added dataset: SH2 James Sunderland (v2)

finetunes/james-sunderland-v2/cond_latents_9a58bc9a.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c8e263823d5948b69e58d812ed88fbe5645c03ee460959d6e68a8f6629bb810
+size 13343

finetunes/james-sunderland-v2/dataset/lxgbsm_00000.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa3a9334e6dfbdf8276a1193fbb712f809a45276c37be5df29f148c2fc324584
+size 1234880

finetunes/james-sunderland-v2/dataset/lxgbsm_00001.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d583a0e40379dc041dbd6fb9fffd83d2ba6610339e4a2a293373973bfd81796c
+size 1587680

finetunes/james-sunderland-v2/dataset/lxgbsm_00002.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66e7ba89e497878b27b6764e15d33dbd4c45896117733df734dd8c63e8f6d4b0
+size 1587680

finetunes/james-sunderland-v2/dataset/lxgbsm_00003.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cbde97c7e6cce1caa5a9b470fa0f00df7d04407240435aa1b18c03d3bdf2510
+size 1323080

finetunes/james-sunderland-v2/dataset/lxgbsm_00004.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d6f665396d46818bbb75c05db5e67e6342a2d586427f2437d39df1c078de1f4
+size 1404224

finetunes/james-sunderland-v2/dataset/lxgbsm_00005.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7175857259a6932e552c011e803d3cf67bd84851e1b6090a032088d49f3a056e
+size 1305440

finetunes/james-sunderland-v2/dataset/lxgbsm_00006.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96eadc75411e796c7a92d1ecc83fd5da86e58ce5bbaca8a9627e9c2646a5c66f
+size 1573568

finetunes/james-sunderland-v2/dataset/lxgbsm_00007.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aecd8dd965f3ccc91b3c0a03548590a93ef94abc99cb568a09a7bc1bfd65ac68
+size 811520

finetunes/james-sunderland-v2/dataset/lxgbsm_00008.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8581aa01bd79ee275374eb99e34e94d3d3ebd4d12bef4dcff3eb10ae61b6c99
+size 1185484

finetunes/james-sunderland-v2/dataset/lxgbsm_00009.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fc917ad96ed1776fc8952735eb41956419b82efd635c0c211a039f0d95e184b
+size 1495956

finetunes/james-sunderland-v2/dataset/lxgbsm_00010.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5148663f1847680f3fdc3afe40b51ce8aa34b44182f440f098ef7e014d168a92
+size 296432

finetunes/james-sunderland-v2/dataset/lxgbsm_00011.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af1449010c075cc581c349de386cfd98ecd421a2763e86acf9db1cc6ee05fdbd
+size 606896

finetunes/james-sunderland-v2/dataset/lxgbsm_00012.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7895ad94887ad6cc759e04ddaac80aab7836104222cd58d7c953cddac5a4cc89
+size 550444

finetunes/james-sunderland-v2/dataset/lxgbsm_00013.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee7c4fad7a9efda7413fd74ff8dbb76db95c3027c8904d7207ee6a03e31468d7
+size 462252

finetunes/james-sunderland-v2/dataset/lxgbsm_00014.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0385c4977ac17cdf665ce4a4ead641a26f4ded2cf265cb78f7c7a103a427bb6d
+size 423440

finetunes/james-sunderland-v2/dataset/lxgbsm_00015.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c5c9def95cf337aec90a47ed0d95f7bde69da928ba1d76e2720b63036a105ed
+size 342296

finetunes/james-sunderland-v2/dataset/lxgbsm_00016.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bda34911ec3fbd4057d433c55bff93a31018062beca44896a373e6e366a43742
+size 331712

finetunes/james-sunderland-v2/dataset/lxgbsm_00017.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d18fddf652f76f351e9308795c755884163ae8d84d15f0d6410362cf7b40df80
+size 317600

finetunes/james-sunderland-v2/dataset/lxgbsm_00018.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f199524f7b217f5e8de979118fd62d04697a4262295039308ea827eaa599b9c5
+size 437552

finetunes/james-sunderland-v2/dataset/lxgbsm_00019.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdfa30c56345d6c809b18fa3c0fadb8431e1bdb373f212be251a47985260c587
+size 331712

finetunes/james-sunderland-v2/dataset/lxgbsm_00020.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e6a63424f17aa653370f053c350fb5c0ea9009c3f2ec2ad160dc024272b645d
+size 398744

finetunes/james-sunderland-v2/dataset/lxgbsm_00021.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d8f5336ce59fcf739122d97e9e6e836d0a2013774465f0e2b5a440dd521390e
+size 314072

finetunes/james-sunderland-v2/dataset/lxgbsm_00022.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae9d5c5a2392f6fae8ba989236a445ab6e188121f3a9028778502d641ce68d5c
+size 268208

finetunes/james-sunderland-v2/dataset/lxgbsm_00023.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d016113f32b555e86be30b1ce5831a8070cdd02490fe7edb3ff0afb61423d07
+size 494000

finetunes/james-sunderland-v2/dataset/lxgbsm_00024.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4de3ee75a047008ef7d3c91546ca55535840571589e19a2e41f72c8e74566615
+size 176480

finetunes/james-sunderland-v2/dataset/lxgbsm_00025.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7e6074038449455d9ecb40e541c155e69bf636bf0c259ce146b4dca928d0bb0
+size 289376

finetunes/james-sunderland-v2/dataset/lxgbsm_00026.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a096af623b5671af14db11286d732f0cb856b5ef792a9694cb9e77f4787b50b1
+size 338768

finetunes/james-sunderland-v2/dataset/lxgbsm_00027.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f03a5055ac9e6e5744afeef481083cdd6d66886c600f6d1ead43e7e050ba46b0
+size 331712

finetunes/james-sunderland-v2/dataset/lxgbsm_00028.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eca4137c035820c531347f8b65e29e4339286df5a0ac891cc3e8c28ce1ee000a
+size 338768

finetunes/james-sunderland-v2/dataset/lxgbsm_00029.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d107c0519eb95e65405a78eb677ba9d7773c73e876566088304f0cdb54e81671
+size 677456

finetunes/james-sunderland-v2/dataset/lxgbsm_00030.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e84632d291e37ab1120ea51480b3be24df85f0c7df393bd83aad2e3b8bd7303a
+size 211760

