{ "product": "Spum Intelligence", "capability": "Spum Voice", "date": "2026-11-02", "summary": [ { "variant": "Kokoro upstream FP32", "hardware": "Apple M5 Max \u00b7 CPU", "samples": 6, "median_rtf": 0.08794767948288963, "median_wall_seconds": 0.6787347915087594, "english_asr_wer": 0.0, "english_asr_samples": 6, "median_first_pcm_chunk_seconds": null }, { "variant": "Spum Voice Kokoro INT8", "hardware": "Apple M5 Max \u00b7 CPU", "samples": 6, "median_rtf": 0.6371333182856251, "median_wall_seconds": 4.19384008400084, "english_asr_wer": 0.0, "english_asr_samples": 6, "median_first_pcm_chunk_seconds": null }, { "variant": "eSpeak NG reference", "hardware": "Apple M5 Max \u00b7 CPU", "samples": 6, "median_rtf": 0.0017264593960506299, "median_wall_seconds": 0.010881438007345423, "english_asr_wer": 0.07272727272727272, "english_asr_samples": 6, "median_first_pcm_chunk_seconds": null }, { "variant": "maya-default", "hardware": "H200 DGX \u00b7 GPU 0", "samples": 10, "median_rtf": 0.1347347155007275, "median_wall_seconds": 0.843412480520783, "english_asr_wer": 0.0, "english_asr_samples": 6, "median_first_pcm_chunk_seconds": 0.01462548301788047 }, { "variant": "maya-matched", "hardware": "H200 DGX \u00b7 GPU 0", "samples": 10, "median_rtf": 0.13512920872017276, "median_wall_seconds": 0.7580617444764357, "english_asr_wer": 0.0, "english_asr_samples": 6, "median_first_pcm_chunk_seconds": 0.014686818030895665 }, { "variant": "spum-voice", "hardware": "H200 DGX \u00b7 GPU 0", "samples": 10, "median_rtf": 0.13511768123218576, "median_wall_seconds": 0.7726958449929953, "english_asr_wer": 0.0, "english_asr_samples": 6, "median_first_pcm_chunk_seconds": 0.014717849961016327 } ], "runs": [ { "name": "Spum Voice / Maya comparison", "hardware": "NVIDIA H200", "gpu_index_requested": 0, "load_seconds": 31.23677752900403, "warmup_excluded": true, "upstream_commit": "c6b376868e268551d065c4206193e96257bef4a1", "scope": "same weights and upstream inference implementation; Spum adds a pronunciation/prosody profile, not a new model or native kernel", "rows": [ { "prompt_id": "en-maths", "language": "en", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "spoken_input": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "variant": "maya-default", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-en-maths-0-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 6.64, "wall_seconds": 0.894308848946821, "time_to_first_chunk_seconds": 0.02123777399538085, "rtf": 0.13468506761247304, "peak_gpu_memory_bytes": 7357096960, "clipping_fraction": 0.0, "rms": 0.06133734986772281, "sha256": "94111c0dadeca9b704f76da5f46b6d632dcea92366a37c231af3fa0cc500830e", "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.1 }, { "prompt_id": "hi-teach", "language": "hi", "text": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "spoken_input": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "variant": "maya-default", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-hi-teach-0-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 4.32, "wall_seconds": 0.5806478020385839, "time_to_first_chunk_seconds": 0.015321705024689436, "rtf": 0.13440921343485737, "peak_gpu_memory_bytes": 7357127168, "clipping_fraction": 0.0, "rms": 0.0642686176610898, "sha256": "ae41cd5d257585066ca5ba2479e41e9accef799405f4ea83fa004cbecd63d04d", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-device", "language": "en", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "spoken_input": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "variant": "spum-voice", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-device-0-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7874487779918127, "time_to_first_chunk_seconds": 0.04565203300444409, "rtf": 0.1406158532128237, "peak_gpu_memory_bytes": 7357092352, "clipping_fraction": 0.0, "rms": 0.06063270497544476, "sha256": "a696ee8982ecb1c2329752b861805b2610a95d50de058bec1cd997270ed5973f", "asr_text": "The scanner is ready, the image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.0 }, { "prompt_id": "hi-physics", "language": "hi", "text": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "spoken_input": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "variant": "spum-voice", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-physics-1-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7565471669659019, "time_to_first_chunk_seconds": 0.014875630964525044, "rtf": 0.13509770838676818, "peak_gpu_memory_bytes": 7357144576, "clipping_fraction": 0.0, "rms": 0.05477384147474004, "sha256": "7adae9e6fc711f0662aac8d349807921e787c11164215aed03dd4dc5f15c79ad", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-maths", "language": "en", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "spoken_input": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "variant": "spum-voice", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-maths-1-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 7.28, "wall_seconds": 0.9792416309937835, "time_to_first_chunk_seconds": 0.014503373997285962, "rtf": 0.13451121304859662, "peak_gpu_memory_bytes": 7357096960, "clipping_fraction": 0.0, "rms": 0.060182075704026665, "sha256": "36dffc4bcc12e9451250be2e2ba1367170a04a2caa48dd0dc475aae020d0ec45", "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.1 }, { "prompt_id": "hi-physics", "language": "hi", "text": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "spoken_input": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "variant": "maya-default", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-hi-physics-0-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 6.24, "wall_seconds": 0.8421370360301808, "time_to_first_chunk_seconds": 0.015962135046720505, "rtf": 0.13495785833816998, "peak_gpu_memory_bytes": 7357144576, "clipping_fraction": 0.0, "rms": 0.0614957979247253, "sha256": "772a4c0224e6bb40cd7cbbbb1c0db8b2d99d05ff79a2c0b6ca1f32c9c567c36f", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-device", "language": "en", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "spoken_input": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "variant": "maya-default", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-en-device-0-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7547924349782988, "time_to_first_chunk_seconds": 0.014582364994566888, "rtf": 0.13478436338898195, "peak_gpu_memory_bytes": 7357092352, "clipping_fraction": 0.0, "rms": 0.06423835634355203, "sha256": "814997954ac3e5aa8e5862462efcc47774f1a66fb92660b71bfb48d6ad6dc92d", "asr_text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.0 }, { "prompt_id": "en-device", "language": "en", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "spoken_input": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "variant": "maya-default", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-en-device-1-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7552221429650672, "time_to_first_chunk_seconds": 0.014361917972564697, "rtf": 0.1348610969580477, "peak_gpu_memory_bytes": 7357092352, "clipping_fraction": 0.0, "rms": 0.06423835634355203, "sha256": "814997954ac3e5aa8e5862462efcc47774f1a66fb92660b71bfb48d6ad6dc92d", "asr_text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.0 }, { "prompt_id": "en-numbers", "language": "en", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "spoken_input": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "variant": "maya-matched", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-numbers-1-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 7.36, "wall_seconds": 0.9911063259933144, "time_to_first_chunk_seconds": 0.014679873012937605, "rtf": 0.1346611855969177, "peak_gpu_memory_bytes": 7357109760, "clipping_fraction": 0.0, "rms": 0.06153496989956635, "sha256": "d74453c6abc7dc1f1c72dc8effac4fe8372ff300108ac27382ce8274fd6307c0", "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.36363636363636365 }, { "prompt_id": "en-maths", "language": "en", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "spoken_input": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "variant": "maya-default", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-en-maths-1-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 6.64, "wall_seconds": 0.893175879959017, "time_to_first_chunk_seconds": 0.014381504966877401, "rtf": 0.134514439752864, "peak_gpu_memory_bytes": 