Upload folder using huggingface_hub

.gitattributes

@@ -71,3 +71,15 @@ HeartMuLa-GGUF/heartmula-f16.gguf filter=lfs diff=lfs merge=lfs -text
 HeartMuLa-GGUF/heartmula-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
 Qwen3-TTS-12Hz-1.7B-Base-GGUF/qwen3-tts-12hz-1.7b-base-q8_0_v2.gguf filter=lfs diff=lfs merge=lfs -text
 Voxtral-Mini-4B-Realtime-2602-GGUF/voxtral-mini-4b-realtime-2602-q4_k.gguf filter=lfs diff=lfs merge=lfs -text
+BS-RoFormer-ep368-GGUF/bs-roformer-ep368-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+Confucius4-TTS-GGUF/confucius4-tts-orig.gguf filter=lfs diff=lfs merge=lfs -text
+DotTTS-SOAR-GGUF/dots-tts-soar-bf16.gguf filter=lfs diff=lfs merge=lfs -text
+DotTTS-SOAR-GGUF/dots-tts-soar-orig.gguf filter=lfs diff=lfs merge=lfs -text
+DramaBox-GGUF/dramabox-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+Fun-ASR-Nano-2512-GGUF/fun-asr-nano-2512-f16.gguf filter=lfs diff=lfs merge=lfs -text
+Fun-ASR-Nano-2512-GGUF/fun-asr-nano-2512-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+Inflect-Micro-v2-GGUF/inflect-micro-v2-orig.gguf filter=lfs diff=lfs merge=lfs -text
+Kroko-ASR-GGUF/kroko-en-community-64-l-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+Parakeet-TDT-0.6B-v3-GGUF/parakeet-tdt-0.6b-v3-f16.gguf filter=lfs diff=lfs merge=lfs -text
+Parakeet-TDT-0.6B-v3-GGUF/parakeet-tdt-0.6b-v3-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+RVC-GGUF/rvc-f16.gguf filter=lfs diff=lfs merge=lfs -text

BS-RoFormer-ep368-GGUF/bs-roformer-ep368-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a55a8cad369d00f6e0fb208bb0cd87e30e25430772b8491e20a4eace6423ad2
+size 172532256

Confucius4-TTS-GGUF/confucius4-tts-orig.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eec4ab3fae3cda1e8ec2b96599442def30dbc980ea817f42b94e073158536244
+size 8192757760

DotTTS-SOAR-GGUF/dots-tts-soar-bf16.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5292089b053ffa55dddaa3eb5e888c7adefee75006f8d0e6c717a13c4933a41c
+size 4788783360

DotTTS-SOAR-GGUF/dots-tts-soar-orig.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4406cc394b9c25f5df92e3c053f64013cad1e2286271bc4509b6db795d13e67
+size 5165040864

DramaBox-GGUF/dramabox-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75e7e80fc748defb188cb902c34c62bc12539a7bba477215dccf59a7218a451e
+size 18942803808

Fun-ASR-Nano-2512-GGUF/MANIFEST.json

@@ -0,0 +1,19 @@
+{
+  "source": {
+    "repo": "FunAudioLLM/Fun-ASR-Nano-2512-hf",
+    "revision": "854d88f94205cd17d2afdb24332130d86fbe654a",
+    "model_safetensors_sha256": "335ca3e74917f1156690400e2c344350112950165789cf78ce3d0a367affd821"
+  },
+  "artifacts": [
+    {
+      "file": "fun-asr-nano-2512-q8_0.gguf",
+      "bytes": 1045334432,
+      "sha256": "4d727357574b079b7f43336b2930f39da086ca02f5d8d50872090b4c1c3d5e0a"
+    },
+    {
+      "file": "fun-asr-nano-2512-f16.gguf",
+      "bytes": 1675708832,
+      "sha256": "3d906c3ccfed07efef88ff53d6cc94b788b9d2edf1492a5679d041b43e98c5be"
+    }
+  ]
+}

