Doomersion API

Doomersion Video API

Transcription engine for short-form video.

Word-level timing. Speaker ID. Dialect. 90+ languages.
Submit a URL, poll for completion, and pull clean structured transcripts into your own bucket.

Get API key Explore docs
Localization Language detected eng · 90+ locales
Transcription Structured transcript speaker_0 · diarized
Timestamping Word-level timing start_ms / end_ms

Capabilities

One job. Six layers of structure.

Captioning and localization ship on every request. Timestamping and speaker tags come with the transcript. Dialect detection and video screenshots are production-ready in our pipeline and can be enabled on your API key on request.

Captioning

Short-form ASR tuned on compressed audio, fast talkers, and music beds. 98%+ word accuracy with production-ready transcripts.

Localization

Detect language on every clip as an ISO 639-3 code. Route captions, translation, and catalog workflows without a separate detect step.

Timestamping

Word-level start_ms / end_ms on every token. Sync captions, scrubbers, and highlight reels to the frame.

Speaker tags

Diarization with real names and roles when metadata allows. Not just speaker_0: Host, Guest, Lecturer, Narrator.

Video screenshots

On request

Timed frame grabs aligned to speech. Use them for moderation, thumbnails, scene search, or multimodal context.

Dialect

On request

Classify spoken variety beyond language code: Arabic dialects, Mandarin/Cantonese, Brazilian vs European Portuguese, and more.

Live demo

Timestamps, languages, and speakers on real clips.

Three @doomlingo TikToks from our pipeline. Toggle the tabs or watch them auto-rotate: word-level timing, bilingual localization, and a live talker indicator for diarization.

Benchmark

Built for short-form. Measured against the field.

Tuned on millions of social clips: compressed audio, fast talkers, music beds, and rapid cuts. 98%+ caption accuracy with word-level timing and speaker ID out of the box.

  1. 01 Doomersion ASR Ours 98%+
  2. 02 AssemblyAI ~93%
  3. 03 Whisper Large v3 ~92%
  4. 04 Deepgram Nova ~91%
  5. 05 YouTube auto-captions ~85%

Competitor figures are approximate published word-accuracy ranges on clean / general audio. Doomersion's 98%+ is measured on short-form social video in production.

Coverage

Localize in 90+ languages

Detected language lands in every transcript as an ISO 639-3 code. Use it to route captions, translations, and downstream localization without a separate detect step.

  • AFR Afrikaans
  • ARA Arabic
  • HYE Armenian
  • ASM Assamese
  • AZE Azerbaijani
  • BEL Belarusian
  • BEN Bengali
  • BOS Bosnian
  • BUL Bulgarian
  • CAT Catalan
  • CEB Cebuano
  • NYA Chichewa
  • HRV Croatian
  • CES Czech
  • DAN Danish
  • NLD Dutch
  • ENG English
  • EST Estonian
  • FIL Filipino
  • FIN Finnish
  • FRA French
  • GLG Galician
  • KAT Georgian
  • DEU German
  • ELL Greek
  • GUJ Gujarati
  • HAU Hausa
  • HEB Hebrew
  • HIN Hindi
  • HUN Hungarian
  • ISL Icelandic
  • IND Indonesian
  • GLE Irish
  • ITA Italian
  • JPN Japanese
  • JAV Javanese
  • KAN Kannada
  • KAZ Kazakh
  • KIR Kirghiz
  • KOR Korean
  • LAV Latvian
  • LIN Lingala
  • LIT Lithuanian
  • LTZ Luxembourgish
  • MKD Macedonian
  • MSA Malay
  • MAL Malayalam
  • CMN Mandarin Chinese
  • MAR Marathi
  • NEP Nepali
  • NOR Norwegian
  • PUS Pashto
  • FAS Persian
  • POL Polish
  • POR Portuguese
  • PAN Punjabi
  • RON Romanian
  • RUS Russian
  • SRP Serbian
  • SND Sindhi
  • SLK Slovak
  • SLV Slovenian
  • SOM Somali
  • SPA Spanish
  • SWA Swahili
  • SWE Swedish
  • TAM Tamil
  • TEL Telugu
  • THA Thai
  • TUR Turkish
  • UKR Ukrainian
  • URD Urdu
  • VIE Vietnamese
  • CYM Welsh

Showing 74 of 90+ languages in production. Same stack powers Creator Studio and the Video API.