finetunes/james-sunderland-v2/dataset/train.txt

@@ -0,0 +1,68 @@
+lxgbsm_00000.wav|I'm kind of lost. I'm looking for Silent Hill. Silent Hill? Is this the right way?
+lxgbsm_00001.wav|Well, there's the hotel 2, I guess. There's the hotel 2, I guess. The one on the lake? I wonder if it's still there.
+lxgbsm_00002.wav|You're coming with me? Is it dangerous? I'll be careful. I guess I really don't care if it's dangerous or not.
+lxgbsm_00003.wav|I'm going to town either way. I'm looking for someone. I'm looking for someone.
+lxgbsm_00004.wav|Someone very important to me. I'm looking for Mary. Have you seen her?
+lxgbsm_00005.wav|Without Mary I just can't go on. I'd do anything if I could be with her again.
+lxgbsm_00006.wav|I'm gonna go look for her. I wish I knew. I killed her three years ago. Last year.
+lxgbsm_00007.wav|But I got a letter from her. I got a letter from her. How do you know about
+lxgbsm_00008.wav|Mary? You really seem to care about her. Is that why you're here too? You're not
+lxgbsm_00009.wav|Mary. You're really not Mary. You're alive! I thought that thing killed you. Are you
+lxgbsm_00010.wav|You hurt bad?
+lxgbsm_00011.wav|Your face, your voice, just your hair and...
+lxgbsm_00012.wav|I was confused.
+lxgbsm_00013.wav|There are all sorts of strange things around here.
+lxgbsm_00014.wav|This is no place for a kid.
+lxgbsm_00015.wav|Eddie told me.
+lxgbsm_00016.wav|Who's Rachel?
+lxgbsm_00017.wav|What's that letter?
+lxgbsm_00018.wav|It's time to end this nightmare.
+lxgbsm_00019.wav|What the hell is it?
+lxgbsm_00020.wav|I didn't come here to play, you know.
+lxgbsm_00021.wav|You liar!
+lxgbsm_00022.wav|Don't be ridiculous.
+lxgbsm_00023.wav|How can you sit there and eat pizza?
+lxgbsm_00024.wav|You!
+lxgbsm_00025.wav|It was you, wasn't it?
+lxgbsm_00026.wav|You're the one who stepped on my hand.
+lxgbsm_00027.wav|I'm done with you.
+lxgbsm_00028.wav|Forget you!
+lxgbsm_00029.wav|Someone very important to me.
+lxgbsm_00030.wav|You're not Mary.
+ynoeld_00000.wav|It's you. It's you. But I don't need you anymore.
+ynoeld_00001.wav|How old are you? Are you still here? Are you okay? What happened to you?
+ynoeld_00002.wav|We met in the apartment building. We met in the apartment building. Sure is quiet here, huh?
+ynoeld_00003.wav|What are you doing, Laura? Aren't you Maria?
+ynoeld_00004.wav|So, you're Maria? You're Maria? What's a little girl like you doing here, anyway?
+ynoeld_00005.wav|I didn't mean to scare you.
+ynoeld_00006.wav|Are you alone here, Eddie?
+ynoeld_00007.wav|James Sunderland.
+ynoeld_00008.wav|James. My, uh, my name's James.
+ynoeld_00009.wav|Eddie? Laura? Laura!
+ynoeld_00010.wav|Maria? Maria? Mary? Mary? Mary?
+ynoeld_00011.wav|Is that her name? Who is that girl, anyhow?
+ynoeld_00012.wav|Yeah. Yeah. Yes! Okay.
+ynoeld_00013.wav|That's right. Oh, yeah.
+ynoeld_00014.wav|I understand now.
+ynoeld_00015.wav|I believe you.
+ynoeld_00016.wav|No.
+ynoeld_00017.wav|No.
+ynoeld_00018.wav|No, I...
+ynoeld_00019.wav|No, you're not.
+ynoeld_00020.wav|Nothing.
+ynoeld_00021.wav|Oh, nothing.
+ynoeld_00022.wav|I guess I really don't care.
+ynoeld_00023.wav|I... don't know.
+ynoeld_00024.wav|Sorry.
+ynoeld_00025.wav|What?
+ynoeld_00026.wav|Do what?
+ynoeld_00027.wav|Tell me what happened.
+ynoeld_00028.wav|Why can't you just tell me?
+ynoeld_00029.wav|All I want from you is an answer.
+ynoeld_00030.wav|How do you know about that?
+ynoeld_00031.wav|Is there any other way?
+ynoeld_00032.wav|Don't you remember?
+ynoeld_00033.wav|Later, okay?
+ynoeld_00034.wav|Stay right there. I'll be there soon.
+ynoeld_00035.wav|I'll be there soon.
+ynoeld_00036.wav|Thanks.