Fun-ASR-Nano-2512-GGUF/README.md

@@ -0,0 +1,73 @@
+---
+library_name: audio.cpp
+license: other
+license_name: funasr-model-license-1.1
+license_link: https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512-hf/blob/main/README.md
+base_model: FunAudioLLM/Fun-ASR-Nano-2512-hf
+pipeline_tag: automatic-speech-recognition
+tags:
+- audio.cpp
+- gguf
+- speech-recognition
+- multilingual
+- funasr
+---
+
+# Fun-ASR-Nano-2512 GGUF
+
+Standalone audio.cpp GGUF builds of
+[FunAudioLLM/Fun-ASR-Nano-2512-hf](https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512-hf).
+Each file embeds the model configuration, processor configuration, tokenizer,
+chat template, and the audio.cpp model package specification.
+
+## Files
+
+| File | Size | SHA256 |
+| --- | ---: | --- |
+| `fun-asr-nano-2512-q8_0.gguf` | 1,045,334,432 bytes | `4d727357574b079b7f43336b2930f39da086ca02f5d8d50872090b4c1c3d5e0a` |
+| `fun-asr-nano-2512-f16.gguf` | 1,675,708,832 bytes | `3d906c3ccfed07efef88ff53d6cc94b788b9d2edf1492a5679d041b43e98c5be` |
+
+The source checkpoint is pinned to revision
+`854d88f94205cd17d2afdb24332130d86fbe654a`. The source
+`model.safetensors` SHA256 is
+`335ca3e74917f1156690400e2c344350112950165789cf78ce3d0a367affd821`.
+
+## audio.cpp
+
+```bash
+audiocpp_cli \
+  --task asr \
+  --family fun_asr_nano \
+  --model fun-asr-nano-2512-q8_0.gguf \
+  --backend cuda \
+  --audio speech.wav
+```
+
+Fun-ASR-Nano currently provides offline multilingual ASR. It does not expose
+streaming or timestamp output. On CUDA, audio.cpp keeps the Q8_0 encoder and
+adaptor weights native and loads decoder weights as BF16 by default for stable
+logits. An explicit `fun_asr_nano.decoder_weight_type` session option overrides
+that default.
+
+## Reproducibility
+
+The files were generated with audio.cpp's `audiocpp_gguf` converter:
+
+```bash
+audiocpp_gguf \
+  --input model.safetensors \
+  --root /path/to/Fun-ASR-Nano-2512-hf \
+  --output fun-asr-nano-2512-q8_0.gguf \
+  --type q8_0 \
+  --family fun_asr_nano \
+  --model-spec model_specs/fun_asr_nano.json
+```
+
+Both formats were checked with `audiocpp_gguf --inspect` and full reference
+audio transcription on CPU and NVIDIA H100 CUDA.
+
+## License
+
+The original model and these converted weights are governed by the
+FunASR Model Open Source License Agreement v1.1 distributed with the source
+model. Review that agreement before using or redistributing the files.

Fun-ASR-Nano-2512-GGUF/fun-asr-nano-2512-f16.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d906c3ccfed07efef88ff53d6cc94b788b9d2edf1492a5679d041b43e98c5be
+size 1675708832

Fun-ASR-Nano-2512-GGUF/fun-asr-nano-2512-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d727357574b079b7f43336b2930f39da086ca02f5d8d50872090b4c1c3d5e0a
+size 1045334432

Inflect-Micro-v2-GGUF/inflect-micro-v2-orig.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4af1cb6a92cdd8be550e8e7c25805ece222ec0f8e75daf26fc00b4e04ef4b03
+size 72082176

Kroko-ASR-GGUF/kroko-en-community-64-l-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:596c749fcb6c582f2d7aaa9aefcff2929e405867a4d04864a55b2c9c2963baa5
+size 167756928

Parakeet-TDT-0.6B-v3-GGUF/parakeet-tdt-0.6b-v3-f16.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83622c44275694039e543e135ac91b40d94f48fc80a1bf187b5a1cbcd5dd8649
+size 1255384320

Parakeet-TDT-0.6B-v3-GGUF/parakeet-tdt-0.6b-v3-q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:074e61ac1abd3d3efcfc10798d17bf9a975b31768b466fc2578364d206dde64c
+size 915733744

