You need to agree to share your Hugging Face information to access this model
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
Access to this model is by request and is governed by the Hebrish terms of use (https://hebrish.app/terms). Non-commercial use is free; commercial use needs a commercial licence from access@hebrish.app.
Log in or Sign Up to review the conditions and access this model content.
Falafel 110M v1.0: Parakeet TDT-CTC 110M for spoken commands on phone and laptop
Falafel 110M is Hebrish's small speech-recognition model for short spoken commands on a phone or laptop, much of it English-Hebrew code-switching, in one speaker's voice. This is Falafel 110M v1.0, a fine-tune of Parakeet TDT-CTC 110M (transducer branch used for the sherpa build).
Name. Falafel is the code name of Hebrish's voice-command line; later retrains keep it and bump the version. This is the small model for phone and laptop; its larger sibling for the desktop is Falafel (0.6B). The training run id is r39-falafel-110m. Device: phone and laptop, as the sherpa-onnx int8 build (about 137 MB, transducer branch only, feat_dim 80); the .nemo here is the training checkpoint. CDN manifest: https://cdn.hebrish.app/models/falafel-110m/latest.json.
A full fine-tune of nvidia/parakeet-tdt_ctc-110m on one speaker's read-aloud phone and desktop recordings (an Irish English speaker living in Israel), run r39-falafel-110m, trained 2026-11-10 on one L40S (Modal) from stock. Output keeps casing and punctuation; numbers are written in digits.
Other Hebrish models: Falafel v1.0 (desktop, 0.6B) · Sabich, English-Hebrew code-switching, tech and general dictation.
Access and terms of use. Governed by the Hebrish terms of use; access is by request (button on this page, or access@hebrish.app). Non-commercial use is free; commercial use needs a commercial licence.
Results
Not benchmarked. Best validation WER 9.15% (best-epoch=19-val_wer=0.0915.ckpt) on primary_val: 325 held-out command takes, 11.4 min, from one speaker. It picked the checkpoint, so it is an indication of fit, not a benchmark and not comparable with other models' WER.
Training data
Source: danielrosehill/Read-Aloud-Voice-Sentences at revision da82b58d626a; the mix is rebuilt from it by prepare (blend commands).
- Rows tagged
voice-commands(4,816 rows, 170.3 min of the speaker's takes of short spoken commands to a computer or phone, much of it English-Hebrew code-switching; sentences indanielrosehill/Voice-Command-Sentences), plus 28.4 min of his own untagged English and LibriSpeech-PC replay at 2:1 (English share 0.20 of the blend); 21.3 min of that total is priority-word rows repeated 3x. Every other task tag is left out. - Mix: 220.1 min of the speaker's takes plus 14.4 min of LibriSpeech-PC replay (cased, punctuated; CC-BY-4.0) = 234.5 min, 5,519 clips.
- Validation: 1 in 16 of the speaker's train takes by id hash, used only to pick the checkpoint.
Training
NeMo 2.7.3 on one L40S (Modal), bf16. AdamW, lr 3e-05 (cosine, 10% warmup), 20 epochs,
batch 16, seed 7. Best checkpoint best-epoch=19-val_wer=0.0915.ckpt,
25.0 min of training.
The speaker's share of 220.1 min is 170.3 min of command takes, 28.4 min of his own untagged English and 21.3 min of priority-word repeats (363 rows, repeated 3x).
Usage
import nemo.collections.asr as nemo_asr
from huggingface_hub import hf_hub_download
path = hf_hub_download("Hebrish/Falafel-Parakeet-110M", "Falafel-Parakeet-110M.nemo")
model = nemo_asr.models.ASRModel.restore_from(path)
print(model.transcribe(["clip.wav"])[0].text) # 16 kHz mono
Builds: Hebrish/Falafel-Parakeet-110M-sherpa-onnx-int8 (sherpa-onnx int8, about 137 MB, transducer branch only, feat_dim 80; what the Hebrish apps download from the CDN manifest https://cdn.hebrish.app/models/falafel-110m/latest.json). No GGUF build.
Limitations
- One speaker. It is adapted to one voice, accent and phone; others will see less benefit.
- Tuned for short spoken commands to a computer or phone; long-form dictation and general prose are not what it is for (use Sabich or Sabich-Rolling).
- Not benchmarked. The only number is the best validation WER (9.15%) on
primary_val, 325 held-out command takes (11.4 min) picked by id hash from the same dataset revision. It selected the checkpoint and is not an independent test. - Less accurate than Falafel 0.6B (best validation WER 9.15% against 3.58% on the same
primary_val); it trades accuracy for size and speed on phone and laptop.
Licence
CC-BY-NC-4.0 (non-commercial), with use governed by the Hebrish terms above. Base model nvidia/parakeet-tdt_ctc-110m by NVIDIA, licensed under the CC-BY-4.0. Training data Read-Aloud-Voice-Sentences and LibriSpeech-PC are CC-BY-4.0.
- Downloads last month
- 12
Model tree for Hebrish/Falafel-Parakeet-110M
Base model
nvidia/parakeet-tdt_ctc-110m