Downloads · 30 days
20
19% of all-time downloads
Enriqueag26/keysay-transcription-cleaner-0.8B-v2
keysay-transcription-cleaner-0.8B-v2 is a machine learning model from Enriqueag26. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for mlx. The card lists the license as apache-2.0.
Cleans speech-to-text output for keysay: removes filler words and self-corrections from dictated text.
Downloads · 30 days
20
19% of all-time downloads
All-time downloads
104
Public
Parameters
752M
820 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors800 MB · 98%
How the weights are stored.
U32752M · 100%
From the Hugging Face model README
Cleans speech-to-text output for keysay: removes filler words and self-corrections from dictated text.
Deletion-only. The output is always the input with spans removed — the model never adds, replaces or reorders words. This is what makes it safe to paste straight into a text field, where the user cannot see what was changed.
from mlx_lm import load, generate
model, tokenizer = load("Enriqueag26/keysay-transcription-cleaner-0.8B-v2")
prompt = tokenizer.apply_chat_template(
[{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": "vale pues te llamo cuando salga del trabajo."}],
tokenize=False, add_generation_prompt=True, enable_thinking=False,
)
generate(model, tokenizer, prompt=prompt, max_tokens=256)
# -> "te llamo cuando salga del trabajo."
SYSTEM_PROMPT:
You clean speech-to-text transcriptions into ready-to-send chat messages.
Remove self-corrections (keep only the final version). Remove filler words.
Never rephrase. Never add words. Keep the original language.
Evaluated on 124 held-out hand-authored dictation scenarios, ~56% Spanish. The eval set is deliberately not generated by the model that produced the training data, so blind spots shared between teacher and student stay visible.
| metric | v1 | v2 |
|---|---|---|
| word match | 75.8% | 87.1% |
| delete F1 | 66.3% | 79.1% |
| deletion-only at inference | 99.2% | 98.4% |
| clean messages left untouched | 93.4% | 97.4% |
| Spanish word match | — | 87.5% |
LoRA via mlx_lm on Qwen3.5-0.8B-8bit — 16 layers, rank 8, 1500 iters, lr 3e-5.
3,113 synthetic pairs matched to the measured distribution of real Qwen3-ASR
output: ~33% lowercase-initial, lengths to ~100 words, ~46% Spanish, 48%
requiring no edit. Every label is verified reachable by deletion alone, and
labels are re-judged against an explicit rubric to keep discourse-marker
handling consistent (oye/vale/bueno as fillers vs. as verbs/adjectives).