{
 "name": "MOZAIC",
 "version": "1.0",
 "schema_version": "1.0",
 "release_date": "TO FILL",
 "cite": "TO FILL",
 "licence": "TO FILL",
 "access": "Private now; gated with manual approval when public.",
 "removal_requests": "TO FILL",
 "labels": {
  "RVRA": "real face, real voice",
  "RVFA": "real face, fake voice",
  "FVRA": "fake face, real voice",
  "FVFA": "fake face, fake voice"
 },
 "arms": {
  "main": "the clip as made",
  "vocoded": "the same clip with its voice passed through BigVGAN; the label does not change"
 },
 "source": {
  "dataset": "VoxCeleb2 (dev set)",
  "people": 1000,
  "cite": "Chung, Nagrani, Zisserman. VoxCeleb2: Deep Speaker Recognition. Interspeech 2018"
 },
 "video": {
  "container": "mov",
  "codec": "h264 (libx264), crf 18, preset medium, yuv420p",
  "size": "224x224",
  "fps": 25,
  "lengths": {
   "77f": "77 frames (3.08 s)",
   "153f": "153 frames (6.12 s)"
  }
 },
 "audio": {
  "codec": "pcm_s16le",
  "sample_rate": 16000,
  "channels": 1
 },
 "applied_to_every_clip": {
  "brightness": {
   "mode": "per-clip gamma on Y, solved from the clip's own histogram",
   "target": 72.0,
   "tolerance": 2.0
  },
  "colour": {
   "mode": "per-clip gain on U/V, solved from the clip's own saturation",
   "target": 17.0,
   "tolerance": 0.75
  },
  "loudness": {
   "mode": "peak-capped constant gain",
   "target_i": -23.0,
   "true_peak_ceiling": -2.0
  },
  "silence_trimmed": false,
  "sound_start": "first 50 ms silent, then a 10 ms fade-in",
  "pixel_aspect_ratio": "1:1"
 },
 "generators": {
  "animate": [
   "animate14b-replace-480p"
  ],
  "facefusion": [
   "facefusion-inswapper128fp16"
  ],
  "s2v": [
   "s2v14b-480p-s40-cfg4.5-retime16to25"
  ],
  "seed_vc": [
   "seedvc-s5-cfg0.3"
  ],
  "bigvgan": "nvidia/bigvgan_v2_22khz_80band_256x",
  "draw_seed": 42
 },
 "counts": {
  "clips": 137745,
  "pairs": {
   "77f": 7590,
   "153f": 7424
  },
  "by_label": {
   "RVRA": 30028,
   "RVFA": 15014,
   "FVRA": 61802,
   "FVFA": 30901
  },
  "by_split": {
   "train": 96957,
   "test": 20235,
   "val": 20553
  },
  "by_folder": {
   "videos/FVFA/animate/77f/main": {
    "train": 3090,
    "test": 639,
    "val": 678
   },
   "videos/FVFA/facefusion/153f/main": {
    "train": 4823,
    "test": 995,
    "val": 1017
   },
   "videos/FVFA/facefusion/77f/main": {
    "train": 4935,
    "test": 1028,
    "val": 1057
   },
   "videos/FVFA/s2v/153f/main": {
    "train": 4451,
    "test": 903,
    "val": 925
   },
   "videos/FVFA/s2v/77f/main": {
    "train": 4479,
    "test": 939,
    "val": 942
   },
   "videos/FVRA/animate/77f/main": {
    "train": 3090,
    "test": 639,
    "val": 678
   },
   "videos/FVRA/animate/77f/vocoded": {
    "train": 3090,
    "test": 639,
    "val": 678
   },
   "videos/FVRA/facefusion/153f/main": {
    "train": 4823,
    "test": 995,
    "val": 1017
   },
   "videos/FVRA/facefusion/153f/vocoded": {
    "train": 4823,
    "test": 995,
    "val": 1017
   },
   "videos/FVRA/facefusion/77f/main": {
    "train": 4935,
    "test": 1028,
    "val": 1057
   },
   "videos/FVRA/facefusion/77f/vocoded": {
    "train": 4935,
    "test": 1028,
    "val": 1057
   },
   "videos/FVRA/s2v/153f/main": {
    "train": 4451,
    "test": 903,
    "val": 925
   },
   "videos/FVRA/s2v/153f/vocoded": {
    "train": 4451,
    "test": 903,
    "val": 925
   },
   "videos/FVRA/s2v/77f/main": {
    "train": 4479,
    "test": 939,
    "val": 942
   },
   "videos/FVRA/s2v/77f/vocoded": {
    "train": 4479,
    "test": 939,
    "val": 942
   },
   "videos/RVFA/seed_vc/153f/main": {
    "train": 5226,
    "test": 1099,
    "val": 1099
   },
   "videos/RVFA/seed_vc/77f/main": {
    "train": 5315,
    "test": 1142,
    "val": 1133
   },
   "videos/RVRA/real/153f/main": {
    "train": 5226,
    "test": 1099,
    "val": 1099
   },
   "videos/RVRA/real/153f/vocoded": {
    "train": 5226,
    "test": 1099,
    "val": 1099
   },
   "videos/RVRA/real/77f/main": {
    "train": 5315,
    "test": 1142,
    "val": 1133
   },
   "videos/RVRA/real/77f/vocoded": {
    "train": 5315,
    "test": 1142,
    "val": 1133
   }
  }
 },
 "people_per_split": {
  "train": 700,
  "test": 150,
  "val": 150
 }
}