finetunes/james-sunderland-v2/dataset/train.yaml

@@ -0,0 +1,147 @@
+name: james-finetune
+model: extensibletrainer
+scale: 1
+gpu_ids: [0] # Superfluous, redundant, unnecessary, the way you launch the training script will set this
+start_step: 0
+checkpointing_enabled: true 
+fp16: false
+wandb: false 
+use_tb_logger: true
+
+datasets:
+  train:
+    name: james-train
+    n_workers: 2
+    batch_size: 68
+    mode: paired_voice_audio
+    path: ./training/james/train.txt
+    fetcher_mode: ['lj']
+    phase: train
+    max_wav_length: 255995
+    max_text_length: 200
+    sample_rate: 22050
+    load_conditioning: True
+    num_conditioning_candidates: 2
+    conditioning_length: 44000
+    use_bpe_tokenizer: True
+    tokenizer_vocab: ./models/tortoise/bpe_lowercase_asr_256.json
+    load_aligned_codes: False
+  val: # I really do not care about validation right now
+    name: james-val
+    n_workers: 1
+    batch_size: 1
+    mode: paired_voice_audio
+    path: ./training/james/train.txt
+    fetcher_mode: ['lj']
+    phase: val
+    max_wav_length: 255995
+    max_text_length: 200
+    sample_rate: 22050
+    load_conditioning: True
+    num_conditioning_candidates: 2
+    conditioning_length: 44000
+    use_bpe_tokenizer: True
+    tokenizer_vocab: ./models/tortoise/bpe_lowercase_asr_256.json
+    load_aligned_codes: False
+
+steps:        
+  gpt_train:
+    training: gpt
+    loss_log_buffer: 500
+
+    # Generally follows the recipe from the DALLE paper.
+    optimizer: adamw # this should be adamw_zero if you're using distributed training
+    optimizer_params:
+      lr: !!float 1e-05 # originally: 1e-4
+      weight_decay: !!float 1e-2
+      beta1: 0.9
+      beta2: 0.96
+    clip_grad_eps: 4
+
+    injectors:
+      paired_to_mel:
+        type: torch_mel_spectrogram
+        mel_norm_file: ./models/tortoise/clips_mel_norms.pth
+        in: wav
+        out: paired_mel
+      paired_cond_to_mel:
+        type: for_each
+        subtype: torch_mel_spectrogram
+        mel_norm_file: ./models/tortoise/clips_mel_norms.pth
+        in: conditioning
+        out: paired_conditioning_mel
+      to_codes:
+        type: discrete_token
+        in: paired_mel
+        out: paired_mel_codes
+        dvae_config: "./models/tortoise/train_diffusion_vocoder_22k_level.yml"
+      paired_fwd_text:
+        type: generator
+        generator: gpt
+        in: [paired_conditioning_mel, padded_text, text_lengths, paired_mel_codes, wav_lengths]
+        out: [loss_text_ce, loss_mel_ce, logits]      
+    losses:
+      text_ce:
+        type: direct
+        weight: 0.01
+        key: loss_text_ce
+      mel_ce:
+        type: direct
+        weight: 1
+        key: loss_mel_ce
+
+networks:
+  gpt:
+    type: generator 
+    which_model_G: unified_voice2 # none of the unified_voice*.py files actually match the tortoise inference code... 4 and 3 have "alignment_head" (wtf is that?), 2 lacks the types=1 parameter.
+    kwargs:
+      layers: 30 # originally: 8
+      model_dim: 1024 # originally: 512
+      heads: 16 # originally: 8
+      max_text_tokens: 402 # originally: 120
+      max_mel_tokens: 604 # originally: 250
+      max_conditioning_inputs: 2 # originally: 1
+      mel_length_compression: 1024
+      number_text_tokens: 256 # supposed to be 255 for newer unified_voice files 
+      number_mel_codes: 8194
+      start_mel_token: 8192
+      stop_mel_token: 8193
+      start_text_token: 255
+      train_solo_embeddings: False # missing in uv3/4
+      use_mel_codes_as_input: True # ditto
+      checkpointing: True
+      #types: 1 # this is MISSING, but in my analysis 1 is equivalent to not having it.
+      #only_alignment_head: False  # uv3/4
+
+path:
+  # pretrain_model_gpt: './models/tortoise/autoregressive.pth' 
+  strict_load: true
+  resume_state: './training/james-finetune//training_state//500.state'
+
+train:
+  niter: 2000
+  warmup_iter: -1
+  mega_batch_factor: 34
+  val_freq: 2000
+
+  ema_enabled: false # I really don't think EMA matters
+
+  default_lr_scheme: MultiStepLR
+  gen_lr_steps: [9, 18, 25, 33] #[50000, 100000, 140000, 180000]
+  lr_gamma: 0.5
+
+eval:
+  output_state: gen
+  injectors:
+    gen_inj_eval:
+      type: generator
+      generator: generator
+      in: hq
+      out: [gen, codebook_commitment_loss]
+
+logger: 
+  print_freq: 5
+  save_checkpoint_freq: 50
+  visuals: [gen, mel]
+  visual_debug_rate: 5
+  is_mel_spectrogram: true