Your credentials stay yours

Bucket keys are used only for that job's upload and are never stored after the job finishes.

We don't keep your results

Output lands in your S3-compatible bucket. Doomersion does not store transcripts or video files on our side.

Bring your own bucket

AWS S3, Cloudflare R2, MinIO. Up to 10,000 URLs per job with parallel workers.

API reference

Or build anything with a simple REST surface.

https://api.doomlingo.ai OpenAPI Spec Quick start

Quick Start

Complete flow. You need an API key plus your own S3-compatible bucket credentials. Results are written to your bucket.

01

Submit a video

POST /jobs with YouTube URLs and your bucket object. Replace bucket credentials with yours.

02

Poll until completed

GET /jobs/{id} every 10โ€“30 seconds until status is "completed".

03

Get results

GET /jobs/{id}/results for S3 paths, then download .mp4 and .json from your bucket.

# 1. Submit a video (replace bucket credentials with yours)
curl -X POST https://api.doomlingo.ai/jobs \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "video",
    "urls": ["https://www.youtube.com/watch?v=jNQXAC9IVRw"],
    "bucket": {
      "name": "your-bucket",
      "endpoint": "https://ACCOUNT_ID.r2.cloudflarestorage.com",
      "access_key": "YOUR_ACCESS_KEY",
      "secret_key": "YOUR_SECRET_KEY",
      "region": "auto"
    }
  }'
# โ†’ {"job_id": "550e8400-...", "status": "queued", "video_count": 1}

# 2. Poll until completed (repeat every 15s)
curl https://api.doomlingo.ai/jobs/550e8400-... \
  -H "X-API-Key: YOUR_KEY"
# โ†’ {"status": "completed", "total": 1, "completed": 1, ...}

# 3. Get results
curl https://api.doomlingo.ai/jobs/550e8400-.../results \
  -H "X-API-Key: YOUR_KEY"
# โ†’ {"results": [{"status": "done", "result_path": "s3://bucket/...", ...}]}

Average processing time: 30โ€“90 seconds per video. If a job hasn't completed after 5 minutes, check the health endpoint or contact support.

GET /health

Check if the API is running and how many workers are active. No authentication required.

curl https://api.doomlingo.ai/health
{"status": "ok", "workers": 4}

Authentication

Include your API key in every request header:

X-API-Key: your_api_key_here

Contact mostafa@doomlingo.ai to get an API key.

POST /jobs

Submit YouTube URLs for processing. Each URL will be downloaded, transcribed with speaker diarization and word-level timing, and uploaded to your bucket.

Request Body

FieldTypeDescription
type string required "video" for individual videos, "channel" for channel listing
urls string[] required YouTube URLs (max 10,000 per job)
bucket object required Your S3-compatible output bucket. Object with name, endpoint, access_key, secret_key, and region. The token must allow PutObject (upload). Works with AWS S3, Cloudflare R2, MinIO, etc.

Bucket object

FieldRequiredDescription
nameyesBucket name
endpointyesS3 API endpoint URL (e.g. https://<account>.r2.cloudflarestorage.com or https://s3.us-east-1.amazonaws.com)
access_keyyesAccess key ID
secret_keyyesSecret access key
regionyesRegion string (auto for Cloudflare R2; e.g. us-east-1 for AWS)

Each job writes {job_id}/{video_id}.mp4 and {job_id}/{video_id}.json into this bucket.

