{ "hf_id": "OpenMOSS-Team/MOSS-Transcribe-Diarize", "meta": { "title": "MOSS-Transcribe-Diarize", "slug": "moss-transcribe-diarize", "provider": "OpenMOSS", "description": "OpenMOSS's 0.9B end-to-end multi-speaker long-audio transcription model with timestamps and speaker labels, served through vLLM's OpenAI-compatible /v1/audio/transcriptions API.", "date_added": "2026-07-08", "date_updated": "2026-08-10", "tasks": [ "multimodal" ], "performance_headline": "0.9B end-to-end transcription with speaker labels and timestamps", "related_recipes": [ "OpenMOSS-Team/MOSS-TTS" ] }, "recommended_command": { "hardware": "h200", "strategy": "single_node_tp", "variant": "default", "node_count": 1, "deploy_type": "single_node", "env": {}, "docker_image": "vllm/vllm-openai:latest", "command": "vllm serve OpenMOSS-Team/MOSS-Transcribe-Diarize \\\n --trust-remote-code \\\n --tensor-parallel-size 1", "argv": [ "vllm", "serve", "OpenMOSS-Team/MOSS-Transcribe-Diarize", "--trust-remote-code", "--tensor-parallel-size", "1" ], "docker_command": "docker run --gpus all \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n vllm/vllm-openai:latest OpenMOSS-Team/MOSS-Transcribe-Diarize \\\n --trust-remote-code \\\n --tensor-parallel-size 1", "docker_argv": [ "docker", "run", "--gpus", "all", "--privileged", "--ipc=host", "-p", "8000:8000", "-v", "~/.cache/huggingface:/root/.cache/huggingface", "vllm/vllm-openai:latest", "OpenMOSS-Team/MOSS-Transcribe-Diarize", "--trust-remote-code", "--tensor-parallel-size", "1" ], "strategy_spec": { "name": "single_node_tp", "deploy_type": "single_node", "display_name": "Tensor Parallel", "short_label": "TP", "orientation": "latency", "description": "Single-node tensor parallel. Splits the model across all local GPUs. TP size is set to the GPU count at deploy time. The simplest multi-GPU strategy — works for all model architectures.\n", "hardware_match": { "min_gpus": 1, "max_gpus": 8, "multi_node": false }, "vllm_args": [], "parallel_flag": "--tensor-parallel-size" }, "hardware_profile": { "brand": "NVIDIA", "generation": "hopper", "compute_arch": "sm90", "display_name": "H200", "description": "NVIDIA H200 SXM 141 GB HBM3e · 8-GPU HGX node", "gpu_count": 8, "vram_gb": 1128, "multi_node": false }, "alternatives": { "kv_store_centralized_mooncake": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/strategies/kv_store_centralized_mooncake.json", "kv_store_distributed_mooncake": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/strategies/kv_store_distributed_mooncake.json" }, "by_hardware": { "h200": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/h200.json", "h100": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/h100.json", "b200": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/b200.json", "gb200": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/gb200.json", "b300": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/b300.json", "gb300": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/gb300.json", "mi300x": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/mi300x.json", "mi325x": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/mi325x.json", "mi350x": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/mi350x.json", "mi355x": "/OpenMOSS-Team/MOSS-Transcribe-Diarize/hw/mi355x.json" } }, "model": { "model_id": "OpenMOSS-Team/MOSS-Transcribe-Diarize", "min_vllm_version": "0.27.0", "architecture": "dense", "parameter_count": "0.9B", "active_parameters": "0.9B", "base_args": [ "--trust-remote-code" ], "base_env": {}, "install": { "pip": { "command": "uv venv\nsource .venv/bin/activate\nuv pip install -U vllm --torch-backend auto" }, "docker": { "command": "docker pull vllm/vllm-openai:latest" }, "extras": [ { "note": "Audio extras are required for loading and preprocessing audio files.", "command": "uv pip install -U \"vllm[audio]\" --torch-backend auto" } ] } }, "features": {}, "opt_in_features": [], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 4, "description": "MOSS-Transcribe-Diarize-0.9B - single-GPU long-form multi-speaker transcription" } }, "hardware_overrides": {}, "guide": "## Overview\n\n[MOSS-Transcribe-Diarize](https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-Diarize)\nis a 0.9B end-to-end speech-to-text model for long-form multi-speaker\ntranscription. It uses a Whisper-style audio encoder and a Qwen3-style causal\ndecoder to generate structured transcripts with text, timestamps, and speaker\nlabels in a single pass.