finetunes/james-sunderland-v2/dataset/whisper.json

@@ -0,0 +1,1741 @@
+{
+	"lxgbsm.wav": {
+		"text": " I'm kind of lost. I'm looking for Silent Hill. Silent Hill? Is this the right way? Well, there's the hotel 2, I guess. There's the hotel 2, I guess. The one on the lake? I wonder if it's still there. You're coming with me? Is it dangerous? I'll be careful. I guess I really don't care if it's dangerous or not. I'm going to town either way. I'm looking for someone. I'm looking for someone. Someone very important to me. I'm looking for Mary. Have you seen her? Without Mary I just can't go on. I'd do anything if I could be with her again. I'm gonna go look for her. I wish I knew. I killed her three years ago. Last year. But I got a letter from her. I got a letter from her. How do you know about Mary? You really seem to care about her. Is that why you're here too? You're not Mary. You're really not Mary. You're alive! I thought that thing killed you. Are you You hurt bad? Your face, your voice, just your hair and... I was confused. There are all sorts of strange things around here. This is no place for a kid. Eddie told me. Who's Rachel? What's that letter? It's time to end this nightmare. What the hell is it? I didn't come here to play, you know. You liar! Don't be ridiculous. How can you sit there and eat pizza? You! It was you, wasn't it? You're the one who stepped on my hand. I'm done with you. Forget you! Someone very important to me. You're not Mary.",
+		"segments": [
+			{
+				"id": 0,
+				"seek": 0,
+				"start": 0.0,
+				"end": 7.0,
+				"text": " I'm kind of lost. I'm looking for Silent Hill. Silent Hill? Is this the right way?",
+				"tokens": [
+					314,
+					1101,
+					1611,
+					286,
+					2626,
+					13,
+					314,
+					1101,
+					2045,
+					329,
+					25083,
+					3327,
+					13,
+					25083,
+					3327,
+					30,
+					1148,
+					428,
+					262,
+					826,
+					835,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.13676407516643566,
+				"compression_ratio": 1.6756756756756757,
+				"no_speech_prob": 0.0679725632071495,
+				"seg-text": [
+					" I'm kind of lost. I'm looking for Silent Hill. Silent Hill? Is this the right way?"
+				]
+			},
+			{
+				"id": 1,
+				"seek": 0,
+				"start": 7.0,
+				"end": 16.0,
+				"text": " Well, there's the hotel 2, I guess. There's the hotel 2, I guess. The one on the lake? I wonder if it's still there.",
+				"tokens": [
+					3894,
+					11,
+					612,
+					338,
+					262,
+					7541,
+					362,
+					11,
+					314,
+					4724,
+					13,
+					1318,
+					338,
+					262,
+					7541,
+					362,
+					11,
+					314,
+					4724,
+					13,
+					383,
+					530,
+					319,
+					262,
+					13546,
+					30,
+					314,
+					4240,
+					611,
+					340,
+					338,
+					991,
+					612,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.13676407516643566,
+				"compression_ratio": 1.6756756756756757,
+				"no_speech_prob": 0.0679725632071495,
+				"seg-text": [
+					" Well, there's the hotel 2, I guess. There's the hotel 2, I guess. The one on the lake? I wonder if it's still there."
+				]
+			},
+			{
+				"id": 2,
+				"seek": 0,
+				"start": 16.0,
+				"end": 25.0,
+				"text": " You're coming with me? Is it dangerous? I'll be careful. I guess I really don't care if it's dangerous or not.",
+				"tokens": [
+					921,
+					821,
+					2406,
+					351,
+					502,
+					30,
+					1148,
+					340,
+					4923,
+					30,
+					314,
+					1183,
+					307,
+					8161,
+					13,
+					314,
+					4724,
+					314,
+					1107,
+					836,
+					470,
+					1337,
+					611,
+					340,
+					338,
+					4923,
+					393,
+					407,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.13676407516643566,
+				"compression_ratio": 1.6756756756756757,
+				"no_speech_prob": 0.0679725632071495,
+				"seg-text": [
+					" You're coming with me? Is it dangerous? I'll be careful. I guess I really don't care if it's dangerous or not."
+				]
+			},
+			{
+				"id": 3,
+				"seek": 2500,
+				"start": 25.0,
+				"end": 32.5,
+				"text": " I'm going to town either way. I'm looking for someone. I'm looking for someone.",
+				"tokens": [
+					314,
+					1101,
+					1016,
+					284,
+					3240,
+					2035,
+					835,
+					13,
+					314,
+					1101,
+					2045,
+					329,
+					2130,
+					13,
+					314,
+					1101,
+					2045,
+					329,
+					2130,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.19361598563916754,
+				"compression_ratio": 1.5684931506849316,
+				"no_speech_prob": 0.09226225316524506,
+				"seg-text": [
+					" I'm going to town either way. I'm looking for someone. I'm looking for someone."
+				]
+			},
+			{
+				"id": 4,
+				"seek": 2500,
+				"start": 32.5,
+				"end": 40.46,
+				"text": " Someone very important to me. I'm looking for Mary. Have you seen her?",
+				"tokens": [
+					17877,
+					845,
+					1593,
+					284,
+					502,
+					13,
+					314,
+					1101,
+					2045,
+					329,
+					5335,
+					13,
+					8192,
+					345,
+					1775,
+					607,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.19361598563916754,
+				"compression_ratio": 1.5684931506849316,
+				"no_speech_prob": 0.09226225316524506,
+				"seg-text": [
+					" Someone very important to me. I'm looking for Mary. Have you seen her?"
+				]
+			},
+			{
+				"id": 5,
+				"seek": 2500,
+				"start": 40.46,
+				"end": 47.86,
+				"text": " Without Mary I just can't go on. I'd do anything if I could be with her again.",
+				"tokens": [
+					9170,
+					5335,
+					314,
+					655,
+					460,
+					470,
+					467,
+					319,
+					13,
+					314,
+					1549,
+					466,
+					1997,
+					611,
+					314,
+					714,
+					307,
+					351,
+					607,
+					757,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.19361598563916754,
+				"compression_ratio": 1.5684931506849316,
+				"no_speech_prob": 0.09226225316524506,
+				"seg-text": [
+					" Without Mary I just can't go on. I'd do anything if I could be with her again."
+				]
+			},
+			{
+				"id": 6,
+				"seek": 4786,
+				"start": 47.86,
+				"end": 56.78,
+				"text": " I'm gonna go look for her. I wish I knew. I killed her three years ago. Last year.",
+				"tokens": [
+					314,
+					1101,
+					8066,
+					467,
+					804,
+					329,
+					607,
+					13,
+					314,
+					4601,
+					314,
+					2993,
+					13,
+					314,
+					2923,
+					607,
+					1115,
+					812,
+					2084,
+					13,
+					4586,
+					614,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1430604533145302,
+				"compression_ratio": 1.7234042553191489,
+				"no_speech_prob": 0.1444021761417389,
+				"seg-text": [
+					" I'm gonna go look for her. I wish I knew. I killed her three years ago. Last year."
+				]
+			},
+			{
+				"id": 7,
+				"seek": 4786,
+				"start": 56.78,
+				"end": 61.379999999999995,
+				"text": " But I got a letter from her. I got a letter from her. How do you know about",
+				"tokens": [
+					887,
+					314,
+					1392,
+					257,
+					3850,
+					422,
+					607,
+					13,
+					314,
+					1392,
+					257,
+					3850,
+					422,
+					607,
+					13,
+					1374,
+					466,
+					345,
+					760,
+					546
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1430604533145302,
+				"compression_ratio": 1.7234042553191489,
+				"no_speech_prob": 0.1444021761417389,
+				"seg-text": [
+					" But I got a letter from her. I got a letter from her. How do you know about"
+				]
+			},
+			{
+				"id": 8,
+				"seek": 4786,
+				"start": 61.379999999999995,
+				"end": 68.1,
+				"text": " Mary? You really seem to care about her. Is that why you're here too? You're not",
+				"tokens": [
+					5335,
+					30,
+					921,
