Transcription

• ElevenLabs Scribe v2

Recipes using this pipeline

• Long video → multiple captioned clips, in one command

  Slice any long video into N captioned shorter clips. Transcribe once, auto-pick the moments with AI, trim + caption each in parallel. Optionally reformat to vertical (9:16) for TikTok / Reels / Shorts, square (1:1) for Instagram, or portrait (4:5); the caption anchor auto-adjusts.

  6 pipelines · CLI

Articles about this pipeline

• How to Trim a Video to a Clean Clip That Never Cuts Mid-Sentence

• How to Turn a Long Video Into Shorts and Reels

• What Is Generative AI? How It Works and What It Makes

• How to Add Subtitles to a Video with an SRT File

• How to Make AI Videos: From Prompt to Finished Video

Transcription API

Call this pipeline from your own code. One request dispatches a run; the model is an input field, not a separate endpoint.

import { createClient } from "@pipe2-ai/sdk";

const client = createClient(process.env.PIPE2_TOKEN!);

const { run_pipeline } = await client.RunPipeline({
  pipeline_slug: "transcription",
  input: {
    source_asset_id: "a1b2c3d4-0000-4000-8000-000000000001",
  },
});

import asyncio
import os

from pipe2 import Pipe2Client

client = Pipe2Client(token=os.environ["PIPE2_TOKEN"])

run = asyncio.run(client.run_pipeline(
    pipeline_slug="transcription",
    input={
        "source_asset_id": "a1b2c3d4-0000-4000-8000-000000000001",
    },
))

client := pipe2.NewClient(os.Getenv("PIPE2_TOKEN"))

input := json.RawMessage(`{
  "source_asset_id": "a1b2c3d4-0000-4000-8000-000000000001"
}`)

run, err := pipe2.RunPipeline(context.Background(), client, "transcription", input)
if err != nil {
    log.Fatal(err)
}

pipe2 pipelines run \
  --pipeline transcription \
  --input-json '{"source_asset_id": "a1b2c3d4-0000-4000-8000-000000000001"}' \
  --wait

Get an API token →

Parameters

source_asset_id required

Audio or video to transcribe. Asset id returned by an upload. See /docs/api/ for the upload flow.

string

corrections

Optional word-boundary substitutions applied to the rendered SRT and TXT, useful when the recognizer mis-hears a specific word (a name, an acronym, a domain term) and writes it the same wrong way every time. Each key is the word as transcribed; the value is the correct spelling. Case-sensitive; matches must align to whole words. Leave empty to ship the transcript exactly as the recognizer wrote it.

object default [object Object]

diarize

Label who is speaking. Turn on for interviews and multi-speaker recordings.

boolean default false

Use it from an AI agent

Point any MCP client at pipe2 and your agent gets these tools. It finds this pipeline, reads the same schema above, then runs it.

• list_pipelines
• get_pipeline_schema
• run_pipeline
• get_pipeline_run_status
• request_upload

{
  "mcpServers": {
    "pipe2ai": {
      "url": "https://mcp.pipe2.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Full API reference→ MCP setup→ CLI reference→

Frequently Asked Questions