7357096960, "clipping_fraction": 0.0, "rms": 0.06133734986772281, "sha256": "94111c0dadeca9b704f76da5f46b6d632dcea92366a37c231af3fa0cc500830e", "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.1 }, { "prompt_id": "hi-teach", "language": "hi", "text": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "spoken_input": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "variant": "maya-default", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-hi-teach-1-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 4.32, "wall_seconds": 0.5850119829992764, "time_to_first_chunk_seconds": 0.014545185025781393, "rtf": 0.13541944050909174, "peak_gpu_memory_bytes": 7357127168, "clipping_fraction": 0.0, "rms": 0.0642686176610898, "sha256": "ae41cd5d257585066ca5ba2479e41e9accef799405f4ea83fa004cbecd63d04d", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-device", "language": "en", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "spoken_input": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "variant": "maya-matched", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-device-0-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7547802039771341, "time_to_first_chunk_seconds": 0.014274539018515497, "rtf": 0.13478217928163108, "peak_gpu_memory_bytes": 7357092352, "clipping_fraction": 0.0, "rms": 0.06063270497544476, "sha256": "a696ee8982ecb1c2329752b861805b2610a95d50de058bec1cd997270ed5973f", "asr_text": "The scanner is ready, the image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.0 }, { "prompt_id": "en-numbers", "language": "en", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "spoken_input": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "variant": "maya-matched", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-numbers-0-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 7.36, "wall_seconds": 0.993201853998471, "time_to_first_chunk_seconds": 0.014773635019082576, "rtf": 0.13494590407587922, "peak_gpu_memory_bytes": 7357109760, "clipping_fraction": 0.0, "rms": 0.06153496989956635, "sha256": "d74453c6abc7dc1f1c72dc8effac4fe8372ff300108ac27382ce8274fd6307c0", "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.36363636363636365 }, { "prompt_id": "hi-physics", "language": "hi", "text": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "spoken_input": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "variant": "maya-default", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-hi-physics-1-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 6.24, "wall_seconds": 0.8446879250113852, "time_to_first_chunk_seconds": 0.014969018986448646, "rtf": 0.13536665464926045, "peak_gpu_memory_bytes": 7357144576, "clipping_fraction": 0.0, "rms": 0.0614957979247253, "sha256": "772a4c0224e6bb40cd7cbbbb1c0db8b2d99d05ff79a2c0b6ca1f32c9c567c36f", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "hi-physics", "language": "hi", "text": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "spoken_input": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "variant": "maya-matched", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-physics-1-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7575219329446554, "time_to_first_chunk_seconds": 0.015594356984365731, "rtf": 0.13527177374011704, "peak_gpu_memory_bytes": 7357144576, "clipping_fraction": 0.0, "rms": 0.05477384147474004, "sha256": "7adae9e6fc711f0662aac8d349807921e787c11164215aed03dd4dc5f15c79ad", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-maths", "language": "en", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "spoken_input": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "variant": "spum-voice", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-maths-0-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 7.28, "wall_seconds": 0.979884980013594, "time_to_first_chunk_seconds": 0.014386577007826418, "rtf": 0.13459958516670248, "peak_gpu_memory_bytes": 7357096960, "clipping_fraction": 0.0, "rms": 0.060182075704026665, "sha256": "36dffc4bcc12e9451250be2e2ba1367170a04a2caa48dd0dc475aae020d0ec45", "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.1 }, { "prompt_id": "en-numbers", "language": "en", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "spoken_input": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "variant": "spum-voice", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-numbers-0-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 7.36, "wall_seconds": 0.9917451619985513, "time_to_first_chunk_seconds": 0.014614093990530819, "rtf": 0.13474798396719448, "peak_gpu_memory_bytes": 7357109760, "clipping_fraction": 0.0, "rms": 0.06153496989956635, "sha256": "d74453c6abc7dc1f1c72dc8effac4fe8372ff300108ac27382ce8274fd6307c0", "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.36363636363636365 }, { "prompt_id": "hi-physics", "language": "hi", "text": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "spoken_input": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "variant": "spum-voice", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-physics-0-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7579429119941778, "time_to_first_chunk_seconds": 0.015281464031431824, "rtf": 0.1353469485703889, "peak_gpu_memory_bytes": 7357144576, "clipping_fraction": 0.0, "rms": 0.05477384147474004, "sha256": "7adae9e6fc711f0662aac8d349807921e787c11164215aed03dd4dc5f15c79ad", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-numbers", "language": "en", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "spoken_input": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "variant": "maya-default", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-en-numbers-1-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 7.44, "wall_seconds": 1.0015032780356705, "time_to_first_chunk_seconds": 0.014613181992899626, "rtf": 0.1346106556499557, "peak_gpu_memory_bytes": 7357109760, "clipping_fraction": 0.0, "rms": 0.06041473810015638, "sha256": "101b3bf968e642b6eeb31db8f6d895888a229b3d6c5351d54eab70b182d87dfa", "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.36363636363636365 }, { "prompt_id": "en-numbers", "language": "en", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "spoken_input": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "variant": "maya-default", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": {}, "audio_file": "audio/maya-en-numbers-0-maya-default.wav", "sample_rate_hz": 24000, "audio_seconds": 7.44, "wall_seconds": 1.001332149025984, "time_to_first_chunk_seconds": 0.014637784042861313, "rtf": 0.13458765443897636, "peak_gpu_memory_bytes": 7357109760, "clipping_fraction": 0.0, "rms": 0.06041473810015638, "sha256": "101b3bf968e642b6eeb31db8f6d895888a229b3d6c5351d54eab70b182d87dfa", "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.36363636363636365 }, { "prompt_id": "en-device", "language": "en", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "spoken_input": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "variant": "spum-voice", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-device-1-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7564483630121686, "time_to_first_chunk_seconds": 0.014436300029046834, "rtf": 0.13508006482360155, "peak_gpu_memory_bytes": 7357092352, "clipping_fraction": 0.0, "rms": 0.06063270497544476, "sha256": "a696ee8982ecb1c2329752b861805b2610a95d50de058bec1cd997270ed5973f", "asr_text": "The scanner is ready, the image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.0 }, { "prompt_id": "hi-physics", "language": "hi", "text": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "spoken_input": "\u092c\u0932 \u0938\u0947 \u0935\u0938\u094d\u0924\u0941 \u0915\u093e \u0906\u0917\u0947 \u092c\u0922\u093c\u0928\u093e \u0936\u0941\u0930\u0942 \u0939\u094b\u0924\u093e \u0939\u0948, \u0932\u0947\u0915\u093f\u0928 \u092c\u093f\u0928\u093e \u092b\u093f\u0938\u0932\u0947 \u0918\u0942\u092e\u0928\u0947 \u0915\u0947 \u0932\u093f\u090f \u0938\u094d\u0925\u0948\u0924\u093f\u0915 \u0918\u0930\u094d\u0937\u0923 \u091f\u0949\u0930\u094d\u0915 \u0926\u0947\u0924\u093e \u0939\u0948\u0964", "variant": "maya-matched", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-physics-0-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.7568385499762371, "time_to_first_chunk_seconds": 0.014950198004953563, "rtf": 0.13514974106718522, "peak_gpu_memory_bytes": 7357144576, "clipping_fraction": 