README.md

@@ -23,7 +23,10 @@ base_model:
 - Aratako/MioCodec-25Hz-44.1kHz-v2
 - Aratako/MioTTS-1.7B
 - Aratako/Semantic-DACVAE-Japanese-32dim
+- Banafo/Kroko-ASR
+- dots-studio/dots.tts-soar
 - fishaudio/s2-pro
+- FunAudioLLM/Fun-ASR-Nano-2512-hf
 - HeartMuLa/HeartCodec-oss-20260123
 - HeartMuLa/HeartMuLa-oss-3B
 - HeartMuLa/HeartMuLaGen
@@ -49,6 +52,7 @@ base_model:
 - microsoft/VibeVoice-1.5B
 - microsoft/VibeVoice-ASR
 - mistralai/Voxtral-Mini-4B-Realtime-2602
+- mirek190/audio.cpp
 - mlx-community/SeedVC-MLX
 - mlx-community/index-tts2-mlx
 - mlx-community/mel-roformer-mlx
@@ -56,6 +60,8 @@ base_model:
 - mlx-community/wavlm-base-plus-mlx
 - nvidia/diar_sortformer_4spk-v1
 - nvidia/nemotron-3.5-asr-streaming-0.6b
+- nvidia/parakeet-tdt-0.6b-v3
+- owensong/Inflect-Micro-v2
 - stabilityai/stable-audio-3-medium
 - stabilityai/stable-audio-3-small-music
 - stabilityai/stable-audio-3-small-sfx
@@ -80,16 +86,23 @@ for the full matrix and drift notes.
 | Directory | Files | audio.cpp family | Tested | Original model license |
 |---|---|---|---|---|
 | `ACE-Step1.5-GGUF` | `base/ace-step-1.5-base-bf16.gguf`, `base/ace-step-1.5-base-q8_0.gguf`, `turbo/ace-step-1.5-turbo-bf16.gguf`, `turbo/ace-step-1.5-turbo-q8_0.gguf` | `ace_step` | 16-bit + Q8 drift | See original model package |
+| `BS-RoFormer-ep368-GGUF` | `bs-roformer-ep368-q8_0.gguf` | `bs_roformer` | Q8 pass | See original model package |
 | `Chatterbox-GGUF` | `chatterbox-f16.gguf`, `chatterbox-q8_0.gguf` | `chatterbox` | 16-bit + Q8 ASR-match drift | MIT |
 | `Citrinet-ASR-GGUF` | `citrinet-asr-q8_0.gguf` | `citrinet_asr` | Q8 pass | CC-BY-4.0 |
+| `Confucius4-TTS-GGUF` | `confucius4-tts-orig.gguf` | `confucius4_tts` | orig pass | See original model package |
+| `DotTTS-SOAR-GGUF` | `dots-tts-soar-orig.gguf`, `dots-tts-soar-bf16.gguf` | `dots_tts` | experimental | See original model package |
+| `DramaBox-GGUF` | `dramabox-q8_0.gguf` | `dramabox` | Q8 pass | See original model package |
 | `Fish-Audio-S2-Pro-GGUF` | `fish-audio-s2-pro-bf16.gguf`, `fish-audio-s2-pro-q8_0.gguf` | `fish_audio` | 16-bit + Q8 pass | See original model package |
+| `Fun-ASR-Nano-2512-GGUF` | `fun-asr-nano-2512-f16.gguf`, `fun-asr-nano-2512-q8_0.gguf` | `fun_asr_nano` | 16-bit + Q8 pass | FunASR Model Open Source License Agreement v1.1 |
 | `HeartMuLa-GGUF` | `heartmula-f16.gguf`, `heartmula-q8_0.gguf` | `heartmula` | 16-bit + Q8 drift | Apache-2.0 |
 | `HTDemucs-GGUF` | `htdemucs-f16.gguf`, `htdemucs-q8_0.gguf` | `htdemucs` | 16-bit pass, Q8 drift | See original model package |
 | `Higgs-Audio-v3-STT-GGUF` | `higgs-audio-v3-stt-f16.gguf`, `higgs-audio-v3-stt-q8_0.gguf` | `higgs_audio_stt` | 16-bit + Q8 pass | Apache-2.0 |
 | `Higgs-Audio-v3-TTS-4B-GGUF` | `higgs-audio-v3-tts-4b-bf16.gguf`, `higgs-audio-v3-tts-4b-q8_0.gguf` | `higgs_audio_tts` | 16-bit + Q8 pass | See original model package |
 | `IndexTTS2-GGUF` | `index-tts2-orig.gguf`, `index-tts2-f16.gguf`, `index-tts2-q8_0.gguf` | `index_tts2` | orig + 16-bit pass/drift, Q8 ASR-match drift | bilibili Model Use License Agreement |
+| `Inflect-Micro-v2-GGUF` | `inflect-micro-v2-orig.gguf` | `inflect_v2` | orig pass | See original model package |
 | `Irodori-TTS-500M-v3-GGUF` | `irodori-tts-500m-v3-f16.gguf`, `irodori-tts-500m-v3-q8_0.gguf` | `irodori_tts` | 16-bit pass, Q8 drift | MIT |
