feat(plugins): add Parakeet TDT speech-to-text plugin #281
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,144 @@ | ||
| --- | ||
| # SPDX-FileCopyrightText: © 2025 StreamKit Contributors | ||
| # SPDX-License-Identifier: MPL-2.0 | ||
| title: "plugin::native::parakeet" | ||
| description: "Fast speech-to-text transcription using NVIDIA Parakeet TDT, a transducer-based ASR model. Approximately 10x faster than Whisper on consumer hardware with competitive accuracy. Uses sherpa-onnx for inference. Requires 16kHz mono audio input." | ||
| --- | ||
|
|
||
| `kind`: `plugin::native::parakeet` (original kind: `parakeet`) | ||
|
|
||
| Fast speech-to-text transcription using NVIDIA Parakeet TDT, a transducer-based ASR model. Approximately 10x faster than Whisper on consumer hardware with competitive accuracy. Uses sherpa-onnx for inference. Requires 16kHz mono audio input. | ||
|
|
||
| Source: `target/plugins/release/libparakeet.so` | ||
|
|
||
| ## Categories | ||
| - `ml` | ||
| - `speech` | ||
| - `transcription` | ||
|
|
||
| ## Pins | ||
| ### Inputs | ||
| - `in` accepts `RawAudio(AudioFormat { sample_rate: 16000, channels: 1, sample_format: F32 })` (one) | ||
|
|
||
| ### Outputs | ||
| - `out` produces `Transcription` (broadcast) | ||
|
|
||
| ## Parameters | ||
| | Name | Type | Required | Default | Description | | ||
| | --- | --- | --- | --- | --- | | ||
| | `execution_provider` | `string enum[cpu, cuda, tensorrt]` | no | `cpu` | Execution provider (cpu, cuda, tensorrt) | | ||
| | `max_segment_duration_secs` | `number` | no | `30.0` | Maximum segment duration before forced transcription (seconds)<br />min: `5`<br />max: `120` | | ||
| | `min_silence_duration_ms` | `integer` | no | `700` | Minimum silence duration before transcription (milliseconds)<br />min: `100`<br />max: `5000` | | ||
| | `model_dir` | `string` | no | `models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8` | Path to Parakeet TDT model directory (contains encoder.int8.onnx, decoder.int8.onnx, joiner.int8.onnx, tokens.txt). IMPORTANT: Input audio must be 16kHz mono f32. | | ||
| | `num_threads` | `integer` | no | `4` | Number of threads for inference<br />min: `1`<br />max: `16` | | ||
| | `use_vad` | `boolean` | no | `true` | Enable VAD-based segmentation | | ||
| | `vad_model_path` | `string` | no | `models/silero_vad.onnx` | Path to Silero VAD ONNX model file | | ||
| | `vad_threshold` | `number` | no | `0.5` | VAD speech probability threshold (0.0-1.0)<br />min: `0`<br />max: `1` | | ||
|
|
||
| ## Example Pipeline | ||
|
|
||
| ```yaml | ||
| # | ||
| # skit:input_asset_tags=speech | ||
|
|
||
| name: Speech-to-Text (Parakeet TDT) | ||
| description: Fast English speech transcription using NVIDIA Parakeet TDT (~10x faster than Whisper on CPU) | ||
| mode: oneshot | ||
| steps: | ||
| - kind: streamkit::http_input | ||
|
|
||
| - kind: containers::ogg::demuxer | ||
|
|
||
| - kind: audio::opus::decoder | ||
|
|
||
| - kind: audio::resampler | ||
| params: | ||
| chunk_frames: 960 | ||
| output_frame_size: 960 | ||
| target_sample_rate: 16000 | ||
|
|
||
| - kind: plugin::native::parakeet | ||
| params: | ||
| model_dir: models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8 | ||
| num_threads: 4 | ||
| use_vad: true | ||
| vad_model_path: models/silero_vad.onnx | ||
| vad_threshold: 0.5 | ||
| min_silence_duration_ms: 700 | ||
|
|
||
| - kind: core::json_serialize | ||
| params: | ||
| pretty: false | ||
| newline_delimited: true | ||
|
|
||
| - kind: streamkit::http_output | ||
| params: | ||
| content_type: application/json | ||
| ``` | ||
|
|
||
|
|
||
| <details> | ||
| <summary>Raw JSON Schema</summary> | ||
|
|
||
| ```json | ||
| { | ||
| "properties": { | ||
| "execution_provider": { | ||
| "default": "cpu", | ||
| "description": "Execution provider (cpu, cuda, tensorrt)", | ||
| "enum": [ | ||
| "cpu", | ||
| "cuda", | ||
| "tensorrt" | ||
| ], | ||
| "type": "string" | ||
| }, | ||
| "max_segment_duration_secs": { | ||
| "default": 30.0, | ||
| "description": "Maximum segment duration before forced transcription (seconds)", | ||
| "maximum": 120.0, | ||
| "minimum": 5.0, | ||
| "type": "number" | ||
| }, | ||
| "min_silence_duration_ms": { | ||
| "default": 700, | ||
| "description": "Minimum silence duration before transcription (milliseconds)", | ||
| "maximum": 5000, | ||
| "minimum": 100, | ||
| "type": "integer" | ||
| }, | ||
| "model_dir": { | ||
| "default": "models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8", | ||
| "description": "Path to Parakeet TDT model directory (contains encoder.int8.onnx, decoder.int8.onnx, joiner.int8.onnx, tokens.txt). IMPORTANT: Input audio must be 16kHz mono f32.", | ||
| "type": "string" | ||
| }, | ||
| "num_threads": { | ||
| "default": 4, | ||
| "description": "Number of threads for inference", | ||
| "maximum": 16, | ||
| "minimum": 1, | ||
| "type": "integer" | ||
| }, | ||
| "use_vad": { | ||
| "default": true, | ||
| "description": "Enable VAD-based segmentation", | ||
| "type": "boolean" | ||
| }, | ||
| "vad_model_path": { | ||
| "default": "models/silero_vad.onnx", | ||
| "description": "Path to Silero VAD ONNX model file", | ||
| "type": "string" | ||
| }, | ||
| "vad_threshold": { | ||
| "default": 0.5, | ||
| "description": "VAD speech probability threshold (0.0-1.0)", | ||
| "maximum": 1.0, | ||
| "minimum": 0.0, | ||
| "type": "number" | ||
| } | ||
| }, | ||
| "type": "object" | ||
| } | ||
| ``` | ||
|
|
||
| </details> |
|
Contributor
Author
There was a problem hiding this comment. 🚩 download-models target doesn't include download-parakeet-models The 
Was this helpful? React with 👍 or 👎 to provide feedback. Debug
Contributor
Author
There was a problem hiding this comment. Good catch. At ~660MB the Parakeet model is substantially larger than most other models, so I've kept it out of the default |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Uh oh!
There was an error while loading. Please reload this page.