{
  "implementation": "Voxtype Parakeet TDT",
  "model": "parakeet-tdt-0.6b-v3-int8",
  "backend": "ROCm/MIGraphX",
  "results": [
    {
      "input": "../test-pack/opening-multispeaker.wav",
      "wall_seconds": 14.470561080961488,
      "rtf": 0.1205880090080124,
      "text": "Now can the band do a line dancing song? Like play a song that people could line dance to? Uh-huh. Um trombone, I could play a Latin song. A Latin song. On what? My trombone. It's a very Latin instrument. Okay. Yeah, I mean if you figure something out of five second clip of it, I think they probably what makes the most sense is just pick pick whatever song the line dancers want to do and try to just nail like uh a repeating measure or a bar of it. I mean like probably there's the one there's like a couple that everyone does at the at like the parties, like there's the the church clap, and there's another there's another, there's some of those different songs that everyone likes to do it. Um it should probably be caught and eye Joe because it's Texas thing like maybe boots on the ground. Um, isn't more what about uh that song he's choosing Texas? Oh, there's a line dance to that, but you don't know it yet. Okay, well there's the line dancing, but then there's Latin dancing. Well, okay, so we're mixing up so line dancing, we're gonna try and have the choir do we're gonna try and have the band do it. Right? Um now do we need to do choir plus band? That might be too much. Tim, what if we just had like the the band as like an accent since we're gonna have a full band and we could play that actual track of us in Texas?",
      "stdout": "Loading audio file: \"../test-pack/opening-multispeaker.wav\"\nAudio format: 16000 Hz, 1 channel(s), Int\nProcessing 1920000 samples (120.00s)...\n2026-09-26T21:23:14.305601Z  WARN Could not auto-detect model type from \"<local-path>\", defaulting to TDT. Set model_type in config to override.\n2026-09-26T21:23:14.305610Z  INFO Loading Parakeet Tdt model from \"<local-path>\"\n2026-09-26T21:23:14.305612Z  INFO Configuring MIGraphX execution provider for AMD GPU acceleration\n2026-09-26T21:23:16.445677Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:23:18.407134Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:23:18.438962Z  INFO Parakeet Tdt model loaded in 4.13s\n2026-09-26T21:23:28.623098Z  INFO Parakeet Tdt transcription completed in 10.18s: \"Now can the band do a line dancing song? Like play...\"\n\nNow can the band do a line dancing song? Like play a song that people could line dance to? Uh-huh. Um trombone, I could play a Latin song. A Latin song. On what? My trombone. It's a very Latin instrument. Okay. Yeah, I mean if you figure something out of five second clip of it, I think they probably what makes the most sense is just pick pick whatever song the line dancers want to do and try to just nail like uh a repeating measure or a bar of it. I mean like probably there's the one there's like a couple that everyone does at the at like the parties, like there's the the church clap, and there's another there's another, there's some of those different songs that everyone likes to do it. Um it should probably be caught and eye Joe because it's Texas thing like maybe boots on the ground. Um, isn't more what about uh that song he's choosing Texas? Oh, there's a line dance to that, but you don't know it yet. Okay, well there's the line dancing, but then there's Latin dancing. Well, okay, so we're mixing up so line dancing, we're gonna try and have the choir do we're gonna try and have the band do it. Right? Um now do we need to do choir plus band? That might be too much. Tim, what if we just had like the the band as like an accent since we're gonna have a full band and we could play that actual track of us in Texas?\n",
      "stderr": "",
      "vram_baseline_bytes": 649109504,
      "vram_peak_bytes": 857559040,
      "vram_delta_bytes": 208449536,
      "vram_samples": 83
    },
    {
      "input": "../test-pack/mid-meeting-turns.wav",
      "wall_seconds": 13.495365774957463,
      "rtf": 0.11246138145797886,
      "text": "Band everybody. Straight into but the band is playing a Christian song now. So the right Great Among the Nations. Whatever it is, okay. So it's like and then straight into all hands singing together, arm in arm. Kind of like you know if you've ever been to a Broadway show when all the actors come out together. Yeah. You know what I mean? And they're all arm in arm in their different costumes, and they're singing together, and the whole crowd is singing. And then and then it's just Mia Bringing us in for a landing. And then that's it. How does everybody feel? Comments, questions, queries, attitudes, frustrations. We're we're we have very short notice, but I think this can be really awesome and fun. Um we all just gotta help each other. Everybody feel okay. I think it's great. Um let's raise my hand to join the choir and I'll be down to learn the blind ends. Okay, thank you, my wife. Yep, sure. Thank you, sis um anyone else. I just want to make sure I have the order correct. So it's starting with the solo, watch your click this link. Okay, got it, did not see that. Okay, perfect, got it, thank you. What about the baptistry voice? What's up, Dylan? So so Dylan and I sing was like doing uh you guys know the backstreet boys back street boys.",
