AssemblyAI Prompts

In your Python code, transcribe the podcast episode at [AUDIO_URL] with AssemblyAI's Universal-3.5 Pro model, enabling speaker diarization and the summarization add-on. Then send a LeMUR prompt: list every chapter of this episode with a timestamp, a one sentence description, and the speaker who introduced it. Save the chapter list to [OUTPUT_FILE] in Markdown format for my show notes.

Set up a real-time streaming pipeline over WebSocket using AssemblyAI's Universal-3.6 Pro Realtime model for [CALL_CENTER_NAME]. Enable the sentiment analysis add-on to score every agent utterance and the PII redaction add-on to mask [PII_TYPE] before storage. Push any call whose average sentiment drops below [THRESHOLD_SCORE] to the supervisor dashboard in [DASHBOARD_TOOL].

Transcribe my bilingual support calls at [CALL_RECORDING_URL] with AssemblyAI's native code-switching support on the Universal-3.5 Pro model, which handles [LANGUAGE_PAIR] such as Hinglish out of the box. Request per sentence language labels, enable word level timestamps, and then ask LeMUR to summarize the customer's issue in English and list the action items for [TEAM_NAME].

Process the deposition recording at [AUDIO_URL] for the case [CASE_NAME] with Universal-3.5 Pro, turning on speaker diarization to separate the attorney, witness, and court reporter. Request verbatim transcription with word level timestamps and punctuation, then export the result as [EXPORT_FORMAT] for filing. Flag this job for HIPAA style audit logging since it contains [SENSITIVE_CONTENT_TYPE].

Bulk transcribe my archive of [NUMBER_OF_VIDEOS] webinar recordings in [SOURCE_FOLDER] using the budget friendly Universal-2 model, which covers 99 plus languages at $0.15 per hour. Add the topic detection add-on to tag each video with topics like [TOPIC_A] and [TOPIC_B], then build a searchable index in [DATABASE_NAME] so my team can find any mention of [SEARCH_KEYWORD] in seconds.

Build a voice agent for [BUSINESS_NAME] using AssemblyAI's Voice Agent API at the flat $4.50 per hour rate. Configure it to answer order status questions by looking up [ORDER_SYSTEM], set the voice to [VOICE_NAME], and define the greeting: welcome the caller, ask for their order number, then read back the status. Log every conversation to [LOG_DESTINATION] with full transcripts.

Transcribe the interview at [AUDIO_URL] with the Universal-3.5 Pro model and boost recognition of my product names by passing a keyterms_prompt list: [TERM_ONE], [TERM_TWO], [TERM_THREE]. Enable automatic punctuation and formatting, then run LeMUR with this question: extract every direct quote from [SPEAKER_NAME] about [TOPIC], keeping each quote under 30 words for my article.

Screen user uploaded audio at [UPLOAD_URL] through AssemblyAI's content moderation add-on before publishing on [PLATFORM_NAME]. Set the severity threshold to [SEVERITY_LEVEL], combine it with the entity detection add-on to log brand mentions of [BRAND_NAME], and automatically quarantine any file flagged above the threshold into [QUARANTINE_FOLDER] for human review.

Generate accurate captions for my YouTube video [VIDEO_TITLE] by transcribing the audio file at [AUDIO_URL] with Universal-3.5 Pro and word level timestamps enabled. Export the result as an SRT file named [FILE_NAME], keeping each caption line under [MAX_CHARS] characters so they display well on mobile screens, then upload directly to the video.

Use AssemblyAI's LLM Gateway to run [LLM_MODEL_NAME] over the last [NUMBER_OF_CALLS] sales call transcripts stored in [STORAGE_NAME]. Ask: what are the top three objections prospects raised about [PRODUCT_NAME], and which objection appears most often in deals lost to [COMPETITOR_NAME]? Return the answer as a JSON object with objection names, counts, and one example quote per objection.