Ir para o conteúdo Studio Crie imagens e vídeos sem viver dentro da cadeia de ferramentas de IA.
Um espaço de trabalho local-primeiro para transformar fluxos de trabalho ComfyUI em ferramentas que você pode usar, entender e retornar.
© 2026 Tea Time Labs
Criação local com IA, organizada.
Tem um workflow do ComfyUI? Gere uma documentação legível direto no navegador. Torne um workflow legível. Experimentar agora ↗
Ipa Whipstr Base 48khz Cv 21 | ST Studio
Recursos / Ipa Whipstr Base 48khz Cv 21 Ipa Whipstr Base 48khz Cv 21 A deep learning automatic speech recognition (ASR) system for transcribing speech audio into IPA text using transformer-based sequence-to-sequence models.
A deep learning automatic speech recognition (ASR) system for transcribing speech audio into IPA text using transformer-based sequence-to-sequence models.
Tipo Text model Versão checkpoint model
Fonte do modelo
Tipo Text model
Tamanho do arquivo 179 MB
Versão vcheckpoint
Atualizado 22 de set. de 2026
Downloads 14
Fonte Hugging Face Neurlang Whipstr STT (ASR)
A deep learning automatic speech recognition (ASR) system for transcribing speech audio into IPA text using transformer-based sequence-to-sequence models.
Artefatos de modelo
1 artefato checkpoint.pt
pt · 179 MB · SHA-256 dd7df9a815a5…f9f2 · Hugging Face
Baixar Trechos de fonte
2 trechos
Language: Universal (IPA), 74+ languages
Model Dataset: Common Voice 21
Model-Native Sample Rates: 8000 Hz, 16000 Hz, 24000 Hz, 32000 Hz, 48000 Hz
Degraded-Performance Sample Rates: 11025 Hz, 22050 Hz, 44100 Hz
License: GPL v2
Release: 2026-07-07
Size: 186 MBTotal parameters:
Encoder: 7 220 576Transformer: 7 537 184Total: 14 757 760
CER: 48.68% (51.32% success rate)
Note: Averaged across all supported languages, works better on higher resource languages
WER: 92.58% (7.42% success rate)
Note: Averaged across all supported languages, works better on higher resource languages
Training Details:
Hardware: Nvidia SparkBatch size: 1Samples: 960000Duration: 1:11:10:00Runs: Jul 5 14:58 - Jul 5 17:45, Jul 5 22:41 - Jul 6 04:38, Jul 6 04:46 - Jul 7 07:12 (shut down at Jul 7 07:36)
Inference code git clone https://github.com/neurlang/whipstr.git
cd whipstr/
uv run --with torch --with transformers --with phase-spectrogram stt_infer_hf.py --audio /home/m/Downloads/LJ001-0001.wav --model neurlang/ipa-whipstr-base-48khz-cv-21
Output: Loading weights: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 139/139 [00:00<00:00, 13371.75it/s]
Transcription: ˈɪt ˈæt juː pˈæst ðə kˈɔldz tˈɑɹk fˈeɪs vˈæst ðə ɹˈɛpɪtʃən lˈupənəl ˈi ˈæz bɪhˈeɪviɚ wˈi sˈɑloʊks lˈɛkəmˌɑɹəl hˈæn jˈæt lˈɚnd bˈeɪsɪks sˈɛktɪs ˈækmɪkmɪkəɹɪkənɪkəksɪksɪksɪksɪksɪksɪksɪksɪkstəkstəksɪksɪk
Explaination: IPA Ground truth Comment ˈɪt It Exact match. ˈæt juː gets you /ɡɛts/ was apparently lost, leaving something like "at you." pˈæst past Good match. ðə the Exact. kˈɔldz cold- Reasonable; final /z/ is likely a transcription artifact. tˈɑɹk start The /s/ was dropped and /st/ became /k/, a common recognition error in noisy speech. fˈeɪs phase Very close. vˈæst fast /f/→/v/ substitution. ðə
End of training data: Step 960000 is chosen as this release