      "stdout": "Loading audio file: \"../test-pack/mid-meeting-turns.wav\"\nAudio format: 16000 Hz, 1 channel(s), Int\nProcessing 1920000 samples (120.00s)...\n2026-09-26T21:23:28.922046Z  WARN Could not auto-detect model type from \"<local-path>\", defaulting to TDT. Set model_type in config to override.\n2026-09-26T21:23:28.922055Z  INFO Loading Parakeet Tdt model from \"<local-path>\"\n2026-09-26T21:23:28.922057Z  INFO Configuring MIGraphX execution provider for AMD GPU acceleration\n2026-09-26T21:23:30.094336Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:23:32.049370Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:23:32.073489Z  INFO Parakeet Tdt model loaded in 3.15s\n2026-09-26T21:23:42.280371Z  INFO Parakeet Tdt transcription completed in 10.21s: \"Band everybody. Straight into but the band is play...\"\n\nBand everybody. Straight into but the band is playing a Christian song now. So the right Great Among the Nations. Whatever it is, okay. So it's like and then straight into all hands singing together, arm in arm. Kind of like you know if you've ever been to a Broadway show when all the actors come out together. Yeah. You know what I mean? And they're all arm in arm in their different costumes, and they're singing together, and the whole crowd is singing. And then and then it's just Mia Bringing us in for a landing. And then that's it. How does everybody feel? Comments, questions, queries, attitudes, frustrations. We're we're we have very short notice, but I think this can be really awesome and fun. Um we all just gotta help each other. Everybody feel okay. I think it's great. Um let's raise my hand to join the choir and I'll be down to learn the blind ends. Okay, thank you, my wife. Yep, sure. Thank you, sis um anyone else. I just want to make sure I have the order correct. So it's starting with the solo, watch your click this link. Okay, got it, did not see that. Okay, perfect, got it, thank you. What about the baptistry voice? What's up, Dylan? So so Dylan and I sing was like doing uh you guys know the backstreet boys back street boys.\n",
      "stderr": "",
      "vram_baseline_bytes": 649109504,
      "vram_peak_bytes": 857559040,
      "vram_delta_bytes": 208449536,
      "vram_samples": 80
    },
    {
      "input": "../test-pack/signoff.wav",
      "wall_seconds": 17.06408589601051,
      "rtf": 0.14220071580008758,
      "text": "Then that gives you guys a pool to kind of work with that. And then we're off to the races. Yes. Okay. Thank you. It's an honor. All right. Let's uh close up. I'm going to pray for a second. Dear God, thank you for all these wonderful disciples here, God, who've given their heart to you, God, and really want to put together a great presentation. I pray that you are with each of them, God, as they go out and inspire and encourage and talk to people about their various different um parts of the presentation. And that it's a really amazing encouragement for everybody. We love you. We pray this name of Son, Jesus Christ. Love y'all. Bye. Bye, guys.",
      "stdout": "Loading audio file: \"../test-pack/signoff.wav\"\nAudio format: 16000 Hz, 1 channel(s), Int\nProcessing 1920000 samples (120.00s)...\n2026-09-26T21:23:42.652330Z  WARN Could not auto-detect model type from \"<local-path>\", defaulting to TDT. Set model_type in config to override.\n2026-09-26T21:23:42.652343Z  INFO Loading Parakeet Tdt model from \"<local-path>\"\n2026-09-26T21:23:42.652346Z  INFO Configuring MIGraphX execution provider for AMD GPU acceleration\n2026-09-26T21:23:44.397266Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:23:46.776057Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:23:46.809547Z  INFO Parakeet Tdt model loaded in 4.16s\n2026-09-26T21:23:59.527299Z  INFO Parakeet Tdt transcription completed in 12.72s: \"Then that gives you guys a pool to kind of work wi...\"\n\nThen that gives you guys a pool to kind of work with that. And then we're off to the races. Yes. Okay. Thank you. It's an honor. All right. Let's uh close up. I'm going to pray for a second. Dear God, thank you for all these wonderful disciples here, God, who've given their heart to you, God, and really want to put together a great presentation. I pray that you are with each of them, God, as they go out and inspire and encourage and talk to people about their various different um parts of the presentation. And that it's a really amazing encouragement for everybody. We love you. We pray this name of Son, Jesus Christ. Love y'all. Bye. Bye, guys.\n",
      "stderr": "",
      "vram_baseline_bytes": 649109504,
      "vram_peak_bytes": 857563136,
      "vram_delta_bytes": 208453632,
      "vram_samples": 79
    },
    {
      "input": "../test-pack/pretrim-tail.wav",
      "wall_seconds": 11.795096833026037,
      "rtf": 0.14009759636338415,
      "text": "",
      "stdout": "Loading audio file: \"../test-pack/pretrim-tail.wav\"\nAudio format: 16000 Hz, 1 channel(s), Int\nProcessing 1347072 samples (84.19s)...\n2026-09-26T21:23:59.926328Z  WARN Could not auto-detect model type from \"<local-path>\", defaulting to TDT. Set model_type in config to override.\n2026-09-26T21:23:59.926350Z  INFO Loading Parakeet Tdt model from \"<local-path>\"\n2026-09-26T21:23:59.926352Z  INFO Configuring MIGraphX execution provider for AMD GPU acceleration\n2026-09-26T21:24:01.638887Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:24:04.108269Z  WARN [MIGraphX EP] MIGraphX ENV Override Variables Set:\n2026-09-26T21:24:04.155438Z  INFO Parakeet Tdt model loaded in 4.23s\n2026-09-26T21:24:11.585197Z  INFO Parakeet Tdt transcription completed in 7.43s: \"\"\n\n\n",
      "stderr": "",
      "vram_baseline_bytes": 649109504,
      "vram_peak_bytes": 649183232,
      "vram_delta_bytes": 73728,
      "vram_samples": 54
    }
  ]
}