0.0, "rms": 0.05477384147474004, "sha256": "7adae9e6fc711f0662aac8d349807921e787c11164215aed03dd4dc5f15c79ad", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-maths", "language": "en", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "spoken_input": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "variant": "maya-matched", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-maths-0-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 7.28, "wall_seconds": 0.9800831940374337, "time_to_first_chunk_seconds": 0.014693763048853725, "rtf": 0.13462681236777935, "peak_gpu_memory_bytes": 7357096960, "clipping_fraction": 0.0, "rms": 0.060182075704026665, "sha256": "36dffc4bcc12e9451250be2e2ba1367170a04a2caa48dd0dc475aae020d0ec45", "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.1 }, { "prompt_id": "en-maths", "language": "en", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "spoken_input": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "variant": "maya-matched", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-maths-1-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 7.28, "wall_seconds": 0.9835911639966071, "time_to_first_chunk_seconds": 0.014504966966342181, "rtf": 0.13510867637316032, "peak_gpu_memory_bytes": 7357096960, "clipping_fraction": 0.0, "rms": 0.060182075704026665, "sha256": "36dffc4bcc12e9451250be2e2ba1367170a04a2caa48dd0dc475aae020d0ec45", "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.1 }, { "prompt_id": "hi-teach", "language": "hi", "text": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "spoken_input": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "variant": "maya-matched", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-teach-0-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 4.72, "wall_seconds": 0.6421188510139473, "time_to_first_chunk_seconds": 0.014835945970844477, "rtf": 0.13604212945210747, "peak_gpu_memory_bytes": 7357127168, "clipping_fraction": 0.0, "rms": 0.06582644787626382, "sha256": "01baf4757d34a023b70b0e3a504b876da17d33a6bbd52041a9a70401ff90c032", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "hi-teach", "language": "hi", "text": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "spoken_input": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "variant": "maya-matched", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-teach-1-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 4.72, "wall_seconds": 0.6418338859803043, "time_to_first_chunk_seconds": 0.014635052008088678, "rtf": 0.13598175550430178, "peak_gpu_memory_bytes": 7357127168, "clipping_fraction": 0.0, "rms": 0.06582644787626382, "sha256": "01baf4757d34a023b70b0e3a504b876da17d33a6bbd52041a9a70401ff90c032", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "hi-teach", "language": "hi", "text": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "spoken_input": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948? ... ", "variant": "spum-voice", "repeat": 0, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-teach-0-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 4.64, "wall_seconds": 0.6305885469773784, "time_to_first_chunk_seconds": 0.014783019956666976, "rtf": 0.13590270408995225, "peak_gpu_memory_bytes": 7357127168, "clipping_fraction": 0.0, "rms": 0.06342990018892473, "sha256": "82bc5d069b4efced9567ee3915392d56a1c17afd9ca50f8567df06aff5a3576c", "hardware_group": "H200 DGX \u00b7 GPU 0" }, { "prompt_id": "en-device", "language": "en", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "spoken_input": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "variant": "maya-matched", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-device-1-maya-matched.wav", "sample_rate_hz": 24000, "audio_seconds": 5.6, "wall_seconds": 0.758601556008216, "time_to_first_chunk_seconds": 0.01461270498111844, "rtf": 0.1354645635728957, "peak_gpu_memory_bytes": 7357092352, "clipping_fraction": 0.0, "rms": 0.06063270497544476, "sha256": "a696ee8982ecb1c2329752b861805b2610a95d50de058bec1cd997270ed5973f", "asr_text": "The scanner is ready, the image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.0 }, { "prompt_id": "en-numbers", "language": "en", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "spoken_input": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "variant": "spum-voice", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-en-numbers-1-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 7.36, "wall_seconds": 0.9946131340111606, "time_to_first_chunk_seconds": 0.014813044050242752, "rtf": 0.13513765407760334, "peak_gpu_memory_bytes": 7357109760, "clipping_fraction": 0.0, "rms": 0.06153496989956635, "sha256": "d74453c6abc7dc1f1c72dc8effac4fe8372ff300108ac27382ce8274fd6307c0", "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "H200 DGX \u00b7 GPU 0", "raw_asr_wer": 0.36363636363636365 }, { "prompt_id": "hi-teach", "language": "hi", "text": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948?", "spoken_input": "\u0905\u0917\u0930 \u092c\u0932 \u0915\u0940 \u0930\u0947\u0916\u093e \u0915\u0947\u0902\u0926\u094d\u0930 \u0938\u0947 \u0917\u0941\u091c\u0930\u0924\u0940 \u0939\u0948, \u0924\u094b \u0909\u0938 \u092c\u0932 \u0915\u093e \u091f\u0949\u0930\u094d\u0915 \u0936\u0942\u0928\u094d\u092f \u0915\u094d\u092f\u094b\u0902 \u0939\u094b\u0924\u093e \u0939\u0948? ... ", "variant": "spum-voice", "repeat": 1, "voice": "Ritvik", "seed": 42, "sampling": { "temperature": 0.78, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.08 }, "audio_file": "audio/maya-hi-teach-1-spum-voice.wav", "sample_rate_hz": 24000, "audio_seconds": 4.64, "wall_seconds": 0.6291178109822795, "time_to_first_chunk_seconds": 0.014652679965365678, "rtf": 0.13558573512549127, "peak_gpu_memory_bytes": 7357127168, "clipping_fraction": 0.0, "rms": 0.06342990018892473, "sha256": "82bc5d069b4efced9567ee3915392d56a1c17afd9ca50f8567df06aff5a3576c", "hardware_group": "H200 DGX \u00b7 GPU 0" } ] }, { "name": "Spum Voice / Kokoro comparison", "hardware": "Apple M5 Max CPU", "threads": 4, "upstream_load_seconds": 6.431118667009287, "important": "Upstream warm persistent FP32 model; Spum INT8 CLI wall time includes a fresh process/model load. This is an end-user deployment comparison, not an identical precision kernel comparison. Different text frontends may produce different pronunciations.", "rows": [ { "variant": "Kokoro upstream FP32", "prompt_id": "device", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "repeat": 0, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-device-0-kokoro-upstream.wav", "audio_seconds": 6.8, "wall_seconds": 0.5980571249965578, "time_to_first_yield_seconds": 0.5980246670078486, "rtf": 0.08794957720537615, "asr_text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.0 }, { "variant": "Spum Voice Kokoro INT8", "prompt_id": "device", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "repeat": 0, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-device-0-spum-voice.wav", "audio_seconds": 5.795333333333334, "wall_seconds": 5.988404249976156, "rtf": 1.0333148941636068, "native_metrics": { "audio_seconds": 5.795333333333334, "inference_seconds": 3.086293459, "model": "kokoro-en-v1-int8", "model_load_seconds": 0.681584917, "real_time_factor": 0.5325480488323939, "sample_rate": 24000, "total_seconds": 3.768747333, "voice": "af_heart" }, "asr_text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.0 }, { "variant": "Kokoro upstream FP32", "prompt_id": "numbers", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "repeat": 0, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-numbers-0-kokoro-upstream.wav", "audio_seconds": 8.7, "wall_seconds": 0.768033958011074, "time_to_first_yield_seconds": 0.7680183330085129, "rtf": 0.0882797652886292, "asr_text": "The exposure is 80 milliseconds. The signal-to-noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.36363636363636365 }, { "variant": "Spum Voice Kokoro INT8", "prompt_id": "numbers", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "repeat": 0, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-numbers-0-spum-voice.wav", "audio_seconds": 7.089125, "wall_seconds": 4.587878083024407, "rtf": 0.6471712775588534, "native_metrics": { "audio_seconds": 7.089125, "inference_seconds": 3.665386208, "model": "kokoro-en-v1-int8", "model_load_seconds": 0.844721208, "real_time_factor": 0.5170435290674096, "sample_rate": 24000, "total_seconds": 