+					1107,
+					1283,
+					284,
+					1337,
+					546,
+					607,
+					13,
+					1148,
+					326,
+					1521,
+					345,
+					821,
+					994,
+					1165,
+					30,
+					921,
+					821,
+					407
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1430604533145302,
+				"compression_ratio": 1.7234042553191489,
+				"no_speech_prob": 0.1444021761417389,
+				"seg-text": [
+					" Mary? You really seem to care about her. Is that why you're here too? You're not"
+				]
+			},
+			{
+				"id": 9,
+				"seek": 4786,
+				"start": 68.1,
+				"end": 76.58,
+				"text": " Mary. You're really not Mary. You're alive! I thought that thing killed you. Are you",
+				"tokens": [
+					5335,
+					13,
+					921,
+					821,
+					1107,
+					407,
+					5335,
+					13,
+					921,
+					821,
+					6776,
+					0,
+					314,
+					1807,
+					326,
+					1517,
+					2923,
+					345,
+					13,
+					4231,
+					345
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1430604533145302,
+				"compression_ratio": 1.7234042553191489,
+				"no_speech_prob": 0.1444021761417389,
+				"seg-text": [
+					" Mary. You're really not Mary. You're alive! I thought that thing killed you. Are you"
+				]
+			},
+			{
+				"id": 10,
+				"seek": 7658,
+				"start": 76.58,
+				"end": 78.26,
+				"text": " You hurt bad?",
+				"tokens": [
+					921,
+					5938,
+					2089,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" You hurt bad?"
+				]
+			},
+			{
+				"id": 11,
+				"seek": 7658,
+				"start": 78.26,
+				"end": 81.7,
+				"text": " Your face, your voice, just your hair and...",
+				"tokens": [
+					3406,
+					1986,
+					11,
+					534,
+					3809,
+					11,
+					655,
+					534,
+					4190,
+					290,
+					986
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" Your face, your voice, just your hair and..."
+				]
+			},
+			{
+				"id": 12,
+				"seek": 7658,
+				"start": 81.7,
+				"end": 84.82,
+				"text": " I was confused.",
+				"tokens": [
+					314,
+					373,
+					10416,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" I was confused."
+				]
+			},
+			{
+				"id": 13,
+				"seek": 7658,
+				"start": 84.82,
+				"end": 87.44,
+				"text": " There are all sorts of strange things around here.",
+				"tokens": [
+					1318,
+					389,
+					477,
+					10524,
+					286,
+					6283,
+					1243,
+					1088,
+					994,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" There are all sorts of strange things around here."
+				]
+			},
+			{
+				"id": 14,
+				"seek": 7658,
+				"start": 87.44,
+				"end": 89.84,
+				"text": " This is no place for a kid.",
+				"tokens": [
+					770,
+					318,
+					645,
+					1295,
+					329,
+					257,
+					5141,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" This is no place for a kid."
+				]
+			},
+			{
+				"id": 15,
+				"seek": 7658,
+				"start": 89.84,
+				"end": 91.78,
+				"text": " Eddie told me.",
+				"tokens": [
+					19478,
+					1297,
+					502,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" Eddie told me."
+				]
+			},
+			{
+				"id": 16,
+				"seek": 7658,
+				"start": 91.78,
+				"end": 93.66,
+				"text": " Who's Rachel?",
+				"tokens": [
+					5338,
+					338,
+					15984,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" Who's Rachel?"
+				]
+			},
+			{
+				"id": 17,
+				"seek": 7658,
+				"start": 93.66,
+				"end": 95.46,
+				"text": " What's that letter?",
+				"tokens": [
+					1867,
+					338,
+					326,
+					3850,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" What's that letter?"
+				]
+			},
+			{
+				"id": 18,
+				"seek": 7658,
+				"start": 95.46,
+				"end": 97.94,
+				"text": " It's time to end this nightmare.",
+				"tokens": [
+					632,
+					338,
+					640,
+					284,
+					886,
+					428,
+					17123,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" It's time to end this nightmare."
+				]
+			},
+			{
+				"id": 19,
+				"seek": 7658,
+				"start": 97.94,
+				"end": 99.82,
+				"text": " What the hell is it?",
+				"tokens": [
+					1867,
+					262,
+					5968,
+					318,
+					340,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" What the hell is it?"
+				]
+			},
+			{
+				"id": 20,
+				"seek": 7658,
+				"start": 99.82,
+				"end": 102.08,
+				"text": " I didn't come here to play, you know.",
+				"tokens": [
+					314,
+					1422,
+					470,
+					1282,
+					994,
+					284,
+					711,
+					11,
+					345,
+					760,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" I didn't come here to play, you know."
+				]
+			},
+			{
+				"id": 21,
+				"seek": 7658,
+				"start": 102.08,
+				"end": 103.86,
+				"text": " You liar!",
+				"tokens": [
+					921,
+					31866,
+					0
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" You liar!"
+				]
+			},
+			{
+				"id": 22,
+				"seek": 7658,
+				"start": 103.86,
+				"end": 105.38,
+				"text": " Don't be ridiculous.",
+				"tokens": [
+					2094,
+					470,
+					307,
+					11441,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.16667714849248663,
+				"compression_ratio": 1.4840182648401827,
+				"no_speech_prob": 0.6022158265113831,
+				"seg-text": [
+					" Don't be ridiculous."
+				]
+			},
+			{
+				"id": 23,
+				"seek": 10538,
+				"start": 105.38,
+				"end": 108.17999999999999,
+				"text": " How can you sit there and eat pizza?",
+				"tokens": [
+					1374,
+					460,
+					345,
+					1650,
+					612,
+					290,
+					4483,
+					14256,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.25711625995058,
+				"compression_ratio": 1.3308823529411764,
+				"no_speech_prob": 0.058869555592536926,
+				"seg-text": [
+					" How can you sit there and eat pizza?"
+				]
+			},
+			{
+				"id": 24,
+				"seek": 10538,
+				"start": 108.17999999999999,
+				"end": 109.17999999999999,
+				"text": " You!",
+				"tokens": [
+					921,
+					0
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.25711625995058,
+				"compression_ratio": 1.3308823529411764,
+				"no_speech_prob": 0.058869555592536926,
+				"seg-text": [
+					" You!"
+				]
+			},
+			{
+				"id": 25,
+				"seek": 10538,
+				"start": 109.17999999999999,
+				"end": 110.82,
+				"text": " It was you, wasn't it?",
+				"tokens": [
+					632,
+					373,
+					345,
+					11,
+					2492,
+					470,
+					340,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.25711625995058,
+				"compression_ratio": 1.3308823529411764,
+				"no_speech_prob": 0.058869555592536926,
+				"seg-text": [
+					" It was you, wasn't it?"
+				]
+			},
+			{
+				"id": 26,
+				"seek": 10538,
+				"start": 110.82,
+				"end": 112.74,
+				"text": " You're the one who stepped on my hand.",
+				"tokens": [
+					921,
+					821,
+					262,
+					530,
+					508,
+					10764,
+					319,
+					616,
+					1021,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.25711625995058,
+				"compression_ratio": 1.3308823529411764,
+				"no_speech_prob": 0.058869555592536926,
+				"seg-text": [
+					" You're the one who stepped on my hand."
+				]
+			},
+			{
+				"id": 27,
+				"seek": 10538,
+				"start": 112.74,
+				"end": 114.61999999999999,
+				"text": " I'm done with you.",
+				"tokens": [
+					314,
+					1101,
+					1760,
+					351,
+					345,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.25711625995058,
+				"compression_ratio": 1.3308823529411764,
+				"no_speech_prob": 0.058869555592536926,
+				"seg-text": [
+					" I'm done with you."
+				]
+			},
+			{
+				"id": 28,
+				"seek": 10538,
+				"start": 114.61999999999999,
+				"end": 116.53999999999999,
+				"text": " Forget you!",
+				"tokens": [
+					29624,
+					345,
+					0
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.25711625995058,
