Batch Audio Transcription

Transcribe meetings, podcasts, interviews, and customer calls at scale using the Runcrate Whisper API. This guide covers single-file transcription, batch processing across a folder, and output format options.

What you’ll build

A pipeline that transcribes multiple audio files using Whisper large-v3 through the Runcrate API. The pipeline reads a folder of recordings, transcribes each one, and writes the results as plain text, SRT subtitles, or structured JSON — your choice.

Single file transcription

curl https://api.runcrate.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer rc_live_YOUR_API_KEY" \
  -F model="openai/whisper-large-v3" \
  -F file=@meeting-2025-05-19.mp3

from openai import OpenAI

client = OpenAI(
    base_url="https://api.runcrate.ai/v1",
    api_key="rc_live_YOUR_API_KEY",
)

transcript = client.audio.transcriptions.create(
    model="openai/whisper-large-v3",
    file=open("meeting-2025-05-19.mp3", "rb"),
)
print(transcript.text)

import OpenAI from "openai";
import fs from "fs";

const client = new OpenAI({
  baseURL: "https://api.runcrate.ai/v1",
  apiKey: "rc_live_YOUR_API_KEY",
});

const transcript = await client.audio.transcriptions.create({
  model: "openai/whisper-large-v3",
  file: fs.createReadStream("meeting-2025-05-19.mp3"),
});
console.log(transcript.text);

from runcrate import Runcrate

client = Runcrate(api_key="rc_live_YOUR_API_KEY")

with open("meeting-2025-05-19.mp3", "rb") as f:
    result = client.models.transcribe(
        model="openai/whisper-large-v3",
        file=f,
        filename="meeting-2025-05-19.mp3",
    )

print(result.text)

Batch processing

Transcribe every audio file in a folder and write the results to disk. This example uses the Runcrate Python SDK and outputs SRT subtitle files.

import os
from pathlib import Path
from runcrate import Runcrate

client = Runcrate(api_key="rc_live_YOUR_API_KEY")

audio_dir = Path("./recordings")
output_dir = Path("./transcripts")
output_dir.mkdir(exist_ok=True)

SUPPORTED = {".mp3", ".wav", ".m4a", ".flac", ".ogg", ".webm"}

for audio_file in sorted(audio_dir.iterdir()):
    if audio_file.suffix.lower() not in SUPPORTED:
        continue

    with open(audio_file, "rb") as f:
        result = client.models.transcribe(
            model="openai/whisper-large-v3",
            file=f,
            filename=audio_file.name,
            response_format="srt",
        )

    srt_path = output_dir / f"{audio_file.stem}.srt"
    srt_path.write_text(result.text)
    print(f"Transcribed: {audio_file.name} → {srt_path.name}")

Batch processing with the OpenAI SDK

The same pattern works with the OpenAI Python SDK pointed at the Runcrate API:

import os
from pathlib import Path
from openai import OpenAI

client = OpenAI(
    base_url="https://api.runcrate.ai/v1",
    api_key="rc_live_YOUR_API_KEY",
)

audio_dir = Path("./recordings")
output_dir = Path("./transcripts")
output_dir.mkdir(exist_ok=True)

for audio_file in sorted(audio_dir.glob("*.mp3")):
    transcript = client.audio.transcriptions.create(
        model="openai/whisper-large-v3",
        file=open(audio_file, "rb"),
    )

    txt_path = output_dir / f"{audio_file.stem}.txt"
    txt_path.write_text(transcript.text)
    print(f"Transcribed: {audio_file.name} → {txt_path.name}")

Output formats

Format	Extension	Use case
`text`	`.txt`	Plain transcript — search, summarization, RAG ingestion
`json`	`.json`	Structured output with word-level timestamps
`srt`	`.srt`	Subtitle file for video editors (Premiere, DaVinci, Final Cut)
`vtt`	`.vtt`	Web video subtitles (HTML5 `<track>` element)

Pass the format via the response_format parameter:

result = client.models.transcribe(
    model="openai/whisper-large-v3",
    file=f,
    filename="episode-42.mp3",
    response_format="srt",  # or "text", "json", "vtt"
)

Language hints

Whisper auto-detects the spoken language, but you can improve accuracy on non-English audio by passing a language hint:

result = client.models.transcribe(
    model="openai/whisper-large-v3",
    file=f,
    filename="interview-tokyo.mp3",
    language="ja",  # ISO 639-1 code
)

Use cases

Meeting recordings — transcribe and feed into an LLM for searchable notes and action items.
Podcast episodes — generate full transcripts for show notes, blog posts, and SEO.
Customer support calls — bulk-transcribe for quality analysis and compliance review.
Lecture recordings — produce study materials and make content accessible.
Video content — generate SRT/VTT subtitle files for automatic captioning.

Tips

Supported formats. MP3, WAV, M4A, FLAC, OGG, and WebM.
Large files. For recordings longer than ~2 hours, split into chunks before uploading. Tools like ffmpeg make this easy: ffmpeg -i long-meeting.mp3 -f segment -segment_time 1800 -c copy chunk_%03d.mp3.
SRT for video editing. SRT is the most widely supported subtitle format across video editors and media players. Use VTT only if you need web-native <track> elements.
Combine with chat models. Pipe transcripts into a Runcrate chat model for summarization, action-item extraction, or translation — see the RAG Pipeline guide for the pattern.

Inference Examples

Compute Examples

Storage Examples

Batch Audio Transcription

What you’ll build

Single file transcription

Batch processing

Batch processing with the OpenAI SDK

Output formats

Language hints

Use cases

Tips

​What you’ll build

​Single file transcription

​Batch processing

​Batch processing with the OpenAI SDK

​Output formats

​Language hints

​Use cases

​Tips

What you’ll build

Single file transcription

Batch processing

Batch processing with the OpenAI SDK

Output formats

Language hints

Use cases

Tips