4.510625333, "voice": "af_heart" }, "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.36363636363636365 }, { "variant": "Kokoro upstream FP32", "prompt_id": "maths", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "repeat": 0, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-maths-0-kokoro-upstream.wav", "audio_seconds": 7.675, "wall_seconds": 0.6824857080064248, "time_to_first_yield_seconds": 0.6824751660169568, "rtf": 0.08892321928422474, "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.1 }, { "variant": "Spum Voice Kokoro INT8", "prompt_id": "maths", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "repeat": 0, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-maths-0-spum-voice.wav", "audio_seconds": 6.453375, "wall_seconds": 4.031706000008853, "rtf": 0.624743796851857, "native_metrics": { "audio_seconds": 6.453375, "inference_seconds": 3.307278084, "model": "kokoro-en-v1-int8", "model_load_seconds": 0.650743792, "real_time_factor": 0.5124881296995758, "sample_rate": 24000, "total_seconds": 3.958534875, "voice": "af_heart" }, "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.1 }, { "variant": "Kokoro upstream FP32", "prompt_id": "device", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "repeat": 1, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-device-1-kokoro-upstream.wav", "audio_seconds": 6.8, "wall_seconds": 0.5945442910015117, "time_to_first_yield_seconds": 0.5945347080123611, "rtf": 0.08743298397081055, "asr_text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.0 }, { "variant": "Spum Voice Kokoro INT8", "prompt_id": "device", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "repeat": 1, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-device-1-spum-voice.wav", "audio_seconds": 5.795791666666666, "wall_seconds": 3.822870875010267, "rtf": 0.6595942530158119, "native_metrics": { "audio_seconds": 5.795791666666666, "inference_seconds": 3.094330083, "model": "kokoro-en-v1-int8", "model_load_seconds": 0.656430416, "real_time_factor": 0.5338925656690559, "sample_rate": 24000, "total_seconds": 3.751231583, "voice": "af_heart" }, "asr_text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.0 }, { "variant": "Kokoro upstream FP32", "prompt_id": "numbers", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "repeat": 1, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-numbers-1-kokoro-upstream.wav", "audio_seconds": 8.7, "wall_seconds": 0.7537456669961102, "time_to_first_yield_seconds": 0.7537356249813456, "rtf": 0.08663743298805865, "asr_text": "The exposure is 80 milliseconds. The signal-to-noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.36363636363636365 }, { "variant": "Spum Voice Kokoro INT8", "prompt_id": "numbers", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "repeat": 1, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-numbers-1-spum-voice.wav", "audio_seconds": 7.08975, "wall_seconds": 4.340328334015794, "rtf": 0.6121976563370773, "native_metrics": { "audio_seconds": 7.08975, "inference_seconds": 3.617576416, "model": "kokoro-en-v1-int8", "model_load_seconds": 0.651289084, "real_time_factor": 0.5102544400014104, "sample_rate": 24000, "total_seconds": 4.269452042, "voice": "af_heart" }, "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.36363636363636365 }, { "variant": "Kokoro upstream FP32", "prompt_id": "maths", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "repeat": 1, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-maths-1-kokoro-upstream.wav", "audio_seconds": 7.675, "wall_seconds": 0.6749838750110939, "time_to_first_yield_seconds": 0.6749732090102043, "rtf": 0.08794578176040312, "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.1 }, { "variant": "Spum Voice Kokoro INT8", "prompt_id": "maths", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "repeat": 1, "voice": "af_heart", "threads": 4, "audio_file": "audio/kokoro-maths-1-spum-voice.wav", "audio_seconds": 6.454125, "wall_seconds": 4.047351833985886, "rtf": 0.6270953590123968, "native_metrics": { "audio_seconds": 6.454125, "inference_seconds": 3.322892833, "model": "kokoro-en-v1-int8", "model_load_seconds": 0.649903208, "real_time_factor": 