+				"compression_ratio": 1.3308823529411764,
+				"no_speech_prob": 0.058869555592536926,
+				"seg-text": [
+					" Forget you!"
+				]
+			},
+			{
+				"id": 29,
+				"seek": 10538,
+				"start": 116.53999999999999,
+				"end": 120.38,
+				"text": " Someone very important to me.",
+				"tokens": [
+					17877,
+					845,
+					1593,
+					284,
+					502,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.25711625995058,
+				"compression_ratio": 1.3308823529411764,
+				"no_speech_prob": 0.058869555592536926,
+				"seg-text": [
+					" Someone very important to me."
+				]
+			},
+			{
+				"id": 30,
+				"seek": 12038,
+				"start": 120.38,
+				"end": 121.58,
+				"text": " You're not Mary.",
+				"tokens": [
+					50363,
+					921,
+					821,
+					407,
+					5335,
+					13,
+					50423
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.49764543771743774,
+				"compression_ratio": 0.6666666666666666,
+				"no_speech_prob": 0.273059606552124,
+				"seg-text": [
+					" You're not Mary."
+				]
+			}
+		],
+		"language": "en"
+	},
+	"ynoeld.wav": {
+		"text": " It's you. It's you. But I don't need you anymore. How old are you? Are you still here? Are you okay? What happened to you? We met in the apartment building. We met in the apartment building. Sure is quiet here, huh? What are you doing, Laura? Aren't you Maria? So, you're Maria? You're Maria? What's a little girl like you doing here, anyway? I didn't mean to scare you. Are you alone here, Eddie? James Sunderland. James. My, uh, my name's James. Eddie? Laura? Laura! Maria? Maria? Mary? Mary? Mary? Is that her name? Who is that girl, anyhow? Yeah. Yeah. Yes! Okay. That's right. Oh, yeah. I understand now. I believe you. No. No. No, I... No, you're not. Nothing. Oh, nothing. I guess I really don't care. I... don't know. Sorry. What? Do what? Tell me what happened. Why can't you just tell me? All I want from you is an answer. How do you know about that? Is there any other way? Don't you remember? Later, okay? Stay right there. I'll be there soon. I'll be there soon. Thanks.",
+		"segments": [
+			{
+				"id": 0,
+				"seek": 0,
+				"start": 0.0,
+				"end": 5.0,
+				"text": " It's you. It's you. But I don't need you anymore.",
+				"tokens": [
+					632,
+					338,
+					345,
+					13,
+					632,
+					338,
+					345,
+					13,
+					887,
+					314,
+					836,
+					470,
+					761,
+					345,
+					7471,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1540551746592802,
+				"compression_ratio": 1.7720207253886011,
+				"no_speech_prob": 0.13705560564994812,
+				"seg-text": [
+					" It's you. It's you. But I don't need you anymore."
+				]
+			},
+			{
+				"id": 1,
+				"seek": 0,
+				"start": 5.0,
+				"end": 11.0,
+				"text": " How old are you? Are you still here? Are you okay? What happened to you?",
+				"tokens": [
+					1374,
+					1468,
+					389,
+					345,
+					30,
+					4231,
+					345,
+					991,
+					994,
+					30,
+					4231,
+					345,
+					8788,
+					30,
+					1867,
+					3022,
+					284,
+					345,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1540551746592802,
+				"compression_ratio": 1.7720207253886011,
+				"no_speech_prob": 0.13705560564994812,
+				"seg-text": [
+					" How old are you? Are you still here? Are you okay? What happened to you?"
+				]
+			},
+			{
+				"id": 2,
+				"seek": 0,
+				"start": 11.0,
+				"end": 18.0,
+				"text": " We met in the apartment building. We met in the apartment building. Sure is quiet here, huh?",
+				"tokens": [
+					775,
+					1138,
+					287,
+					262,
+					7962,
+					2615,
+					13,
+					775,
+					1138,
+					287,
+					262,
+					7962,
+					2615,
+					13,
+					10889,
+					318,
+					5897,
+					994,
+					11,
+					24926,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1540551746592802,
+				"compression_ratio": 1.7720207253886011,
+				"no_speech_prob": 0.13705560564994812,
+				"seg-text": [
+					" We met in the apartment building. We met in the apartment building. Sure is quiet here, huh?"
+				]
+			},
+			{
+				"id": 3,
+				"seek": 0,
+				"start": 18.0,
+				"end": 22.0,
+				"text": " What are you doing, Laura? Aren't you Maria?",
+				"tokens": [
+					1867,
+					389,
+					345,
+					1804,
+					11,
+					16753,
+					30,
+					9843,
+					470,
+					345,
+					14200,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1540551746592802,
+				"compression_ratio": 1.7720207253886011,
+				"no_speech_prob": 0.13705560564994812,
+				"seg-text": [
+					" What are you doing, Laura? Aren't you Maria?"
+				]
+			},
+			{
+				"id": 4,
+				"seek": 0,
+				"start": 22.0,
+				"end": 29.0,
+				"text": " So, you're Maria? You're Maria? What's a little girl like you doing here, anyway?",
+				"tokens": [
+					1406,
+					11,
+					345,
+					821,
+					14200,
+					30,
+					921,
+					821,
+					14200,
+					30,
+					1867,
+					338,
+					257,
+					1310,
+					2576,
+					588,
+					345,
+					1804,
+					994,
+					11,
+					6949,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.1540551746592802,
+				"compression_ratio": 1.7720207253886011,
+				"no_speech_prob": 0.13705560564994812,
+				"seg-text": [
+					" So, you're Maria? You're Maria? What's a little girl like you doing here, anyway?"
+				]
+			},
+			{
+				"id": 5,
+				"seek": 2900,
+				"start": 29.0,
+				"end": 31.0,
+				"text": " I didn't mean to scare you.",
+				"tokens": [
+					314,
+					1422,
+					470,
+					1612,
+					284,
+					19437,
+					345,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" I didn't mean to scare you."
+				]
+			},
+			{
+				"id": 6,
+				"seek": 2900,
+				"start": 31.0,
+				"end": 33.0,
+				"text": " Are you alone here, Eddie?",
+				"tokens": [
+					4231,
+					345,
+					3436,
+					994,
+					11,
+					19478,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" Are you alone here, Eddie?"
+				]
+			},
+			{
+				"id": 7,
+				"seek": 2900,
+				"start": 33.0,
+				"end": 35.0,
+				"text": " James Sunderland.",
+				"tokens": [
+					3700,
+					35706,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" James Sunderland."
+				]
+			},
+			{
+				"id": 8,
+				"seek": 2900,
+				"start": 35.0,
+				"end": 38.5,
+				"text": " James. My, uh, my name's James.",
+				"tokens": [
+					3700,
+					13,
+					2011,
+					11,
+					21480,
+					11,
+					616,
+					1438,
+					338,
+					3700,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" James. My, uh, my name's James."
+				]
+			},
+			{
+				"id": 9,
+				"seek": 2900,
+				"start": 38.5,
+				"end": 42.0,
+				"text": " Eddie? Laura? Laura!",
+				"tokens": [
+					19478,
+					30,
+					16753,
+					30,
+					16753,
+					0
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" Eddie? Laura? Laura!"
+				]
+			},
+			{
+				"id": 10,
+				"seek": 2900,
+				"start": 42.0,
+				"end": 48.5,
+				"text": " Maria? Maria? Mary? Mary? Mary?",
+				"tokens": [
+					14200,
+					30,
+					14200,
+					30,
+					5335,
+					30,
+					5335,
+					30,
+					5335,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" Maria? Maria? Mary? Mary? Mary?"
+				]
+			},
+			{
+				"id": 11,
+				"seek": 2900,
+				"start": 48.5,
+				"end": 52.0,
+				"text": " Is that her name? Who is that girl, anyhow?",
+				"tokens": [
+					1148,
+					326,
+					607,
+					1438,
+					30,
+					5338,
+					318,
+					326,
+					2576,
+					11,
+					597,
+					4919,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" Is that her name? Who is that girl, anyhow?"
+				]
+			},
+			{
+				"id": 12,
+				"seek": 2900,
+				"start": 52.0,
+				"end": 56.0,
+				"text": " Yeah. Yeah. Yes! Okay.",
