---

name: recognition-engines description: "Recognizer engine selection and transcription APIs for SpeechRecognition 3.17.0." disable-model-invocation: true metadata: disco-role: operating license: NOASSERTION

Recognition Engines

Use this sub-skill when a task is about choosing or calling a SpeechRecognition recognizer engine, interpreting transcription return values, or diagnosing engine-specific optional dependency, credential, language, model, or network failures.

Use this for

• Recognizer.recognize_google, recognize_google_cloud, recognize_sphinx, recognize_vosk, recognize_whisper, recognize_faster_whisper, recognize_openai, recognize_groq, and recognize_cohere_api.
• Legacy/service methods still present on Recognizer: recognize_wit, recognize_azure, recognize_houndify, recognize_ibm, recognize_amazon, recognize_lex, recognize_assemblyai, recognize_tensorflow, and recognize_api.
• show_all, verbose, show_dict, confidence tuple variants, raw response objects, and UnknownValueError versus RequestError versus SetupError handling.
• Optional extras, SDK credentials, local Sphinx/Vosk/Whisper/Faster-Whisper model notes, and OpenAI-compatible base URL setup.

Route elsewhere

• File loading, AudioData conversion, segmentation, upload-size chunking, WAV/AIFF/FLAC details: route to audio-data.
• Microphone acquisition, calibration, listen, background callbacks, and capture-time thresholds: route to capture-listening.
• sprc download vosk, CLI help, environment probes, and model-download side effects: route to cli-model-setup.
• Repository edits, recognizer tests, CI, docs updates, and maintainer workflows: route to repo-development.

Read first

1. references/engine-selection.md to choose a local, default web, cloud SDK, OpenAI-compatible, or legacy engine.
2. references/api-reference.md for verified signatures, parameters, return shapes, and exceptions.
3. references/cloud-and-engine-reference.md for optional extras, credentials, local model requirements, and raw response behavior.
4. references/troubleshooting.md when an engine fails or returns no transcript.

Bundled helpers

• scripts/transcribe_file.py: inspect an audio file without network access by default, or transcribe a user-supplied file with an explicitly selected engine.
• scripts/sphinx_keyword_grammar_template.py: PocketSphinx keyword and grammar template with argument validation and no bundled source-example dependency.

These helpers are examples/templates. They read user-supplied audio paths and the installed SpeechRecognition package; they do not require the original source repository.