A 17 MB speech recogniser that often names words before they end
A single-author preprint retrains a small open streaming model so its median word appears about 90 ms before the speaker finishes it, each word tagged with a usable confidence. The cost is more words taken back, and it is less accurate than larger models.

AI illustrationWhat the study shows
Streaming speech recognisers, which turn audio into text as it arrives, normally output a word only after the speaker has finished it. This preprint, which has not been peer reviewed, describes Earlybird, a model that tries to output each word while it is still being spoken, plus a guess at the next two words.
The author starts from Moonshine streaming Tiny, an open model released under the MIT licence, and trims its vocabulary to bring it from 34M to 29M parameters. It is then retrained on recordings cut off at random 80 ms points. At each cut, the target text includes every word that is 'due', set at roughly 300 ms before the word ends, written out in full even if only its opening sounds have been heard. Because each later target only extends the earlier one, the model is never trained to write text it later deletes. Quantisation, storing weights in fewer bits, shrinks the file to 16.8 MB at 4 bits.
Each word carries a confidence: the product of the probabilities the model gave its text pieces. No extra calibration step is applied.
The author fed every model audio in 80 ms chunks and logged when each word first appeared correctly and then stayed unchanged. Processing time was left out, so the numbers reflect what each model chooses to output, not hardware speed. Comparisons were with Moonshine Tiny's official runtime, Whistle (a model of similar file size) and Whisper base re-run every 0.48 s.
The author reports that on LibriSpeech test-clean (67 utterances) Earlybird's median word arrives 90 ms before it ends, against 270 ms after for Moonshine Tiny, 400 ms after for Whistle and 90 ms after for Whisper base. It was earliest on all six timed sets, covering audiobooks, TED talks, public speech, earnings calls and synthetic speech, with medians between 46 and 90 ms early. On LibriSpeech clean, 68.8% of words were right before they ended, against 3.4%, 9.1% and 25.4% for the three baselines.
For words still being spoken, expected calibration error (ECE, the average gap between stated confidence and actual hit rate) was 3.7 to 5.5 points. Words shown at 90% confidence or above were right 91% to 96% of the time.
On the eight English sets of the Open ASR Leaderboard, run on about 200 clips per set, the 4-bit model's mean word error rate (WER) was 12.6%, against 14.4% for Moonshine Tiny's runtime, 12.2% for Whistle and 9.7% for Whisper base. The weights and a loader are public under the MIT licence.
Our reading
In our reading, the useful idea is simple and cheap: change only the training targets, not the architecture, and a small recogniser learns to commit to words early. The gains in earliness are large and consistent across six test sets and three training seeds. The model also runs at about 39 ms per 80 ms step on one server CPU thread.
The confidence result matters as much as the timing. An early word is only worth acting on if the system knows when to trust it, and here the raw scores track accuracy well without any tuning. That gives a downstream agent a straightforward threshold to work with.
The trade-off is real. Earlybird takes back 1.29 words for every word spoken on clean audiobooks, against 0.16 for Whistle and 0.39 for Whisper base, and 2 to 6 on spontaneous or noisy speech. It is also less accurate than Whisper base, which is 2.5 times larger, on every set except clean audiobooks. We rate it level III, open source available: weights are released and it has public benchmark results, but no named organisation reports using it.
Read the rest of this story
Membership is free. You get every article in full and the Issue every Sunday.
We send a sign-in link, no password needed. Already a member? Sign inReadiness assessment
| Level | Status | Evidence |
|---|---|---|
| I Lab only | Met | A working trained model with a described method, ablation (padding frames) and seed-variance checks. |
| II Benchmarked | Met | Scored with the public Open ASR Leaderboard code on its eight English sets (mean WER 12.6% at 4-bit), with the author's Whisper base TED-LIUM score matching the published entry. Caveat: the scores come from about 200-clip subsets, are the author's own runs, and are not on the official leaderboard. The earliness metric is the author's own. |
| III Open source | Met | Weights and a loader are released under the MIT licence at https://huggingface.co/sahilmahendrakar/Earlybird-29M. The training code and data pipeline are not mentioned as released. I have not verified the repository. |
| IV In production | Not yet | No named organisation is documented using it in production, and no voice-agent deployment or end-to-end latency result is reported. |
What would move it up
Documented deployment by a named organisation, or a voice-agent test showing that early words and confidence cut end-to-end response time without errors from revisions, would support level 4. Full-set or official leaderboard results, independent replication, on-device speed measurements and training code would strengthen levels 2 and 3.
Checked automatically every week, last on 16 November.
The study
Sahil Mahendrakar. Earlybird: Predictive Streaming Speech Recognition in 29M Parameters. arXiv, 10 November 2026. doi:10.48550/arxiv.2611.14421
Notes
- Mahendrakar, S. Earlybird: Predictive Streaming Speech Recognition in 29M Parameters. arXiv preprint 2611.14421, November 2026. Not peer reviewed. CC BY licence. Source
- Earlybird-29M weights and loader, MIT licence. Source
- Moonshine streaming models and runtime, the starting point for Earlybird. Source
- Open ASR Leaderboard, whose evaluation code was used for the accuracy scores. Source
Edited and signed by
Stefan PerdalEditor and publisher86 stories editedStefan read the study, set the readiness level and approved this story. An AI system found the paper and wrote the first draft.
How this was made
An AI model read the study and wrote this article in the structure Science Underway uses. An editor reviewed it and approved it for publication. Check the original study before you rely on the details.
How we workNo corrections