+				"tokens": [
+					9425,
+					13,
+					9425,
+					13,
+					3363,
+					0,
+					16805,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" Yeah. Yeah. Yes! Okay."
+				]
+			},
+			{
+				"id": 13,
+				"seek": 2900,
+				"start": 56.0,
+				"end": 58.5,
+				"text": " That's right. Oh, yeah.",
+				"tokens": [
+					1320,
+					338,
+					826,
+					13,
+					3966,
+					11,
+					10194,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.27043379114029253,
+				"compression_ratio": 1.521472392638037,
+				"no_speech_prob": 0.15195460617542267,
+				"seg-text": [
+					" That's right. Oh, yeah."
+				]
+			},
+			{
+				"id": 14,
+				"seek": 5850,
+				"start": 58.5,
+				"end": 60.22,
+				"text": " I understand now.",
+				"tokens": [
+					314,
+					1833,
+					783,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" I understand now."
+				]
+			},
+			{
+				"id": 15,
+				"seek": 5850,
+				"start": 60.22,
+				"end": 61.58,
+				"text": " I believe you.",
+				"tokens": [
+					314,
+					1975,
+					345,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" I believe you."
+				]
+			},
+			{
+				"id": 16,
+				"seek": 5850,
+				"start": 61.58,
+				"end": 62.58,
+				"text": " No.",
+				"tokens": [
+					1400,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" No."
+				]
+			},
+			{
+				"id": 17,
+				"seek": 5850,
+				"start": 62.58,
+				"end": 63.58,
+				"text": " No.",
+				"tokens": [
+					1400,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" No."
+				]
+			},
+			{
+				"id": 18,
+				"seek": 5850,
+				"start": 63.58,
+				"end": 65.58,
+				"text": " No, I...",
+				"tokens": [
+					1400,
+					11,
+					314,
+					986
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" No, I..."
+				]
+			},
+			{
+				"id": 19,
+				"seek": 5850,
+				"start": 65.58,
+				"end": 68.06,
+				"text": " No, you're not.",
+				"tokens": [
+					1400,
+					11,
+					345,
+					821,
+					407,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" No, you're not."
+				]
+			},
+			{
+				"id": 20,
+				"seek": 5850,
+				"start": 68.06,
+				"end": 69.06,
+				"text": " Nothing.",
+				"tokens": [
+					10528,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" Nothing."
+				]
+			},
+			{
+				"id": 21,
+				"seek": 5850,
+				"start": 69.06,
+				"end": 70.06,
+				"text": " Oh, nothing.",
+				"tokens": [
+					3966,
+					11,
+					2147,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" Oh, nothing."
+				]
+			},
+			{
+				"id": 22,
+				"seek": 5850,
+				"start": 70.06,
+				"end": 72.86,
+				"text": " I guess I really don't care.",
+				"tokens": [
+					314,
+					4724,
+					314,
+					1107,
+					836,
+					470,
+					1337,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" I guess I really don't care."
+				]
+			},
+			{
+				"id": 23,
+				"seek": 5850,
+				"start": 72.86,
+				"end": 75.66,
+				"text": " I... don't know.",
+				"tokens": [
+					314,
+					986,
+					836,
+					470,
+					760,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" I... don't know."
+				]
+			},
+			{
+				"id": 24,
+				"seek": 5850,
+				"start": 75.66,
+				"end": 76.66,
+				"text": " Sorry.",
+				"tokens": [
+					19061,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" Sorry."
+				]
+			},
+			{
+				"id": 25,
+				"seek": 5850,
+				"start": 76.66,
+				"end": 77.66,
+				"text": " What?",
+				"tokens": [
+					1867,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" What?"
+				]
+			},
+			{
+				"id": 26,
+				"seek": 5850,
+				"start": 77.66,
+				"end": 78.66,
+				"text": " Do what?",
+				"tokens": [
+					2141,
+					644,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" Do what?"
+				]
+			},
+			{
+				"id": 27,
+				"seek": 5850,
+				"start": 78.66,
+				"end": 81.66,
+				"text": " Tell me what happened.",
+				"tokens": [
+					14026,
+					502,
+					644,
+					3022,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" Tell me what happened."
+				]
+			},
+			{
+				"id": 28,
+				"seek": 5850,
+				"start": 81.66,
+				"end": 83.66,
+				"text": " Why can't you just tell me?",
+				"tokens": [
+					4162,
+					460,
+					470,
+					345,
+					655,
+					1560,
+					502,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" Why can't you just tell me?"
+				]
+			},
+			{
+				"id": 29,
+				"seek": 5850,
+				"start": 83.66,
+				"end": 85.66,
+				"text": " All I want from you is an answer.",
+				"tokens": [
+					1439,
+					314,
+					765,
+					422,
+					345,
+					318,
+					281,
+					3280,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" All I want from you is an answer."
+				]
+			},
+			{
+				"id": 30,
+				"seek": 5850,
+				"start": 85.66,
+				"end": 87.66,
+				"text": " How do you know about that?",
+				"tokens": [
+					1374,
+					466,
+					345,
+					760,
+					546,
+					326,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.3413577498051158,
+				"compression_ratio": 1.4972067039106145,
+				"no_speech_prob": 0.08281701803207397,
+				"seg-text": [
+					" How do you know about that?"
+				]
+			},
+			{
+				"id": 31,
+				"seek": 8766,
+				"start": 87.66,
+				"end": 89.66,
+				"text": " Is there any other way?",
+				"tokens": [
+					1148,
+					612,
+					597,
+					584,
+					835,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.15933908586916717,
+				"compression_ratio": 1.2842105263157895,
+				"no_speech_prob": 0.11254733055830002,
+				"seg-text": [
+					" Is there any other way?"
+				]
+			},
+			{
+				"id": 32,
+				"seek": 8766,
+				"start": 89.66,
+				"end": 91.66,
+				"text": " Don't you remember?",
+				"tokens": [
+					2094,
+					470,
+					345,
+					3505,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.15933908586916717,
+				"compression_ratio": 1.2842105263157895,
+				"no_speech_prob": 0.11254733055830002,
+				"seg-text": [
+					" Don't you remember?"
+				]
+			},
+			{
+				"id": 33,
+				"seek": 8766,
+				"start": 91.66,
+				"end": 92.66,
+				"text": " Later, okay?",
+				"tokens": [
+					11450,
+					11,
+					8788,
+					30
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.15933908586916717,
+				"compression_ratio": 1.2842105263157895,
+				"no_speech_prob": 0.11254733055830002,
+				"seg-text": [
+					" Later, okay?"
+				]
+			},
+			{
+				"id": 34,
+				"seek": 8766,
+				"start": 92.66,
+				"end": 95.66,
+				"text": " Stay right there. I'll be there soon.",
+				"tokens": [
+					16160,
+					826,
+					612,
+					13,
+					314,
+					1183,
+					307,
+					612,
+					2582,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.15933908586916717,
+				"compression_ratio": 1.2842105263157895,
+				"no_speech_prob": 0.11254733055830002,
+				"seg-text": [
+					" Stay right there. I'll be there soon."
+				]
+			},
+			{
+				"id": 35,
+				"seek": 8766,
+				"start": 95.66,
+				"end": 97.66,
+				"text": " I'll be there soon.",
+				"tokens": [
+					314,
+					1183,
+					307,
+					612,
+					2582,
+					13
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.15933908586916717,
+				"compression_ratio": 1.2842105263157895,
+				"no_speech_prob": 0.11254733055830002,
+				"seg-text": [
+					" I'll be there soon."
+				]
+			},
+			{
+				"id": 36,
+				"seek": 9766,
+				"start": 97.66,
+				"end": 98.66,
+				"text": " Thanks.",
+				"tokens": [
+					50363,
+					6930,
+					13,
+					50413
+				],
+				"temperature": 0.0,
+				"avg_logprob": -0.7431881904602051,
+				"compression_ratio": 0.4666666666666667,
+				"no_speech_prob": 0.008630666881799698,
+				"seg-text": [
+					" Thanks."
+				]
+			}
+		],
+		"language": "en"
+	}
+}