\n\nExample output:\n\n```text\n[0.11][S01] Good morning! [1.03]\n[1.11][S02] Morning, let's start the meeting. [2.42]\n```\n\nKey capabilities:\n\n- End-to-end multi-speaker transcription without a separate ASR, diarization,\n or alignment pipeline.\n- Structured text output with speaker IDs and timestamps.\n- Long-audio transcription for meetings, interviews, podcasts, and customer\n support recordings.\n- Prompt-based hotwords for names, organizations, product names, project\n codes, and domain vocabulary.\n\n## Prerequisites\n\nInstall vLLM with audio dependencies.\n\n```bash\nuv venv\nsource .venv/bin/activate\n\nuv pip install -U \"vllm[audio]\" --torch-backend auto\n```\n\n## Launching with vLLM\n\n```bash\nvllm serve OpenMOSS-Team/MOSS-Transcribe-Diarize \\\n --trust-remote-code \\\n --served-model-name moss-mtd\n```\n\nThe model uses custom Hugging Face processor code, so `--trust-remote-code` is\nrequired.\n\n## Client Usage\n\n### OpenAI Transcription API\n\n```python\nfrom openai import OpenAI\n\nclient = OpenAI(base_url=\"http://localhost:8000/v1\", api_key=\"EMPTY\")\n\nwith open(\"meeting.wav\", \"rb\") as audio_file:\n transcription = client.audio.transcriptions.create(\n model=\"moss-mtd\",\n file=audio_file,\n response_format=\"diarized_json\",\n temperature=0,\n )\n\nprint(transcription.text)\n\nfor segment in transcription.segments:\n print(\n f\"[{segment.start:.2f} - {segment.end:.2f}] \"\n f\"{segment.speaker}: {segment.text}\"\n )\n```\n\n### cURL\n\n```bash\ncurl http://localhost:8000/v1/audio/transcriptions \\\n -F \"model=moss-mtd\" \\\n -F \"file=@meeting.wav\" \\\n -F \"response_format=diarized_json\" \\\n -F \"temperature=0\"\n```\n\n## Output Format\n\nWith `response_format=diarized_json`, vLLM returns an OpenAI-compatible\nresponse with structured speaker segments:\n\n```json\n{\n \"task\": \"transcribe\",\n \"duration\": 5.1,\n \"text\": \"Let's review the deployment plan. The vLLM server is already running.\",\n \"segments\": [\n {\n \"type\": \"transcript.text.segment\",\n \"id\": \"seg_0\",\n \"start\": 0.41,\n \"end\": 2.8,\n \"text\": \"Let's review the deployment plan.\",\n \"speaker\": \"S01\"\n },\n {\n \"type\": \"transcript.text.segment\",\n \"id\": \"seg_1\",\n \"start\": 3.12,\n \"end\": 5.1,\n \"text\": \"The vLLM server is already running.\",\n \"speaker\": \"S02\"\n }\n ],\n \"usage\": {\n \"type\": \"duration\",\n \"seconds\": 6\n }\n}\n```\n\nSpeaker IDs such as `S01`, `S02`, and `S03` are generated by the model and are\nintended to remain consistent within the audio.\n\nUse `response_format=json` if you want the model's native timestamped\ntranscript instead.\n\n```text\n[0.41][S01] Let's review the deployment plan. [2.80]\n[3.12][S02] The vLLM server is already running. [5.10]\n```\n\n## References\n\n- [Model card](https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-Diarize)\n- [GitHub repository](https://github.com/OpenMOSS/MOSS-Transcribe-Diarize)\n- [vLLM audio support](https://docs.vllm.ai/en/latest/features/multimodal_inputs.html)\n", "source": { "source": "vllm", "repo": "vllm-project/vllm", "ref": "main", "sha": "389ffebe555705c85337025d5df108d0384d365e", "path": "recipes/models/OpenMOSS-Team/MOSS-Transcribe-Diarize.yaml", "edit_url": "https://github.com/vllm-project/vllm/edit/main/recipes/models/OpenMOSS-Team/MOSS-Transcribe-Diarize.yaml", "overlays": [] } }