0.5148479202060697, "sample_rate": 24000, "total_seconds": 3.973306792, "voice": "af_heart" }, "asr_text": "The integral from 0 to infinity of exponential minus x squared is the square root of pi divided by 2.", "reference_word_count": 20, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.1 }, { "variant": "eSpeak NG reference", "prompt_id": "device", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "repeat": 0, "audio_file": "audio/kokoro-device-0-espeak.wav", "audio_seconds": 6.311927437641724, "wall_seconds": 0.23423554099281318, "rtf": 0.03710998634045273, "asr_text": "A scanner is ready. The image has been saved. Please remove the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 2, "asr_wer": 0.1111111111111111, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.1111111111111111 }, { "variant": "eSpeak NG reference", "prompt_id": "numbers", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "repeat": 0, "audio_file": "audio/kokoro-numbers-0-espeak.wav", "audio_seconds": 7.794557823129252, "wall_seconds": 0.013755125022726133, "rtf": 0.001764708830808816, "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.36363636363636365 }, { "variant": "eSpeak NG reference", "prompt_id": "maths", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "repeat": 0, "audio_file": "audio/kokoro-maths-0-espeak.wav", "audio_seconds": 6.2687074829931975, "wall_seconds": 0.011045792023651302, "rtf": 0.0017620525528776357, "asr_text": "The integral from 0 to infinity of exponential minus x squared into square root of pi divided by 2", "reference_word_count": 20, "asr_word_edits": 2, "asr_wer": 0.1, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.2 }, { "variant": "eSpeak NG reference", "prompt_id": "device", "text": "The scanner is ready. The image has been saved. Please review the result before moving the optical stage.", "repeat": 1, "audio_file": "audio/kokoro-device-1-espeak.wav", "audio_seconds": 6.311927437641724, "wall_seconds": 0.010672625008737668, "rtf": 0.0016908662392236243, "asr_text": "A scanner is ready. The image has been saved. Please remove the result before moving the optical stage.", "reference_word_count": 18, "asr_word_edits": 2, "asr_wer": 0.1111111111111111, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.1111111111111111 }, { "variant": "eSpeak NG reference", "prompt_id": "numbers", "text": "The exposure is eighty milliseconds. The signal to noise ratio is seventy one point one, at twelve point five frames per second.", "repeat": 1, "audio_file": "audio/kokoro-numbers-1-espeak.wav", "audio_seconds": 7.794557823129252, "wall_seconds": 0.010717083991039544, "rtf": 0.0013749444463979864, "asr_text": "The exposure is 80 milliseconds. The signal to noise ratio is 71.1 at 12.5 frames per second.", "reference_word_count": 17, "asr_word_edits": 0, "asr_wer": 0.0, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.36363636363636365 }, { "variant": "eSpeak NG reference", "prompt_id": "maths", "text": "The integral from zero to infinity of exponential minus x squared is the square root of pi, divided by two.", "repeat": 1, "audio_file": "audio/kokoro-maths-1-espeak.wav", "audio_seconds": 6.2687074829931975, "wall_seconds": 0.010468708002008498, "rtf": 0.0016699946568586534, "asr_text": "The integral from 0 to infinity of exponential minus x squared into square root of pi divided by 2", "reference_word_count": 20, "asr_word_edits": 2, "asr_wer": 0.1, "hardware_group": "Apple M5 Max \u00b7 CPU", "raw_asr_wer": 0.2 } ] } ], "quality_method": "Whisper base.en automatic transcription, normalized using the upstream Whisper English normalizer. ASR WER is a proxy, not human-rated naturalness or medical correctness. Hindi samples require listening review.", "claims": "Spum uses upstream Maya and Kokoro weights. This benchmark does not establish faster Maya kernels, a new voice model, superior naturalness, or superiority across models.", "sources": { "Maya": "https://huggingface.co/maya-research/maya2-native", "Kokoro": "https://huggingface.co/hexgrad/Kokoro-82M", "eSpeak NG": "https://github.com/espeak-ng/espeak-ng", "Chatterbox": "https://github.com/resemble-ai/chatterbox", "Qwen3-TTS": "https://github.com/QwenLM/Qwen3-TTS" } }