Example

# Submit 3 videos for processing
curl -X POST https://api.doomlingo.ai/jobs \
  -H "X-API-Key: your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "video",
    "urls": [
      "https://www.youtube.com/shorts/abc123def45",
      "https://www.youtube.com/watch?v=xyz789ghi01",
      "https://youtu.be/mno456pqr78"
    ],
    "bucket": {
      "name": "your-bucket",
      "endpoint": "https://ACCOUNT_ID.r2.cloudflarestorage.com",
      "access_key": "YOUR_ACCESS_KEY",
      "secret_key": "YOUR_SECRET_KEY",
      "region": "auto"
    }
  }'

Response

{
  "job_id": "550e8400-e29b-41d4-a716-446655440000",
  "status": "queued",
  "video_count": 3
}

GET /jobs/{job_id}

Poll the status of a submitted job. A job moves through: queued โ†’ processing โ†’ completed.

Example

curl https://api.doomlingo.ai/jobs/550e8400-... \
  -H "X-API-Key: your_key"

Response

{
  "job_id": "550e8400-...",
  "type": "video",
  "status": "processing",
  "total": 3,
  "completed": 1,
  "failed": 0,
  "created_at": 1780000000.0,
  "completed_at": null
}

Poll every 10โ€“30 seconds until status is "completed".

GET /jobs/{job_id}/results

Get per-video results. Works at any stage. You can check individual video status even while the job is still processing. Each completed video includes the S3 path where the video and transcript were uploaded. The transcript includes speaker identification (real names inferred from video metadata).

Response

{
  "job_id": "550e8400-...",
  "status": "completed",
  "results": [
    {
      "video_url": "https://www.youtube.com/shorts/abc123def45",
      "status": "done",
      "result_path": "s3://bucket/550e8400-.../abc123def45",
      "error": null
    },
    {
      "video_url": "https://www.youtube.com/watch?v=xyz789ghi01",
      "status": "done",
      "result_path": "s3://bucket/550e8400-.../xyz789ghi01",
      "error": null
    }
  ]
}

Output Format

For each video, two files are uploaded to your bucket:

FileDescription
{job_id}/{video_id}.mp4The downloaded video file
{job_id}/{video_id}.jsonTranscript with timing & speaker data

Transcript JSON

{
  "video_url": "https://www.youtube.com/shorts/abc123def45",
  "video_id": "abc123def45",
  "duration_s": 28.4,
  "language": "eng",
  "transcript": "How do you say good morning in Japanese?",
  "metadata": {
    "channel_name": "Doomlingo",
    "channel_id": "UCxxxxxxxx",
    "channel_url": "https://www.youtube.com/@doomlingo",
    "title": "How to say good morning in Japanese",
    "description": "...",
    "upload_date": "2026-05-15",
    "view_count": 125000,
    "like_count": 8500
  },
  "speakers": [
    {"id": "speaker_0", "name": "Maya", "role": "Host"}
  ],
  "segments": [
    {
      "text": "How do you say good morning in Japanese?",
      "speaker": "Maya",
      "start_ms": 420,
      "end_ms": 2840,
      "words": [
        {"word": "How", "start_ms": 420, "end_ms": 610},
        {"word": "do", "start_ms": 630, "end_ms": 720},
        {"word": "you", "start_ms": 740, "end_ms": 860},
        {"word": "say", "start_ms": 880, "end_ms": 1040}
      ]
    }
  ]
}

Field Reference

FieldTypeDescription
video_idstringPlatform video ID (YouTube 11-character ID, or the source ID for other supported URLs)
duration_snumberVideo duration in seconds
languagestringDetected language (ISO 639-3 code: eng, ara, jpn, etc.)
transcriptstringFull transcript as plain text
metadataobjectSource video metadata (channel name, title, description, view/like counts)
speakersarrayIdentified speakers with real names and roles (e.g. host, guest, lecturer)
speakers[].namestringReal name of the speaker, inferred from video metadata and transcript
speakers[].rolestringSpeaker's role (e.g. Lecturer, Host, Guest, Narrator)
segmentsarraySpeech segments, split by speaker turn
segments[].speakerstringReal name of the speaker (resolved from speakers list)
segments[].start_msnumberSegment start time in milliseconds
segments[].end_msnumberSegment end time in milliseconds
segments[].wordsarrayWord-level timestamps within the segment
segments[].words[].wordstringSpoken token
segments[].words[].start_msnumberWord start time in milliseconds
segments[].words[].end_msnumberWord end time in milliseconds

Transcript Preview FREE

Before processing a video with the full pipeline, you can preview YouTube's auto-generated transcript for free. Run this from your machine. No API credits, no API key needed.

