huseinzolkepliscicom commited on
Commit
cab53b7
·
verified ·
1 Parent(s): 8a257c8

card: add PESQ column + bitrate-vs-UTMOSv2 plot

Browse files
Files changed (1) hide show
  1. README.md +10 -8
README.md CHANGED
@@ -48,14 +48,16 @@ codecs span a ~30× bitrate range. DAC/NeMo/EnCodec are high-bitrate multi-codeb
48
  *acoustic* codecs; WideCodec is a **single-codebook, 50-tokens/s** code at **~0.8 kbps** —
49
  the only one here whose token stream an autoregressive LLM/TTS model can practically predict.
50
 
51
- | codec | native SR | tokens/s | codebooks | ~bitrate | UTMOSv2 | mel-L1 ↓ | HF≥11k | rolloff |
52
- |---|--:|--:|--:|--:|--:|--:|--:|--:|
53
- | ground truth | | | | | 2.822 | | 0.00244 | 8263 |
54
- | nvidia nemo44k | 44.1k | ~86 | RVQ (many) | ~6–9 kbps | **2.903** | 0.379 | 0.00196 | 8372 |
55
- | **WideCodec (ours, gs-2.0M)** | 44.1k | **50** | **1** | **~0.8 kbps** | **2.788** | 0.571 | 0.00223 | 9011 |
56
- | dac | 44.1k | 86 | 9 | ~8 kbps | 2.672 | **0.341** | 0.00156 | 8185 |
57
- | snac44k | 44.1k | multi-scale | 3–4 | ~2.6 kbps | 2.340 | 0.493 | 0.00158 | 8402 |
58
- | encodec48k | 48k | 150 | RVQ | 24 kbps | 2.042 | 0.458 | 0.00169 | 8800 |
 
 
59
 
60
  **Bitrate = frame-rate × Σ log₂(codebook size).** WideCodec's FSQ is `levels=[4]×8,
61
  num_quantizers=1` → 8·log₂4 = **16 bits/frame** × 50 frames/s = **800 bps = 0.8 kbps**
 
48
  *acoustic* codecs; WideCodec is a **single-codebook, 50-tokens/s** code at **~0.8 kbps** —
49
  the only one here whose token stream an autoregressive LLM/TTS model can practically predict.
50
 
51
+ ![bitrate vs UTMOSv2](bitrate_vs_utmosv2.png)
52
+
53
+ | codec | native SR | tokens/s | codebooks | ~bitrate | UTMOSv2 | PESQ-WB | mel-L1 | HF≥11k | rolloff |
54
+ |---|--:|--:|--:|--:|--:|--:|--:|--:|--:|
55
+ | ground truth | | | | | 2.822 | | — | 0.00244 | 8263 |
56
+ | nvidia nemo44k | 44.1k | ~86 | RVQ (many) | ~6–9 kbps | **2.903** | **3.880** | 0.379 | 0.00196 | 8372 |
57
+ | **WideCodec (ours, gs-2.0M)** | 44.1k | **50** | **1** | **~0.8 kbps** | **2.788** | 2.079 | 0.571 | 0.00223 | 9011 |
58
+ | dac | 44.1k | 86 | 9 | ~8 kbps | 2.672 | 3.874 | **0.341** | 0.00156 | 8185 |
59
+ | snac44k | 44.1k | multi-scale | 3–4 | ~2.6 kbps | 2.340 | 2.600 | 0.493 | 0.00158 | 8402 |
60
+ | encodec48k | 48k | 150 | RVQ | 24 kbps | 2.042 | 3.115 | 0.458 | 0.00169 | 8800 |
61
 
62
  **Bitrate = frame-rate × Σ log₂(codebook size).** WideCodec's FSQ is `levels=[4]×8,
63
  num_quantizers=1` → 8·log₂4 = **16 bits/frame** × 50 frames/s = **800 bps = 0.8 kbps**