parakeet-ctc-1.1b

Name: parakeet-ctc-1.1b
Author: nvidia

by nvidia

Fast Conformer CTC model for high-throughput English speech recognition

1.7Mdl/month

58likes

1.1Bparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

nvidia/parakeet-ctc-1.1b

Feature URI

mixpeek://transcription@v1/nvidia_parakeet_ctc_1b_v1

Overview

Parakeet CTC 1.1B is NVIDIA's XXL Fast Conformer model with a CTC decoder, trained on 64K hours of English speech. The convolutional-transformer hybrid architecture processes audio up to 11 hours in a single pass on an A100 80GB GPU, achieving a real-time factor of 1,336x (1,336 hours of audio transcribed per hour of compute).

On Mixpeek, Parakeet CTC powers high-throughput English transcription for large audio and video libraries where speed matters. Its non-autoregressive CTC decoding enables massive parallelism, making it ideal for batch processing millions of hours of content.

Architecture

Fast Conformer encoder (CNN + Transformer hybrid) with 1.1B parameters and CTC (Connectionist Temporal Classification) decoder. Processes 80-channel log-mel spectrograms. Supports local attention for processing audio segments up to 11 hours on A100 80GB.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so parakeet-ctc-1.1b runs
// on your side and the output is upserted through POST
// /v1/namespaces/{namespace_id}/documents/upsert. On Enterprise the other
// path is to upload the weights instead: POST /v1/namespaces/{id}/models
// accepts the huggingface format and a custom plugin loads them.
const res = await fetch(
  "https://api.mixpeek.com/v1/namespaces/ns_your_namespace/documents/upsert",
  {
    method: "POST",
    headers: {
      Authorization: "Bearer API_KEY",
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      collection_id: "col_your_collection",
      documents: [
        {
          document_id: "asset-00412",
          // The model produces text, so it lands in payload. Give the
          // collection a text vector index and embed that text to make it
          // searchable rather than only filterable.
          payload: { extracted_text: modelOutput, source_key: "archive/2026/asset-00412" },
          vectors: { "text-embedding": embeddingOfModelOutput },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// universal_extractor@v1 runs google/gemini-embedding-2
// (3072-d) over a bucket, with no inference of your own.