AssemblyAI transcription stuck in processing — how to fix?

AssemblyAI transcription stuck: (1) poll GET /v2/transcript/ID every 3-5 seconds until status = 'completed' or 'error' — do not assume it completed; (2) if status = 'error', the error field explains why (usually audio URL inaccessible, or corrupted audio); (3) most transcriptions complete in 30-90 seconds for files under 1 hour; (4) use webhooks for production to avoid polling: pass webhook_url in the transcript POST body.

AssemblyAI audio URL error — how to upload local files?

AssemblyAI needs a publicly accessible URL. Upload local files to AssemblyAI first: POST the file binary to https://api.assemblyai.com/v2/upload with the authorization header. The response returns an upload_url. Use that upload_url as the audio_url in your /v2/transcript POST. The upload_url is valid for 24 hours.

Free tier 4 min read

AssemblyAI Not Working?

API 401 (lowercase authorization header), audio URL not accessible, transcription stuck in processing, webhook not firing, or LeMUR not responding? Check live status and fix it fast.

AssemblyAI — live status

Updated every 5 minutes. Full history at prismix.dev/service/assemblyai.

Full status →

What's wrong? Diagnose fast

🔑

API 401 — non-standard auth header

AssemblyAI uses a lowercase header with no "Bearer": authorization: YOUR_KEY (not "Authorization: Bearer"). This trips up many developers. Python SDK: aai.settings.api_key = "YOUR_KEY". Generate key at assemblyai.com/app/account.

📡

Audio URL not accessible (400)

AssemblyAI fetches your audio from the URL you provide. Localhost URLs fail. Fix: upload your file to AssemblyAI first: POST to /v2/upload with the file binary (authorization header required). Response gives upload_url — use that in your transcript POST. Valid 24 hours.

⏳

Transcription stuck in queued or processing

Poll GET /v2/transcript/ID every 3-5s until status = "completed" or "error". "error" = check the error field for the message. Common causes: audio URL expired or restricted, unsupported format (use WAV/MP3/MP4/WebM/FLAC), or corrupted audio file.

🖥

Webhook not receiving calls

webhook_url must be a public HTTPS URL (not localhost). AssemblyAI sends a POST when transcription completes or fails. Must respond with HTTP 200 in under 10 seconds. Test your endpoint with webhook.site. If not firing, check the transcript status via polling as fallback.

🧠

LeMUR not responding or failing

LeMUR requires a completed transcript_id — you cannot call LeMUR on a pending transcript. Check transcript.status = "completed" before calling LeMUR. LeMUR max input: ~100k tokens from transcripts. If context is too long, use context_window=short in the request or split transcripts.

⏺

Realtime streaming issues

AssemblyAI Realtime uses WebSocket (wss://api.assemblyai.com/v2/realtime/ws?sample_rate=16000). Key in the URL query param: ?token=YOUR_KEY. Audio must be 16-bit PCM raw bytes, 16kHz sample rate. If connection drops: VPNs block WebSocket — test without VPN. Also: token expires after 24h.

AssemblyAI API quick reference

Upload local file + transcribe (curl)

# Step 1: Upload local file — authorization is LOWERCASE, no Bearer
curl -X POST "https://api.assemblyai.com/v2/upload" \
  -H "authorization: YOUR_API_KEY" \
  -H "content-type: application/octet-stream" \
  --data-binary @/path/to/audio.mp3
# Returns: {"upload_url":"https://cdn.assemblyai.com/upload/..."}

# Step 2: Start transcription
curl -X POST "https://api.assemblyai.com/v2/transcript" \
  -H "authorization: YOUR_API_KEY" \
  -H "content-type: application/json" \
  -d '{"audio_url": "https://cdn.assemblyai.com/upload/..."}'
# Returns: {"id":"TRANSCRIPT_ID","status":"queued"}

Poll for completion + get result

# Poll every 3-5 seconds
curl "https://api.assemblyai.com/v2/transcript/TRANSCRIPT_ID" \
  -H "authorization: YOUR_API_KEY"
# status: "queued" → "processing" → "completed" | "error"
# If "error": check the "error" field in the response

Python SDK (recommended for production)

import assemblyai as aai

aai.settings.api_key = "YOUR_API_KEY"

transcriber = aai.Transcriber()
transcript = transcriber.transcribe("https://public-url/audio.mp3")
# OR for local file:
transcript = transcriber.transcribe("/local/path/audio.mp3")

if transcript.status == aai.TranscriptStatus.error:
    print(transcript.error)
else:
    print(transcript.text)

Step-by-step fix

1

Check live AssemblyAI status

Visit prismix.dev/service/assemblyai. Also assemblyai.statuspage.io. Transcription, LeMUR, and Realtime can have independent status.
2

Fix API 401

AssemblyAI uses a non-standard header: authorization: YOUR_KEY (lowercase, no "Bearer" prefix). This differs from most APIs. Python SDK: aai.settings.api_key = "YOUR_KEY". Key at assemblyai.com/app/account.
3

Fix audio URL not accessible

Your audio_url must be publicly reachable. For local files: upload to AssemblyAI first with POST /v2/upload (binary body, authorization header). Use the returned upload_url as your audio_url. Upload URLs are valid 24 hours.
4

Fix transcription stuck

Poll GET /v2/transcript/ID every 3-5 seconds until status = "completed" or "error". For production: use webhook_url in the transcript POST to receive a callback instead of polling.
5

Fix webhook not firing

Webhook URL must be a public HTTPS endpoint. If developing locally, use ngrok or webhook.site to expose a local URL. Your endpoint must respond with HTTP 200 within 10 seconds. AssemblyAI sends: {"transcript_id": "...", "status": "completed"}.

🔔

Get alerted when AssemblyAI goes down

Star AssemblyAI on Prismix and get emailed the moment status changes. Free, no credit card.