Audio Understanding - Speech Recognition (Video)

<aside> 🛠

Recognize Speech in a Video and Convert It to Text

Recognize Speech in a Video and Convert It to Text Typing up what was said in a recording is tedious, and manually pulling the audio out first is an extra chore. This node does both in one shot — hand it a single video and get back a clean text transcript of everything spoken in it, audio extraction handled for you. It runs on OpenAI’s Whisper Large V3, state-of-the-art open-source speech recognition that holds up on accents, background noise, and technical vocabulary. Language auto-detects by default, or force English, Korean, Japanese, or Chinese.

🏆 7.44 mean WER on the Open ASR Leaderboard, running at 145.51× real-time (RTFx)

</aside>

What it does

This node takes one video file as input, pulls out its audio track, and returns a text transcript of everything spoken in it. The underlying model is OpenAI's Whisper Large V3 — one of the strongest open-source speech recognition models available, trained by OpenAI on 5 million hours of internet audio and known for handling accents, background noise, and technical vocabulary well. The studio exposes a curated language picker (Auto, English, Korean, Japanese, Chinese); the model itself supports 99 languages, so if you feed it audio in another language it will typically still work in Auto mode. This node is designed for the "one video → one transcript" workflow. If you have a list of clips and need timed transcripts per clip, use the sibling node Audio Understanding - Speech Recognition (Clips) instead.

Problem it solves

One-shot video transcription – Feed a full video, get back the full spoken text — no manual audio extraction, no chunking
Multilingual out of the box – Auto-detects the spoken language; explicit picker for the four most common languages in the studio's typical workflows
Meeting / lecture / interview recordings – Turn long-form video recordings into readable text
Podcast or video captioning – Get a full transcript for further editing into subtitles or show notes
Content search & indexing – Make video content searchable via its transcript text
Robust to real-world audio – Whisper Large V3 handles background noise, accents, and varied recording quality well

Input/Output

<aside>

Input: A single video file
- The node extracts the audio track from the video automatically </aside>

<aside>

Output: Plain-text transcript of everything spoken in the video </aside>

<aside>

Options:
- Spoken language (dropdown, default Auto)
  - Auto — model detects the language from the audio itself (recommended for mixed or unknown-language content)
  - English — force English
  - Korean — force Korean
  - Japanese — force Japanese
  - Chinese — force Chinese
  - Note: under the hood the model supports 99 languages; the studio's picker is scoped to the four most-used explicit choices
- Output mode (dropdown, default Transcribe)
  - Transcribe — keeps the spoken language (audio in language X → text in language X)
  - Translate to English — uses Whisper's native speech-to-English mode (audio in any language → English text)
  - Tip: run the same video twice (Transcribe + Translate to English) to get both transcripts for a bilingual subtitle stack. </aside>

Accuracy & Speed


Model	OpenAI Whisper Large V3 — state-of-the-art open-source speech recognition
Mean WER (Open ASR Leaderboard)	7.44 — among the strongest open ASR models
Real-Time Factor (RTFx)	145.51 — much faster than real-time on capable hardware
Improvement over Whisper V2	10 – 20% fewer errors across most languages
Languages	99 supported by the model; UI-exposed choices: Auto, English, Korean, Japanese, Chinese
Long audio handling	Automatic — Whisper's 30-second receptive field is handled behind the scenes via chunked or sequential long-form inference
Model size	1.55B parameters

How the Video variant compares

	Speech Recognition (Video)	Speech Recognition (Clips)
Input	A single video file	A list of video clips
Output	One plain-text transcript	A timed transcript per clip
Best for	One video → one transcript	Per-scene / per-highlight captioning, subtitle alignment
Underlying model	Whisper Large V3	Whisper Large V3

Both run inside CNAPS Studio — for batch clip lists with per-clip timestamps use the sibling Audio Understanding - Speech Recognition (Clips) node, or compare them side-by-side.

Technical Details (for engineers)