { runPipeline: runPipeline2 } = await Promise.resolve().then(() => (init_orchestrator(), orchestrator_exports));
|
|
1320
1714
|
return runPipeline2({ ...options, onProgress });
|
|
1321
1715
|
}
|
|
1322
|
-
const workerPath =
|
|
1716
|
+
const workerPath = join6(dirname(currentFile), "pipeline-worker.mjs");
|
|
1323
1717
|
return new Promise((resolve2, reject) => {
|
|
1324
1718
|
const worker = new Worker(workerPath, { workerData: options });
|
|
1325
1719
|
worker.on(
|
|
@@ -1397,6 +1791,8 @@ var SieveView = (props) => {
|
|
|
1397
1791
|
quality: props.quality,
|
|
1398
1792
|
iouThreshold: props.iouThreshold,
|
|
1399
1793
|
animationThreshold: props.animationThreshold,
|
|
1794
|
+
maxSegmentDuration: props.maxSegmentDuration,
|
|
1795
|
+
concurrency: props.concurrency,
|
|
1400
1796
|
debug: props.debug
|
|
1401
1797
|
},
|
|
1402
1798
|
(phase, percent) => {
|
|
@@ -1544,6 +1940,12 @@ program.name("scene-sieve").description("Extract key frames from video and GIF f
|
|
|
1544
1940
|
).option(
|
|
1545
1941
|
"-at, --anim-threshold <number>",
|
|
1546
1942
|
`Min consecutive frames for animation (default: ${ANIMATION_FRAME_THRESHOLD})`
|
|
1943
|
+
).option(
|
|
1944
|
+
"--max-segment-duration <number>",
|
|
1945
|
+
`Max segment duration in seconds for long video splitting (default: ${DEFAULT_MAX_SEGMENT_DURATION})`
|
|
1946
|
+
).option(
|
|
1947
|
+
"--concurrency <number>",
|
|
1948
|
+
`Number of segments to process in parallel (default: ${DEFAULT_SEGMENT_CONCURRENCY})`
|
|
1547
1949
|
).option("--debug", "Enable debug mode (preserve temp workspace)").action(async (input, opts) => {
|
|
1548
1950
|
const { waitUntilExit } = render(
|
|
1549
1951
|
React2.createElement(SieveView, {
|
|
@@ -1557,6 +1959,8 @@ program.name("scene-sieve").description("Extract key frames from video and GIF f
|
|
|
1557
1959
|
quality: parseInt(opts.quality, 10),
|
|
1558
1960
|
iouThreshold: opts.iouThreshold !== void 0 ? parseFloat(opts.iouThreshold) : void 0,
|
|
1559
1961
|
animationThreshold: opts.animThreshold !== void 0 ? parseInt(opts.animThreshold, 10) : void 0,
|
|
1962
|
+
maxSegmentDuration: opts.maxSegmentDuration !== void 0 ? parseInt(opts.maxSegmentDuration, 10) : void 0,
|
|
1963
|
+
concurrency: opts.concurrency !== void 0 ? parseInt(opts.concurrency, 10) : void 0,
|
|
1560
1964
|
debug: opts.debug ?? false
|
|
1561
1965
|
})
|
|
1562
1966
|
);
|
package/dist/commands/Sieve.d.ts
CHANGED
package/dist/constants.d.ts
CHANGED
|
@@ -27,4 +27,6 @@ export declare const PIXELDIFF_GAUSSIAN_KERNEL = 3;
|
|
|
27
27
|
export declare const PIXELDIFF_BINARY_THRESHOLD = 30;
|
|
28
28
|
export declare const PIXELDIFF_CONTOUR_MIN_AREA = 100;
|
|
29
29
|
export declare const PIXELDIFF_SAMPLE_SPACING = 8;
|
|
30
|
+
export declare const DEFAULT_MAX_SEGMENT_DURATION = 300;
|
|
31
|
+
export declare const DEFAULT_SEGMENT_CONCURRENCY = 2;
|
|
30
32
|
export declare function getTempWorkspaceDir(sessionId: string): string;
|
package/dist/core/extractor.d.ts
CHANGED
|
@@ -1,7 +1,30 @@
|
|
|
1
1
|
import type { FrameNode, ProcessContext } from '../types/index.js';
|
|
2
|
+
export interface FFprobeMetadata {
|
|
3
|
+
format?: {
|
|
4
|
+
format_name?: string;
|
|
5
|
+
duration?: string;
|
|
6
|
+
};
|
|
7
|
+
streams?: Array<{
|
|
8
|
+
codec_type?: string;
|
|
9
|
+
}>;
|
|
10
|
+
}
|
|
2
11
|
/**
|
|
3
12
|
* Extract frames from video/GIF using FFmpeg.
|
|
4
13
|
* Always uses FPS-based extraction. For long videos, FPS is automatically
|
|
5
14
|
* reduced to stay within maxFrames budget.
|
|
6
15
|
*/
|
|
7
16
|
export declare function extractFrames(ctx: ProcessContext): Promise<FrameNode[]>;
|
|
17
|
+
export declare function getVideoMetadata(inputPath: string): Promise<FFprobeMetadata>;
|
|
18
|
+
/**
|
|
19
|
+
* Extract frames from a specific time range of a video using FFmpeg.
|
|
20
|
+
* Uses input seeking (-ss before -i) for fast seek + -t for duration.
|
|
21
|
+
*
|
|
22
|
+
* @param inputPath - Path to the video file
|
|
23
|
+
* @param outputDir - Directory to write extracted frames
|
|
24
|
+
* @param fps - Frames per second for extraction
|
|
25
|
+
* @param scale - Height scale for vision analysis
|
|
26
|
+
* @param startTime - Start time in seconds
|
|
27
|
+
* @param duration - Duration in seconds to extract
|
|
28
|
+
* @returns Array of FrameNode with segment-local timestamps (starting from 0)
|
|
29
|
+
*/
|
|
30
|
+
export declare function extractFramesForRange(inputPath: string, outputDir: string, fps: number, scale: number, startTime: number, duration: number): Promise<FrameNode[]>;
|
package/dist/core/index.d.ts
CHANGED
|
@@ -5,4 +5,5 @@ export { pruneTo, pruneByThreshold, pruneByThresholdWithCap, suppressConsecutive
|
|
|
5
5
|
export { dbscan } from './dbscan.js';
|
|
6
6
|
export type { Point2D } from './dbscan.js';
|
|
7
7
|
export { resolveInput, resolveOptions } from './input-resolver.js';
|
|
8
|
-
export {
|
|
8
|
+
export { shouldSegment, computeSegmentPlan, processSegment, mergeSegmentFrames, runSegmentedPipeline, } from './segmenter.js';
|
|
9
|
+
export { createWorkspace, createSegmentWorkspace, cleanupWorkspace, finalizeOutput, readFramesAsBuffers, writeInputBuffer, writeInputFrames, } from './workspace.js';
|