whisburn

lyamc/whisburn

whisburn is a machine learning model from lyamc. 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 rust. The card lists the license as mit.

Pre-converted weights for whisburn on Burn 0.21.

Downloads · 30 days

78% of all-time downloads

All-time downloads

Public

Repo size

66.3 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors21.9 GB · 33%

At a glance

Library: rust
License: mit
Access: Public
Created: Sep 6, 2026
Updated: Oct 13, 2026
SHA: 896729cd

Model card

From the Hugging Face model README

whisburn Burn 0.21 bundles

Pre-converted weights for whisburn on Burn 0.21.

These are not the Gadersd/whisper-burn Burn 0.8 dumps.

Each JFK-checked model transcribed samples/jfk.wav before upload. Local copies were then deleted.

Models (JFK-checked)

Folder	Source	License
`tiny/`	openai/whisper-tiny	MIT
`tiny_en/`	openai/whisper-tiny.en	MIT
`base/`	openai/whisper-base	MIT
`base_en/`	openai/whisper-base.en	MIT
`small/`	openai/whisper-small	MIT
`small_en/`	openai/whisper-small.en	MIT
`medium/`	openai/whisper-medium	MIT
`medium_en/`	openai/whisper-medium.en	MIT
`large-v3-turbo/`	openai/whisper-large-v3-turbo	MIT
`distil-medium-en/`	distil-whisper/distil-medium.en	MIT
`distil-large-v3/`	distil-whisper/distil-large-v3	MIT
`parakeet-tdt-0.6b-v3/`	nvidia/parakeet-tdt-0.6b-v3	CC-BY-4.0
`parakeet-ctc-0.6b/`	nvidia/parakeet-ctc-0.6b	CC-BY-4.0
`parakeet-ctc-1.1b/`	nvidia/parakeet-ctc-1.1b	CC-BY-4.0
`moonshine-tiny/`	UsefulSensors/moonshine-tiny	MIT
`moonshine-base/`	UsefulSensors/moonshine-base	MIT
`t-one/`	t-tech/T-one	Apache-2.0
`bitnet-asr/`	microsoft/VibeVoice-ASR-BitNet	MIT
`qwen3-0.6b/`	Qwen/Qwen3-0.6B	Apache-2.0
`phonon-2/`	FermionResearch/Phonon-2	CC-BY-4.0
`nemotron-3.5-asr-streaming-0.6b/`	nvidia/nemotron-3.5-asr-streaming-0.6b	OpenMDW-1.1

qwen3-0.6b is the offline English summarizer (not ASR).

phonon-2 is the Phonon-2 five-value checkpoint expanded onto the Parakeet TDT v3 Burn graph (.burn_version 0.21.0, model.mpk.gz). On samples/jfk.wav it returns a partial sentence and drops the opening clause. A 1.62 hour English clip finished in 715 s (RTF 0.12).

nemotron-3.5-asr-streaming-0.6b is the NVIDIA streaming 0.6B RNN-T. Weights stay float32 model.safetensors (.burn_version 0.21.0, no .mpk). Their license is OpenMDW-1.1 (https://openmdw.ai/license/1-1/), not the MIT license in this card. Offline transcription encodes each window with the 56-frame left / 3-frame right attention mask, then greedy RNN-T, using the English prompt (en / en-US). On samples/jfk.wav it returned: "And so my fellow Americans ask not what your country can do for you. Ask what you can do for your country."

Also published

These folders are in the repo. They were not part of the original JFK pass above.

Folder	Source	License	Notes
`parakeet-tdt-0.6b-v2/`	nvidia/parakeet-tdt-0.6b-v2	CC-BY-4.0	`model.mpk.gz`
`qwen3-asr-0.6b/`	Qwen/Qwen3-ASR-0.6B	Apache-2.0	safetensors + `qwen3_runtime.json`
`qwen3-asr-1.7b/`	Qwen/Qwen3-ASR-1.7B	Apache-2.0	safetensors + `qwen3_runtime.json`
`vibevoice-asr/`	microsoft/VibeVoice-ASR-HF	MIT	`vibevoice-asr-q4_k.gguf`. Original JFK run did not finish in 40 minutes.
`silero-vad/`	snakers4/silero-vad	MIT	`vad_runtime.json`
`ten-vad/`	TEN-framework/ten-vad	Apache-2.0 with extra conditions	`vad_runtime.json` + `ten_features.json`
`diarization-3.1/`	pyannote/wespeaker-voxceleb-resnet34-LM	CC-BY-4.0	`embedding.safetensors`
`nemo-diarization/`	nvidia/speakerverification_en_titanet_large	CC-BY-4.0	`embedding.safetensors`

Use

cargo run -p whisburn-cli -- models download tiny_en --verbose
cargo run -p whisburn-cli -- transcribe --input audio.wav --model tiny_en

models download fetches these bundles first and falls back to local conversion.

Library: rust
License: mit
Created: Sep 6, 2026
Updated: Oct 13, 2026