{ "schema_version": 1, "family": "sam_audio", "display_name": "SAM Audio", "description": "Prompt-conditioned audio separation.", "category": "audio_tools", "status": "supported", "tasks": [ "s2s" ], "modes": [ "offline" ], "languages": [ "language_agnostic" ], "capabilities": {}, "runtime": { "tags": [ "gguf" ] }, "options": { "request": [ { "name": "reference_video_file", "type": "path", "description": "Masked video identifying the target source; frames are matched to audio by nearest timestamp. Requires libav runtime libraries. Mutually exclusive with reference_image_file.", "required": false }, { "name": "reference_image_file", "type": "path", "description": "Masked RGB image identifying the target sound source, used as a constant visual prompt throughout the recording. Mask unwanted regions before supplying the image.", "required": false }, { "name": "anchors", "type": "string", "description": "Optional JSON array of time spans: [[\"+\", start_seconds, end_seconds], [\"-\", start_seconds, end_seconds]]. Positive spans identify the target; negative spans identify sounds to exclude.", "required": false }, { "name": "seed", "type": "int", "description": "Random seed; -1 selects a random seed.", "min": -1, "max": 4294967295, "default": 42, "required": false }, { "name": "num_inference_steps", "type": "int", "description": "Midpoint integration steps; each step evaluates the denoiser twice.", "min": 1, "default": 16, "required": false } ], "session": [ { "name": "memory_bounded", "type": "bool", "description": "Experimental bounded-workspace execution with full recording context. Provisional option name.", "default": false, "required": false } ], "load": [] }, "package_defaults": { "download": { "kind": "huggingface_snapshot", "repo": "audio-cpp/SAM-Audio-GGUF", "revision": "main", "gated": false } }, "packages": [ { "id": "sam_audio_small_f32", "display_name": "SAM Audio Small F32 GGUF", "default": true, "format": "gguf", "precision": "f32", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-small-f32.gguf" ] }, { "id": "sam_audio_small_bf16", "display_name": "SAM Audio Small BF16 GGUF", "format": "gguf", "precision": "bf16", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-small-bf16.gguf" ] }, { "id": "sam_audio_small_q8_0", "display_name": "SAM Audio Small Q8_0 GGUF", "format": "gguf", "precision": "q8_0", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-small-q8_0.gguf" ] }, { "id": "sam_audio_base_f32", "display_name": "SAM Audio Base F32 GGUF", "format": "gguf", "precision": "f32", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-base-f32.gguf" ] }, { "id": "sam_audio_base_bf16", "display_name": "SAM Audio Base BF16 GGUF", "format": "gguf", "precision": "bf16", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-base-bf16.gguf" ] }, { "id": "sam_audio_base_q8_0", "display_name": "SAM Audio Base Q8_0 GGUF", "format": "gguf", "precision": "q8_0", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-base-q8_0.gguf" ] }, { "id": "sam_audio_large_f32", "display_name": "SAM Audio Large F32 GGUF", "format": "gguf", "precision": "f32", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-large-f32.gguf" ] }, { "id": "sam_audio_large_bf16", "display_name": "SAM Audio Large BF16 GGUF", "format": "gguf", "precision": "bf16", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-large-bf16.gguf" ] }, { "id": "sam_audio_large_q8_0", "display_name": "SAM Audio Large Q8_0 GGUF", "format": "gguf", "precision": "q8_0", "target_directory": "SAM-Audio-GGUF", "files": [ "sam-audio-large-q8_0.gguf" ] } ], "dependencies": [], "ui": { "recommended_package": "sam_audio_small_f32", "tags": [ "GGUF" ], "docs": [ "docs/models/sam_audio.md", "docs/audio_tools.md", "docs/gguf.md" ] }, "sources": [ { "format": "gguf", "roots": { "model": ".", "weights": "$gguf" }, "files": { "config": "model:config.json", "t5_config": "model:t5_config.json", "tokenizer": "model:tokenizer/spiece.model" }, "tensors": { "weights": { "source": "weights:" } } } ] }