# Waveform Generation on LibriTTS-R clean (test)

Source: [Waveform Generation on LibriTTS-R clean (test)](<https://www.sota2.com/research/sota/waveform-generation-on-libritts-r-clean-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T06:04:39.968Z

Evaluation of speech waveform generation models on the clean test split of the LibriTTS-R dataset, utilizing objective metrics including MCD, Log-F0-RMSE, and perceptual scores like N-MOS and Speech BERT Score.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Waveform Generation |
| Dataset (including any stated split) | LibriTTS-R clean (test) |
| Metric | Speech BERT Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Speech BERT Score](<https://www.sota2.com/research/sota/waveform-generation-on-libritts-r-clean-test.md?metric=Speech%20BERT%20Score>): Higher is better; unit: SCALAR.

- [MCD](<https://www.sota2.com/research/sota/waveform-generation-on-libritts-r-clean-test.md?metric=MCD>): Lower is better; unit: SCALAR.

- [Log-F0-RMSE](<https://www.sota2.com/research/sota/waveform-generation-on-libritts-r-clean-test.md?metric=Log-F0-RMSE>): Lower is better; unit: SCALAR.

- [Speaker Similarity](<https://www.sota2.com/research/sota/waveform-generation-on-libritts-r-clean-test.md?metric=Speaker%20Similarity>): Higher is better; unit: SCALAR.

- [N-MOS](<https://www.sota2.com/research/sota/waveform-generation-on-libritts-r-clean-test.md?metric=N-MOS>): Higher is better; unit: SCALAR.

- [S-MOS](<https://www.sota2.com/research/sota/waveform-generation-on-libritts-r-clean-test.md?metric=S-MOS>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Speech BERT Score | Result date | Report paper |
| --- | --- | --- | --- |
| Clean speech | 100 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| WaveTrainerFit | 95.28 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| WaveTrainerFit | 94.6 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| WaveTrainerFit | 94.39 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| WaveFit | 94.28 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| WaveFit | 94.11 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| WaveFit | 93.3 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| HiFi-GAN V1 | 91.09 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| HiFi-GAN V1 | 90.71 | Not recorded | 6985758ac97a0b3e3b5d5c70 |
| HiFi-GAN V1 | 88.9 | Not recorded | 6985758ac97a0b3e3b5d5c70 |

### Clean speech

Speech BERT Score: 100. Result ID: 699a108d5698665ec49437a9.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### WaveTrainerFit

Speech BERT Score: 95.28. Result ID: 699a108d5698665ec49437a2.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

Recorded evaluation configuration: {"SSL features":"WavLM","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### WaveTrainerFit

Speech BERT Score: 94.6. Result ID: 699a108d5698665ec49437a8.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

Recorded evaluation configuration: {"SSL features":"Whisper encoder","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### WaveTrainerFit

Speech BERT Score: 94.39. Result ID: 699a108d5698665ec49437a5.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

Recorded evaluation configuration: {"SSL features":"XLS-R","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### WaveFit

Speech BERT Score: 94.28. Result ID: 699a108d5698665ec49437a1.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d901e209131cea3aa80.md>)

Recorded evaluation configuration: {"SSL features":"WavLM","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### WaveFit

Speech BERT Score: 94.11. Result ID: 699a108d5698665ec49437a4.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d901e209131cea3aa80.md>)

Recorded evaluation configuration: {"SSL features":"XLS-R","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### WaveFit

Speech BERT Score: 93.3. Result ID: 699a108d5698665ec49437a7.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d901e209131cea3aa80.md>)

Recorded evaluation configuration: {"SSL features":"Whisper encoder","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### HiFi-GAN V1

Speech BERT Score: 91.09. Result ID: 699a108d5698665ec49437a3.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68a01e209131cea1d3a7.md>)

[Implementation](<https://github.com/jik876/hifi-gan>)

Recorded evaluation configuration: {"SSL features":"XLS-R","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### HiFi-GAN V1

Speech BERT Score: 90.71. Result ID: 699a108d5698665ec49437a0.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68a01e209131cea1d3a7.md>)

[Implementation](<https://github.com/jik876/hifi-gan>)

Recorded evaluation configuration: {"SSL features":"WavLM","SSL layer":8,"T":5}. This record does not assert independent reproduction.

### HiFi-GAN V1

Speech BERT Score: 88.9. Result ID: 699a108d5698665ec49437a6.

[Reporting paper](<https://www.sota2.com/research/paper/6985758ac97a0b3e3b5d5c70.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68a01e209131cea1d3a7.md>)

[Implementation](<https://github.com/jik876/hifi-gan>)

Recorded evaluation configuration: {"SSL features":"Whisper encoder","SSL layer":8,"T":5}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
