cn8/ai-transcript

⚠ PRODUCT DISABLED (2026-07-29) — the transcription machine

Learn more about AI Transcript & Video Understanding

Videoasync (job)Instant previewaudiovideotranscription

Input

video_url *string

Drag & drop, paste a URL, or choose a fileVideo — MP4, MOV or WebM

Public URL of the source video or audio file.

Instant sample preview — no cost, no account. Sign up free to run with your own input.

Result

Completed Instant preview

Json

{
  "language": "en",
  "duration_seconds": 12,
  "segments": [
    {
      "start": 0,
      "end": 3.6,
      "text": "Every AI capability behind one key — that's the promise."
    },
    {
      "start": 3.6,
      "end": 8.4,
      "text": "Chat, image, video and voice, all through one API."
    },
    {
      "start": 8.4,
      "end": 11.8,
      "text": "Let me show you how fast you can ship."
    }
  ]
}

Sample output — generated with “Keynote clip”. Sign in to run your own input.

This request costs 0.001 per second.

Overview

⚠ PRODUCT DISABLED (2026-07-29) — the transcription machine (speech-recognition.cinema8.com) is offline and the whole transcript category is pulled from the gateway catalog. Docs kept for re-enable; see service_catalog.yaml for the commented-out category block and the DB cleanup DELETE calls.

A bundled toolkit for turning raw video into structured, searchable, interactive content. Transcription produces word-level timestamps and subtitle files (VTT / SRT). Scene captioning splits the video into shots and writes a visual caption for each one. Speaker diarization labels who spoke when. Content generators turn the transcript into video metadata (title, summary, tags), chaptered timestamps, quiz questions and interactive elements (CTA panels, tag buttons, lead forms). Translation localizes subtitles into another language while preserving timings.

Lightweight endpoints respond synchronously; heavier ones run as async jobs — submit a job, poll until it's ready, and read the result.

Key capabilities

Accurate transcription

Word-level timestamps, automatic language detection, and subtitle output in VTT and/or SRT.

Scene-aware visual captioning

The video is split into shots and each one gets a visual caption. When a transcript is supplied, captions are refined using the surrounding context for tighter labels.

Speaker diarization

Identify who spoke when. Supply the number of speakers, or let the service detect it automatically.

Content generators

Generate metadata (title, summary, tags), chapters, quiz questions and interactive video elements from the transcript.

Multilingual translation

Translate subtitles into another language. The best translation engine is chosen automatically for each language pair, or you can force one with force_backend.

Async jobs with auto-polling

Heavier endpoints return a job ID; poll GET /v1/jobs/{id} until the result is ready.

When to use it

Lecture and webinar enrichment

Transcribe → generate chapters and a quiz → translate to multiple languages → embed interactive CTAs.

Podcast post-production

Transcribe and diarize speakers, then auto-generate show notes (metadata) and chapter timestamps.

Searchable video archives

Build a transcript + scene-caption index for full-text search across video libraries.

Localization pipelines

Translate VTT/SRT subtitles between languages while preserving segment timings.

Input & output

input

A public video or audio URL (mp4 / mov / mkv / avi / webm and common audio formats) for media endpoints, or subtitle content / transcript segments for the text-based generators.

JSON body

output

Async endpoints return a job ID; on completion you get subtitle URLs, segments, captions, a chapter list, a quiz, and so on. Synchronous endpoints (metadata, quiz, interactives) return JSON directly.

JSONVTT / SRT URL

Guides & tips

Typical pipeline order

1transcript-transcribe → POST a video URL, get back VTT/SRT subtitle URLs.
2Download the VTT (or SRT) — that file is your input for the next steps.
3transcript-metadata / transcript-chapters → POST the subtitle content for enrichment (title, summary, chapters).
4transcript-quiz → POST the subtitle to generate quiz questions.
5transcript-translate → POST the subtitle and a target language to localize it.
6transcript-caption → POST the original video URL again for scene-level visual captions.

How async polling works

Async endpoints (transcribe, chapters, caption, translate) return { "status": "accepted", "job": { "id": "<gateway_job_id>" } } with HTTP 202. Poll GET /v1/jobs/{gateway_job_id} until status is completed, then read the result from the job. Each poll returns the job's current status, so you can show progress or simply wait for completion.

Webhooks

Instead of polling, you can configure a webhook on your API key. The final result is delivered there when the job completes (or fails). See Webhooks under settings.

Specs

Latency

Synchronous endpoints: a few seconds. Async transcribe/caption: scales with media length (minutes for long video).

Async

true

Rate Limit

Per API key

Max Input

Duration-dependent; longer media takes longer to process

Schema

Request body

video_urlrequiredstring

Public URL of the source video or audio file.

languagestring

ISO-639-1 language hint (e.g. 'en', 'tr'). Auto-detected if omitted.

subtitle_formatsarray · default: vtt

Subtitle file formats to emit.

enum: vtt, srt

word_timestampsboolean · default: false

Include word-level timestamps in the result.

Pricing

Media endpoints billed by source duration; text-generator endpoints by request.

Service	Unit	Price
Transcribe Video	second	$0.001/sec
Scene Captioning	second	$0.001/sec
Video Metadata	item	$0.04/req
Chapter Generation	item	$0.05/req
Quiz Generation	item	$0.04/req
Subtitle Translate	item	$0.02/req
Audio Description Script	second	$0.003/sec

Async jobs are billed only when they complete successfully.

Audio-only inputs (subtitle only, no media) are billed at the same flat per-request rate.

There's no extra cost for parsing subtitles — it's included in the metadata/chapters/quiz/translate flat fee.

FAQ

Related models

Dubbing

Use transcript-translate output to drive subtitle localization, or send the source video to dubbing for full voice translation.

Speech Clearance

Denoise speech before transcribing for cleaner segments.

Text to Speech

Render translated segments to speech with TTS for localized voice tracks.

Voice Clone Dubbing

Preserve the original speaker while dubbing to another language.