{"title":"Exercise: Audio and Speech","description":null,"content":"# Exercise 1: Transcribe then Summarize\n\nBuild a two-stage audio pipeline: transcribe a spoken audio file to text, then summarize that transcript with a chat model. Speech in, summary out.\n\nMost audio applications are really two steps glued together: a transcription model turns sound into text, and a language model does something useful with the text. Keeping them separate means you can swap the transcription model or the summarizer independently, and you can inspect the transcript when the summary looks wrong.\n\n### Requirements\n\n- Transcribe an audio file with a transcription model.\n- Pass the transcript to a chat model and ask for a short summary.\n- Print both the transcript and the summary.\n- The audio file path is something you supply; the pipeline shape does not depend on the specific clip.\n\n\n**main.py**\n\n```python\nfrom openai import OpenAI\n\nclient = OpenAI() # Uses OPENAI_API_KEY from environment\n\n# Supply a path to an audio file (for example a short .mp3 or .wav)\naudio_path = \"sample.mp3\"\n\n# TODO: open the audio file and transcribe it with client.audio.transcriptions.create\n\n# TODO: pass the transcript text to a chat model and ask for a 1-2 sentence summary\n\n# TODO: print the transcript and the summary\n```\n\n\n### Expected Output\n\nThe transcript reflects the spoken words, and the summary condenses it.\n\n\n```plaintext\nTranscript: Welcome to the team meeting. Today we will review the quarterly roadmap and assign owners for each milestone.\nSummary: The meeting covers the quarterly roadmap and assigns an owner to each milestone.\n```\n\n\n### Solution\n\n\n**main.py**\n\n```python\nfrom openai import OpenAI\n\nclient = OpenAI()\n\naudio_path = \"sample.mp3\"\n\n# Stage 1: transcribe audio to text\nwith open(audio_path, \"rb\") as audio_file:\n transcription = client.audio.transcriptions.create(\n model=\"gpt-4o-transcribe\",\n file=audio_file,\n )\ntranscript = transcription.text\n\n# Stage 2: summarize the transcript\nresponse = client.chat.completions.create(\n model=\"gpt-5.4-mini\",\n messages=[\n {\"role\": \"system\", \"content\": \"Summarize the transcript in one or two sentences.\"},\n {\"role\": \"user\", \"content\": transcript},\n ],\n)\nsummary = response.choices[0].message.content\n\nprint(f\"Transcript: {transcript}\")\nprint(f\"Summary: {summary}\")\n```\n\n\nThe transcription model and the chat model do different jobs and are called independently, with the transcript as the hand-off between them. Opening the file in binary mode (`\"rb\"`) is what the transcription API expects. Because the stages are decoupled, you can upgrade transcription accuracy or change the summary style without touching the other half, which is the practical reason to build audio features as a pipeline rather than one opaque call.\n\n---\n\n# Exercise 2: Extract Action Items from a Meeting\n\nSummaries are useful, but structured output from audio is more so. Transcribe a meeting recording, then pull out the action items as a structured list instead of prose. This is the audio-to-data pipeline behind meeting assistants and call-center analytics.\n\n### Requirements\n\n- Transcribe the audio file to text.\n- Ask the model to extract action items as a JSON object with an `action_items` list.\n- Parse the JSON and print each action item.\n\n\n**main.py**\n\n```python\nfrom openai import OpenAI\nimport json\n\nclient = OpenAI() # Uses OPENAI_API_KEY from environment\n\naudio_path = \"meeting.mp3\"\n\n# TODO: transcribe the audio file\n# TODO: ask the model to extract action items from the transcript as a JSON list of strings\n# TODO: parse and print the action items\n```\n\n\n### Expected Output\n\nThe transcript is turned into a clean, numbered list of action items.\n\n\n```plaintext\n1. Alice to finalize the quarterly roadmap by Friday.\n2. Bob to assign owners to each milestone.\n3. Schedule a follow-up review next week.\n```\n\n\n### Solution\n\n\n**main.py**\n\n```python\nfrom openai import OpenAI\nimport json\n\nclient = OpenAI() # Uses OPENAI_API_KEY from environment\n\naudio_path = \"meeting.mp3\"\n\nwith open(audio_path, \"rb\") as audio_file:\n transcript = client.audio.transcriptions.create(model=\"gpt-4o-transcribe\", file=audio_file).text\n\nresponse = client.chat.completions.create(\n model=\"gpt-5.4-mini\",\n messages=[{\"role\": \"system\", \"content\": \"Extract the action items from the transcript as JSON: {\\\"action_items\\\": [\\\"...\\\"]}.\"},\n {\"role\": \"user\", \"content\": transcript}],\n response_format={\"type\": \"json_object\"},\n)\nitems = json.loads(response.choices[0].message.content)[\"action_items\"]\nfor i, item in enumerate(items, start=1):\n print(f\"{i}. {item}\")\n```\n\n\nChaining transcription into a structured-extraction call turns speech directly into data your application can act on, not just a paragraph someone has to read. Requesting JSON with `response_format` makes the action items a list you can store as tasks or send to a tracker. The two stages stay independent, so you can improve transcription accuracy or change what you extract without touching the other. It needs a real audio file and a transcription-capable model.","pageType":"ai-engineering"}

Get Premium