 | `Irodori-TTS-600M-v3-VoiceDesign-GGUF` | `irodori-tts-600m-v3-voicedesign-f16.gguf`, `irodori-tts-600m-v3-voicedesign-q8_0.gguf` | `irodori_tts` | 16-bit pass, Q8 drift | MIT |
+| `Kroko-ASR-GGUF` | `kroko-en-community-64-l-q8_0.gguf` | `kroko_asr` | Q8 pass | See original model package |
 | `MOSS-TTS-Local-v1.5-GGUF` | `moss-tts-local-v1.5-bf16.gguf`, `moss-tts-local-v1.5-q8_0.gguf` | `moss_tts_local` | 16-bit pass, Q8 ASR-match drift | Apache-2.0 |
 | `MOSS-TTS-Nano-100M-GGUF` | `moss-tts-nano-100m-bf16.gguf`, `moss-tts-nano-100m-q8_0.gguf` | `moss_tts_nano` | 16-bit pass, Q8 ASR-match drift | Apache-2.0 |
 | `Mel-Band-RoFormer-GGUF` | `mel-band-roformer-f16.gguf`, `mel-band-roformer-q8_0.gguf` | `mel_band_roformer` | 16-bit + Q8 drift | MIT |
@@ -97,13 +110,15 @@ for the full matrix and drift notes.
 | `MioTTS-1.7B-GGUF` | `miotts-1.7b-orig.gguf`, `miotts-1.7b-bf16.gguf`, `miotts-1.7b-q8_0.gguf` | `miotts` | orig pass, 16-bit drift, Q8 ASR-match drift | Apache-2.0 |
 | `Nemotron-3.5-ASR-Streaming-0.6B-GGUF` | `nemotron-3.5-asr-streaming-0.6b-f16.gguf`, `nemotron-3.5-asr-streaming-0.6b-q8_0.gguf` | `nemotron_asr` | 16-bit pass, Q8 minor filler drift | OpenMDW-1.1 |
 | `OmniVoice-GGUF` | `omnivoice-bf16.gguf`, `omnivoice-f16.gguf`, `omnivoice-q8_0.gguf` | `omnivoice` | 16-bit + Q8 drift | Apache-2.0 |
+| `Parakeet-TDT-0.6B-v3-GGUF` | `parakeet-tdt-0.6b-v3-f16.gguf`, `parakeet-tdt-0.6b-v3-q8_0.gguf` | `parakeet_tdt` | 16-bit + Q8 pass | See original model package |
 | `PocketTTS-GGUF` | `english/`, `german/`, `italian/`, `portuguese/`, `spanish/` each contain `bf16` and `q8_0` GGUFs | `pocket_tts` | 16-bit pass, Q8 drift | See original model package |
 | `Qwen3-ASR-0.6B-GGUF` | `qwen3-asr-0.6b-f16.gguf`, `qwen3-asr-0.6b-q8_0.gguf` | `qwen3_asr` | 16-bit + Q8 pass | Apache-2.0 |
 | `Qwen3-ASR-1.7B-GGUF` | `qwen3-asr-1.7b-f16.gguf`, `qwen3-asr-1.7b-q8_0.gguf` | `qwen3_asr` | 16-bit + Q8 pass | Apache-2.0 |
 | `Qwen3-ForcedAligner-0.6B-GGUF` | `qwen3-forced-aligner-0.6b-f16.gguf`, `qwen3-forced-aligner-0.6b-q8_0.gguf` | `qwen3_forced_aligner` | 16-bit + Q8 pass | Apache-2.0 |
-| `Qwen3-TTS-12Hz-1.7B-Base-GGUF` | `qwen3-tts-12hz-1.7b-base-orig.gguf`, `qwen3-tts-12hz-1.7b-base-bf16.gguf`, `qwen3-tts-12hz-1.7b-base-q8_0.gguf` | `qwen3_tts` | orig pass, 16-bit + Q8 ASR-match drift | Apache-2.0 |
+| `Qwen3-TTS-12Hz-1.7B-Base-GGUF` | `qwen3-tts-12hz-1.7b-base-orig.gguf`, `qwen3-tts-12hz-1.7b-base-bf16.gguf`, `qwen3-tts-12hz-1.7b-base-q8_0_v2.gguf` | `qwen3_tts` | orig pass, 16-bit + Q8 ASR-match drift | Apache-2.0 |
 | `Qwen3-TTS-12Hz-1.7B-CustomVoice-GGUF` | `qwen3-tts-12hz-1.7b-customvoice-bf16.gguf`, `qwen3-tts-12hz-1.7b-customvoice-q8_0.gguf` | `qwen3_tts` | 16-bit + Q8 ASR-match drift | Apache-2.0 |
 | `Qwen3-TTS-12Hz-1.7B-VoiceDesign-GGUF` | `qwen3-tts-12hz-1.7b-voicedesign-bf16.gguf`, `qwen3-tts-12hz-1.7b-voicedesign-q8_0.gguf` | `qwen3_tts` | 16-bit + Q8 ASR-match drift | Apache-2.0 |
+| `RVC-GGUF` | `rvc-f16.gguf` | `rvc` | F16 pass | See original model package |
 | `SeedVC-MLX-GGUF` | `seed-vc-mlx-orig.gguf`, `seed-vc-mlx-f16.gguf`, `seed-vc-mlx-q8_0.gguf` | `seed_vc` | 16-bit + Q8 drift | GPL-3.0 |
 | `Sortformer-Diar-4spk-v1-GGUF` | `sortformer-diar-4spk-v1-f16.gguf`, `sortformer-diar-4spk-v1-q8_0.gguf` | `sortformer_diar` | 16-bit + Q8 pass | CC-BY-NC-4.0 |
 | `Stable-Audio-3-Medium-GGUF` | `stable-audio-3-medium-f16.gguf`, `stable-audio-3-medium-q8_0.gguf` | `stable_audio` | 16-bit + Q8 drift | Stability AI Community License |

