---
library_name: audio.cpp
license: cc-by-sa-4.0
pipeline_tag: automatic-speech-recognition
base_model:
- oruk/orukeet
tags:
- audio.cpp
- gguf
- parakeet
---

# Orukeet r3 GGUF

Follow or star [audio.cpp](https://github.com/0xShug0/audio.cpp) for inference code and model updates.

Orukeet r3 is a fine-tune of NVIDIA Parakeet-TDT-0.6B-v3. These packages run
with the existing `parakeet_tdt` family; no separate model implementation is required.
Each GGUF includes the model weights, tokenizer, configuration, conversion provenance,
license notices, and audio.cpp model contract.

| File | Weight precision |
|---|---|
| `orukeet-q8_0.gguf` | Q8_0 |
| `orukeet-f16.gguf` | F16 |

## Usage

```bash
audiocpp_cli --model Orukeet-GGUF/orukeet-q8_0.gguf \
  --family parakeet_tdt --task asr --backend cuda \
  --audio input.wav --text-out transcript.txt --words-out words.json --log
```

Use the F16 filename for the F16 package. See the
[Parakeet-TDT documentation](https://github.com/0xShug0/audio.cpp/blob/main/docs/models/parakeet_tdt.md)
for additional options and streaming support.

## Source and conversion

Converted from the official
[oruk/orukeet checkpoint](https://huggingface.co/oruk/orukeet/tree/555136b50265a132d4cea0d35560c26fc4f657ab),
pinned at `555136b50265a132d4cea0d35560c26fc4f657ab`.
The source file is `orukeet-v0.1.0.nemo`, SHA-256:

```text
031c8ddab4845aeced904a7cde8e8aa57993b2e344716cf83a545b079c473b56
```

Conversion uses audio.cpp's
[Orukeet converter](https://github.com/0xShug0/audio.cpp/blob/main/tools/community_models/convert_orukeet.py)
and `audiocpp_gguf`. The checkpoint is mapped to the existing Parakeet tensor
layout, retaining the stock tokenizer and compatible frontend configuration.

## License and attribution

Orukeet r3 weights and fitted kernels are **CC BY-SA 4.0**, retaining NVIDIA's
**CC BY 4.0** foundation attribution. See [LICENSE-WEIGHTS](LICENSE-WEIGHTS),
[NOTICE.md](NOTICE.md), and [conversion provenance](provenance.json).
The accompanying notices are from upstream revision
`b59c13a733fce5cf193230d0fa317ee7f145a108`.