Step WER CER 943000 94.21% 49.58% 944000 93.16% 49.96% 945000 94.09% 50.22% 946000 95.25% 50.32% 947000 96.18% 50.30% 948000 93.97% 50.78% 949000 92.93% 49.63% 950000 94.55% 50.84% 951000 94.90% 50.11% 952000 92.35%
---
license: gpl-2.0
language_bcp47:
- hy-AM
- nan-TW
- pa-IN
- sv-SE
- zh-CN
language:
- hy
- nan
- pa
- sv
- zh
- af
- am
- ar
- as
- az
- ba
- be
- bg
- bn
- ca
- cs
- da
- de
- el
- en
- eo
- es
- et
- eu
- fa
- fi
- fr
- gl
- gn
- he
- hi
- ht
- hu
- ia
- id
- is
- it
- ja
- kk
- ko
- ky
- lo
- lt
- ltg
- mk
- ml
- mn
- mr
- mt
- nl
- pl
- ps
- pt
- ro
- ru
- sd
- sk
- sl
- sq
- sw
- ta
- te
- th
- tk
- tn
- tr
- tt
- ug
- uk
- ur
- uz
- vi
- yue
- zu
pipeline_tag: automatic-speech-recognition
tags:
- whipstr
- stt
- asr
---
# Neurlang Whipstr STT (ASR)
A deep learning automatic speech recognition (ASR) system for transcribing speech audio into IPA text using transformer-based sequence-to-sequence models.
- **Language:** Universal (IPA), 74+ languages
- **Model Github:** neurlang/whipstr [https://github.com/neurlang/whipstr](https://github.com/neurlang/whipstr)
- **Model Dataset:** Common Voice 21
- **Model-Native Sample Rates:** 8000 Hz, 16000 Hz, 24000 Hz, 32000 Hz, 48000 Hz
- **Degraded-Performance Sample Rates:** 11025 Hz, 22050 Hz, 44100 Hz
- **License:** GPL v2
- **Release:** 2026-07-07
- **Size:** 186 MB
- **Total parameters:**
- **Encoder:** 7 220 576
- **Transformer:** 7 537 184
- **Total:** 14 757 760
- **CER:** 48.68% (51.32% success rate)
- **Note:** Averaged across all supported languages, works better on higher resource languages
- **WER:** 92.58% (7.42% success rate)
- **Note:** Averaged across all supported languages, works better on higher resource languages
- **Training Details:**
- **Hardware:** Nvidia Spark
- **Batch size:** 1
- **Samples:** 960000
- **Duration:** 1:11:10:00
- **Runs:** Jul 5 14:58 - Jul 5 17:45, Jul 5 22:41 - Jul 6 04:38, Jul 6 04:46 - Jul 7 07:12 (shut down at Jul 7 07:36)
---
## Inference code
```bash
git clone https://github.com/neurlang/whipstr.git
cd whipstr/
uv run --with torch --with transformers --with phase-spectrogram stt_infer_hf.py --audio /home/m/Downloads/LJ001-0001.wav --model neurlang/ipa-whipstr-base-48khz-cv-21
```
### Output:
```bash
Loading weights: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 139/139 [00:00<00:00, 13371.75it/s]
Transcription: ˈɪt ˈæt juː pˈæst ðə kˈɔldz tˈɑɹk fˈeɪs vˈæst ðə ɹˈɛ...
Source context: 14 downloads · 2 likes · Pipeline automatic-speech-recognition · Repo neurlang/ipa-whipstr-base-48khz-cv-21
ɹˈɛpɪtʃən repetition Clearly recognizable.
lˈupənəl loop / no-EOS This portion collapsed badly; "loop no-EOS" became something like "lupənəl."
ˈi ˈæz behavior (beginning omitted)The start of "behavior" appears to have shifted.
bɪhˈeɪviɚ behavior Excellent match.
sˈɑloʊks saw looks This is actually close to "saw looks," with the pause removed.
lˈɛkəmˌɑɹəl like the model Quite distorted, but you can hear the rough rhythm.
hˈæn hadn't The final /t/ and /d/ were lost.
bˈeɪsɪks basic Essentially correct.
sˈɛktɪs seq2seq This was the biggest failure—only the initial /sɛk/ of "seq2seq" survived, while
loops mechanics "mechanics" disappeared entirely.