RVC-GGUF/rvc-f16.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ddfc32ab63199bcca4bcd95a007a8add35ccff7e2c20e1272f19b9617410b57
+size 1260188416

Voxtral-Mini-4B-Realtime-2602-GGUF/README.md

@@ -0,0 +1,172 @@
+---
+library_name: audio.cpp
+pipeline_tag: automatic-speech-recognition
+tags:
+- audio.cpp
+- gguf
+- voxtral
+- asr
+- streaming-asr
+---
+
+# Voxtral Mini 4B Realtime GGUF for audio.cpp
+
+This repository contains quantized standalone GGUF checkpoints for running Voxtral Mini 4B Realtime ASR with [audio.cpp](https://github.com/0xShug0/audio.cpp). The GGUF files embed the audio.cpp model spec and required sidecars, so the model can be used directly from the checkpoint path without a separate local model-spec directory.
+
+## What audio.cpp does
+
+audio.cpp is a young C++/GGML audio inference framework focusing on CUDA performance. It runs speech and audio models locally with CLI and server interfaces, including ASR, TTS, voice conversion, source separation, diarization, VAD, and audio generation models. The project currently tracks 35+ model families and is growing quickly. For Voxtral Realtime, audio.cpp provides offline and streaming speech recognition from a single GGUF checkpoint.
+
+## Files
+
+| File | Quantization | Size | Recommended use |
+| --- | --- | ---: | --- |
+| `voxtral-mini-4b-realtime-2602-q8_0.gguf` | Q8_0 | 4.8 GiB | Default balanced checkpoint for high-quality ASR with lower memory than BF16. |
+| `voxtral-mini-4b-realtime-2602-q4_k.gguf` | Q4_K | 2.9 GiB | Lower-memory and faster checkpoint for CUDA testing and deployment. Validate output quality for your domain. |
+
+## Performance
+
+Measurements below are from audio.cpp CUDA validation runs. Results vary by GPU, driver, backend, audio length, and decode settings.
+
+### Q8_0 vs BF16 reference
+
+BF16 was used only as a local reference baseline for validation. This repository publishes quantized GGUF checkpoints only.
+
+| Mode | BF16 reference | Q8_0 | Q8_0 improvement |
+| --- | ---: | ---: | ---: |
+| Offline ASR speed | 11.1x-12.5x realtime | 14.7x-16.7x realtime | 1.31x-1.38x faster |
+| Offline ASR peak VRAM | 10,909 MiB | 7,754 MiB | 3,155 MiB lower |
+| Streaming server TTFT | 207.308 ms | 179.896 ms | 27.412 ms lower |
+| Streaming client TTFT | 550.526 ms | 530.558 ms | 19.968 ms lower |
+| Streaming speed | 4.7x realtime | 5.4x realtime | 1.15x faster |
+| Streaming peak VRAM | 12,616 MiB | 8,972 MiB | 3,644 MiB lower |
+
+### Q4_K quick check
+
+| Route | Q8_0 RTF | Q4_K RTF | Q4_K vs Q8_0 |
+| --- | ---: | ---: | ---: |
+| Offline short | 0.0862 | 0.0629 | 1.37x faster |
+| Offline medium | 0.0643 | 0.0476 | 1.35x faster |
+| Offline longer | 0.0576 | 0.0439 | 1.31x faster |
+| Offline sampled | 0.0630 | 0.0500 | 1.26x faster |
+| Streaming path | 0.1036 | 0.0904 | 1.15x faster |
+
+In the quick validation set, Q4_K transcripts matched Q8_0 except for one capitalization-only difference.
+
+## Use with audio.cpp
+
+Build audio.cpp with the helper script for your platform, then use the generated `audiocpp_cli` binary. The project provides build paths for Linux, Windows, and macOS; see the [audio.cpp README](https://github.com/0xShug0/audio.cpp#build) for the current build matrix and detailed requirements.
