Show that WER matches NeMo, with the multilingual results

README.md

@@ -43,30 +43,39 @@ tags:
 
 # scriptrs models
 
-Model files for [scriptrs](https://github.com/avencera/scriptrs), a Rust library for fast NVIDIA Parakeet TDT v2 and v3 speech-to-text on Core ML (macOS) and native CUDA (Linux). With the default `online` feature, scriptrs downloads these files on first use from a pinned revision of this repository.
+Model files for [scriptrs](https://github.com/avencera/scriptrs), a Rust library for fast NVIDIA Parakeet TDT v2 and v3 speech-to-text on Core ML (macOS) and native CUDA (Linux), at the same word error rate as NVIDIA's reference implementation. With the default `online` feature, scriptrs downloads these files on first use from a pinned revision of this repository.
 
 These are format conversions of NVIDIA's models, not new or retrained models. The native CUDA weights are NVIDIA's FP32 weights with their original names; scriptrs converts selected layers to FP16 on the GPU at load time.
 
-## Speed
+## Same accuracy, faster
 
-RTX 5090, Parakeet v3, LibriSpeech test-clean (5.4 h), one utterance at a time:
+scriptrs runs NVIDIA's weights unchanged and gets the same word error rate as NVIDIA's own NeMo implementation, at 3x its speed. All results are Parakeet v3 on an RTX 5090.
+
+LibriSpeech test-clean, English (5.4 h), one utterance at a time:
 
 | | Speed (x real time) | WER |
 |---|---|---|
-| scriptrs CUDA (FP16 default) | 1534x | 1.93 % |
+| **scriptrs CUDA** | **1534x** | **1.93 %** |
 | NVIDIA NeMo (PyTorch) | 477x | 1.93 % |
 | parakeet-rs 0.4.3 (ONNX Runtime) | 253x | 1.93 % |
 
-More scriptrs CUDA results on the same GPU and model:
+### Multilingual
+
+The multilingual check is nearly 6 times the size of LibriSpeech: 31.6 h over ten language sets in six languages, one utterance at a time with automatic language selection. In scriptrs's parity tests, its FP32 mode matches NeMo layer by layer and gives NeMo's exact text. The FP16 default scores the same as FP32:
+
+| Set | Speed, FP16 default | WER, FP16 default | WER, FP32 |
+|---|---|---|---|
+| FLEURS de, es, fr, it, nl, pt (15.8 h) | 1839x | 4.57 % | 4.57 % |
+| MLS es, fr, it, pt (15.8 h) | 1921x | 6.25 % | 6.26 % |
+
+### Batches and long recordings
 
 | Workload | Speed (x real time) | WER |
 |---|---|---|
-| LibriSpeech test-clean, batches of 16 through the queue | 4767x | 1.93 % |
+| LibriSpeech test-clean, batches of 16 | 4767x | 1.93 %, unchanged |
 | Earnings-21, 44 calls (39 h), long-form | 3633x | 8.24 % |
-| FLEURS de, es, fr, it, nl, pt (15.8 h) | 1839x | 4.57 % |
-| MLS es, fr, it, pt (15.8 h) | 1921x | 6.25 % |
 
-More results are in the [scriptrs README](https://github.com/avencera/scriptrs#cuda).
+On Earnings-21, NeMo transcribing each whole call with local attention scores 9.58 %. More results, including harder English sets against Whisper large-v3, are in the [scriptrs README](https://github.com/avencera/scriptrs#cuda).
 
 ## Usage