{ "schema_version": "2", "name": "Maya 2 Native — speak & clone", "references": [ { "id": "ref_text", "label": "Text", "role": "reference", "asset_type": "text", "inputs": [{"id": "in", "label": "Text", "type": "text"}], "outputs": [{"id": "out", "label": "Text", "type": "text"}], "x": 40, "y": 40, "width": 320, "height": 140, "data": {"out": "नमस्ते, आपका ऑर्डर कल पहुँच जाएगा।"} }, { "id": "ref_lang", "label": "Language", "role": "reference", "asset_type": "text", "inputs": [{"id": "in", "label": "Text", "type": "text"}], "outputs": [{"id": "out", "label": "Text", "type": "text"}], "x": 40, "y": 210, "width": 150, "height": 90, "data": {"out": "hi"} }, { "id": "ref_voice", "label": "Voice", "role": "reference", "asset_type": "text", "inputs": [{"id": "in", "label": "Text", "type": "text"}], "outputs": [{"id": "out", "label": "Text", "type": "text"}], "x": 210, "y": 210, "width": 150, "height": 90, "data": {"out": "Ritvik"} }, { "id": "ref_seed", "label": "Seed (0 = random)", "role": "reference", "asset_type": "text", "inputs": [{"id": "in", "label": "Text", "type": "text"}], "outputs": [{"id": "out", "label": "Text", "type": "text"}], "x": 40, "y": 320, "width": 150, "height": 90, "data": {"out": "0"} }, { "id": "ref_clip", "label": "Reference Clip (3–10 s)", "role": "reference", "asset_type": "audio", "inputs": [{"id": "in", "label": "Audio", "type": "audio"}], "outputs": [{"id": "out", "label": "Audio", "type": "audio"}], "x": 40, "y": 440, "width": 240, "height": 120, "data": {} }, { "id": "ref_transcript", "label": "Clip Transcript", "role": "reference", "asset_type": "text", "inputs": [{"id": "in", "label": "Text", "type": "text"}], "outputs": [{"id": "out", "label": "Text", "type": "text"}], "x": 300, "y": 440, "width": 260, "height": 120, "data": {} }, { "id": "ref_reflang", "label": "Clip Language", "role": "reference", "asset_type": "text", "inputs": [{"id": "in", "label": "Text", "type": "text"}], "outputs": [{"id": "out", "label": "Text", "type": "text"}], "x": 40, "y": 590, "width": 150, "height": 90, "data": {"out": "hi"} }, { "id": "ref_clonelang", "label": "Speak As (language)", "role": "reference", "asset_type": "text", "inputs": [{"id": "in", "label": "Text", "type": "text"}], "outputs": [{"id": "out", "label": "Text", "type": "text"}], "x": 210, "y": 590, "width": 150, "height": 90, "data": {"out": "ta"} } ], "operators": [ { "id": "op_speak", "label": "Maya Speak", "role": "operator", "kind": "fn", "fn": "maya_tts", "inputs": [ {"id": "text", "label": "Text", "type": "text", "required": true}, {"id": "lang", "label": "Language", "type": "text"}, {"id": "voice", "label": "Voice", "type": "text"}, {"id": "seed", "label": "Seed", "type": "text"} ], "outputs": [{"id": "out_0", "label": "Speech", "type": "audio", "output_index": 0}], "x": 440, "y": 100, "width": 230, "height": 150, "data": {} }, { "id": "op_clone", "label": "Maya Clone", "role": "operator", "kind": "fn", "fn": "maya_clone", "inputs": [ {"id": "text", "label": "Text", "type": "text", "required": true}, {"id": "ref_audio", "label": "Reference Clip", "type": "audio", "required": true}, {"id": "ref_text", "label": "Transcript", "type": "text", "required": true}, {"id": "ref_lang", "label": "Clip Language", "type": "text"}, {"id": "lang", "label": "Speak As", "type": "text"}, {"id": "seed", "label": "Seed", "type": "text"} ], "outputs": [{"id": "out_0", "label": "Cloned Speech", "type": "audio", "output_index": 0}], "x": 620, "y": 420, "width": 240, "height": 200, "data": {} } ], "subjects": [ { "id": "sub_speech", "label": "Speech", "role": "subject", "asset_type": "audio", "inputs": [{"id": "in", "label": "Audio", "type": "audio"}], "outputs": [{"id": "out", "label": "Audio", "type": "audio"}], "x": 740, "y": 100, "width": 230, "height": 110, "data": {} }, { "id": "sub_clone", "label": "Cloned Speech", "role": "subject", "asset_type": "audio", "inputs": [{"id": "in", "label": "Audio", "type": "audio"}], "outputs": [{"id": "out", "label": "Audio", "type": "audio"}], "x": 920, "y": 420, "width": 230, "height": 110, "data": {} } ], "edges": [ {"id": "e_text", "from_node_id": "ref_text", "from_port_id": "out", "to_node_id": "op_speak", "to_port_id": "text", "type": "text"}, {"id": "e_lang", "from_node_id": "ref_lang", "from_port_id": "out", "to_node_id": "op_speak", "to_port_id": "lang", "type": "text"}, {"id": "e_voice", "from_node_id": "ref_voice", "from_port_id": "out", "to_node_id": "op_speak", "to_port_id": "voice", "type": "text"}, {"id": "e_seed", "from_node_id": "ref_seed", "from_port_id": "out", "to_node_id": "op_speak", "to_port_id": "seed", "type": "text"}, {"id": "e_out", "from_node_id": "op_speak", "from_port_id": "out_0", "to_node_id": "sub_speech", "to_port_id": "in", "type": "audio"}, {"id": "e_text_2", "from_node_id": "ref_text", "from_port_id": "out", "to_node_id": "op_clone", "to_port_id": "text", "type": "text"}, {"id": "e_clip", "from_node_id": "ref_clip", "from_port_id": "out", "to_node_id": "op_clone", "to_port_id": "ref_audio", "type": "audio"}, {"id": "e_transcript", "from_node_id": "ref_transcript", "from_port_id": "out", "to_node_id": "op_clone", "to_port_id": "ref_text", "type": "text"}, {"id": "e_reflang", "from_node_id": "ref_reflang", "from_port_id": "out", "to_node_id": "op_clone", "to_port_id": "ref_lang", "type": "text"}, {"id": "e_clonelang", "from_node_id": "ref_clonelang", "from_port_id": "out", "to_node_id": "op_clone", "to_port_id": "lang", "type": "text"}, {"id": "e_seed_2", "from_node_id": "ref_seed", "from_port_id": "out", "to_node_id": "op_clone", "to_port_id": "seed", "type": "text"}, {"id": "e_out_2", "from_node_id": "op_clone", "from_port_id": "out_0", "to_node_id": "sub_clone", "to_port_id": "in", "type": "audio"} ] }