+
+```bash
+git clone https://github.com/0xShug0/audio.cpp
+cd audio.cpp
+
+# Linux: CUDA, Vulkan, or CPU
+scripts/build_linux.sh --backend cuda --target audiocpp_cli
+
+# Windows: CUDA or CPU presets
+powershell.exe -NoProfile -ExecutionPolicy Bypass -File .\scripts\build_windows.ps1 -Preset windows-cuda-release -Target audiocpp_cli
+
+# macOS: Metal
+scripts/build_metal.sh --target audiocpp_cli
+```
+
+The examples below assume `audiocpp_cli` is on your `PATH`. You can also replace it with the built binary path for your platform, such as `build/linux-cuda-release/bin/audiocpp_cli`.
+
+Run offline ASR with Q8_0:
+
+```bash
+MODEL=/path/to/Voxtral-Mini-4B-Realtime-2602-GGUF/voxtral-mini-4b-realtime-2602-q8_0.gguf
+
+audiocpp_cli \
+  --task asr \
+  --family voxtral_realtime \
+  --model "$MODEL" \
+  --backend cuda \
+  --threads 8 \
+  --audio input.wav \
+  --text-out transcript.txt
+```
+
+Run the Q4_K checkpoint by changing the model path:
+
+```bash
+MODEL=/path/to/Voxtral-Mini-4B-Realtime-2602-GGUF/voxtral-mini-4b-realtime-2602-q4_k.gguf
+```
+
+## Streaming ASR
+
+Streaming from an audio file:
+
+```bash
+audiocpp_cli \
+  --task asr \
+  --family voxtral_realtime \
+  --model "$MODEL" \
+  --backend cuda \
+  --threads 8 \
+  --mode streaming \
+  --audio input.wav \
+  --text-out transcript.txt
+```
+
+Streaming raw 16 kHz mono PCM from `ffmpeg`:
+
+```bash
+ffmpeg -i input.mp3 -ar 16000 -ac 1 -f s16le - \
+  | audiocpp_cli \
+      --task asr \
+      --family voxtral_realtime \
+      --model "$MODEL" \
+      --backend cuda \
+      --threads 8 \
+      --mode streaming \
+      --audio -
+```
+
+For better streaming throughput, batch a few decode steps:
+
+```bash
+audiocpp_cli \
+  --task asr \
+  --family voxtral_realtime \
+  --model "$MODEL" \
+  --backend cuda \
+  --threads 8 \
+  --mode streaming \
+  --audio input.wav \
+  --session-option voxtral_realtime.stream_batch_tokens=4
+```
+
+`stream_batch_tokens=4` improves throughput by amortizing encoder work, with up to roughly `4 * 80 ms` of additional buffering delay.
+
+## Prompting and decoding
+
+audio.cpp exposes Voxtral Realtime as an ASR model. Pass audio to the CLI and audio.cpp builds the model's transcription prompt internally; no free-form text prompt is required for normal transcription.
+
+Decode options can be controlled from the request:
+
+```bash
+audiocpp_cli \
+  --task asr \
+  --family voxtral_realtime \
+  --model "$MODEL" \
+  --backend cuda \
+  --threads 8 \
+  --audio input.wav \
+  --text-out transcript.txt \
+  --request-option max_new_tokens=256 \
+  --do-sample false \
+  --temperature 1.0 \
+  --top-p 1.0 \
+  --top-k 50 \
+  --seed 1234
+```
+
+## Notes
+
+- Task: `asr`
+- Family: `voxtral_realtime`
+- Supported modes: offline and streaming
+- Timestamp output is not currently exposed by audio.cpp for this model
+- The GGUF package is intended to be standalone: model weights, sidecars, and audio.cpp model spec are embedded in the checkpoint