Setup (one time)

pip install youtube-transcript-api

Fetch a transcript

# Python - get transcript for any YouTube video
from youtube_transcript_api import YouTubeTranscriptApi

ytt = YouTubeTranscriptApi()
for t in ytt.list("VIDEO_ID"):
    entries = t.fetch()
    print(f"Language: {t.language_code}, Segments: {len(entries)}")
    for e in entries:
        print(f"  [{e.start:.1f}s] {e.text}")
    break

Example output

Language: en, Segments: 12
  [0.0s] How do you say good morning
  [2.1s] in Japanese?
  [3.4s] Ohayou gozaimasu.
  ...

Typical workflow

1. Fetch free transcript locally       โ†’ preview content, filter videos
2. Pick the videos you want             โ†’ check language, speaker, topic
3. POST /jobs with URLs + your bucket   โ†’ full processing with speaker ID + word timing

The free transcript gives you YouTube's auto-captions. POST /jobs gives you production-grade transcription with word-level millisecond timing, real speaker names, and higher accuracy.

Integration Flow

1. POST /jobs with URLs + bucket credentials  โ†’ get job_id
2. Poll GET /jobs/{id} every 10-30 seconds   โ†’ wait for "completed"
3. GET /jobs/{id}/results                    โ†’ get S3 paths
4. Download .mp4 and .json from your bucket  โ†’ done

Average processing time: 30โ€“90 seconds per video (download + transcription).

Supported URL Formats

Production POST /jobs accepts YouTube URLs today:

  • https://www.youtube.com/shorts/VIDEO_ID
  • https://www.youtube.com/watch?v=VIDEO_ID
  • https://youtu.be/VIDEO_ID
  • https://www.youtube.com/embed/VIDEO_ID
  • https://www.youtube.com/@channelname (channel jobs only)

The live demo above uses real @doomlingo TikTok clips to showcase caption timing, localization, and speakers. Contact mostafa@doomlingo.ai if you need TikTok or other sources on the API.

Limits & Errors

Limits

LimitValue
Max URLs per job10,000
Parallel processing4 videos at a time (configurable)
Concurrent jobsUnlimited

HTTP Status Codes

StatusMeaning
200Success
400Bad request - invalid type, empty URLs, missing/incomplete bucket object, or bucket name only without credentials
401Invalid or missing API key
404Job not found
500Server error

Troubleshooting

SymptomCauseFix
Job stays "queued" for >2 min Workers may be busy or restarting Check GET /health. If workers: 0, contact support.
400 "No bucket configured" bucket missing from the request body Include a full bucket object (name, endpoint, access_key, secret_key, region) in every POST /jobs call.
400 "bucket requires name, endpoint..." bucket object is incomplete All five bucket fields are required. A bucket name string alone is not enough unless support has pre-configured server-side credentials for your account.
Video status "failed" Video unavailable, private, age-restricted, region-locked, or upload to your bucket failed Check the error field in results. Try a different video. Confirm your bucket token allows PutObject and the endpoint/region match your provider.
422 response Missing required header or malformed request body Ensure X-API-Key header is present and body is valid JSON with type, urls, and bucket.
Job not completed after 5 min Long video or download retry in progress Keep polling. If still not done after 10 min, contact support.

API & Enterprise

Get Access

API keys are issued by email. Bring your own S3-compatible bucket: results land in your storage and we don't keep copies on our side.

Request API access

Results delivered to your S3/R2 bucket. We never store your content.