finetunes/james-sunderland-v2/dataset/ynoeld_00000.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e435018e3435a88ca51a8e744e3c842ee53dcab922360402fef6d736c88ca79
+size 882080

finetunes/james-sunderland-v2/dataset/ynoeld_00001.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7bff135be4a8d9a33d7b987b37d3e8fe6b9daa9618b5b0f5cd14287d78c576f
+size 1058480

finetunes/james-sunderland-v2/dataset/ynoeld_00002.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:362f390cd119e0d22c4b29a11fcf95b969c32138f9bd6e65fbd18d97e06b42c7
+size 1234880

finetunes/james-sunderland-v2/dataset/ynoeld_00003.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c34556f613d94aba9f79c7c105ddbd9468425bd5eea55402721dfe54e5e043bc
+size 705680

finetunes/james-sunderland-v2/dataset/ynoeld_00004.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:735ca52243a4626822a96ecfc64a8e304707282c961d06a3fba91ec5a6e4dd9d
+size 1234880

finetunes/james-sunderland-v2/dataset/ynoeld_00005.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d853f8d20a2e497aa5578caed08851a105f80fd6527daedf25f17d55fd907089
+size 352880

finetunes/james-sunderland-v2/dataset/ynoeld_00006.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8333ebc2ef88def5545ab1262fb3cd27dd639bc4d1d28c83d81c78983b365863
+size 352880

finetunes/james-sunderland-v2/dataset/ynoeld_00007.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3968685fff1ea301026b84feb82404d0436f033f8a270926ae3ee702444c66d
+size 352880

finetunes/james-sunderland-v2/dataset/ynoeld_00008.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a04b72788ffe2515c185a7408440e15b5f14ae98f51615ea0d316797d09af409
+size 617480

finetunes/james-sunderland-v2/dataset/ynoeld_00009.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1c82d2159fa64e178c5192400b5c49be356744b0c54aff0fc8da6becd42ce17
+size 617480

finetunes/james-sunderland-v2/dataset/ynoeld_00010.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10cde908ff5d1b9a3c27b428f9eb1f83d632d53f8e554f4d0a6b627c61db5564
+size 1146680

finetunes/james-sunderland-v2/dataset/ynoeld_00011.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57d1d75b7f7dd9fe4994eabc242889d86b8fb68df7853fe69d4585830351b745
+size 617480

finetunes/james-sunderland-v2/dataset/ynoeld_00012.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f70df39da3d3d0d1fc412e0ded5af3f32da788c85f86a21bca1f86cb23e6e8b
+size 705680

finetunes/james-sunderland-v2/dataset/ynoeld_00013.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:051ee0a96a55cf545477550647ab19b37dae9592c255ef8ed938744e24699881
+size 441080

finetunes/james-sunderland-v2/dataset/ynoeld_00014.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc13d0f9c0a70690358e98f66b75e5b0f7f4fb6c30e70c6c17e8b5f8734cd54a
+size 303488