Spoken
Spoken is a podcast transcript API that instantly converts any podcast episode into clean, speaker-labeled Markdown. It eliminates the need for a complex transcription pipeline by combining audio downloading, format conversion, speech-to-text processing, and speaker diarization into a single API call. The API resolves speakers to their real names by analyzing context within the transcript, avoiding generic labels like 'Speaker 1' and 'Speaker 2'. The output is plain Markdown with bold speaker names and per-turn timestamps, optimized to fit within typical LLM context windows (8,000–15,000 tokens per hour-long episode). This makes transcripts immediately usable for AI agents, summarizers, RAG pipelines, and other podcast tools. Spoken is designed for developers who want to integrate podcast content into their applications without managing infrastructure. It offers a simple REST API: search for an episode by text or URL, then fetch the transcript with a single GET request. The service supports any podcast episode from Spotify, YouTube, or other podcast apps. The response includes credit consumption headers, and credits never expire. A demo API key is provided for evaluation without signup. Beyond basic HTTP calls, Spoken can be integrated into AI agents via npx skills, OpenAPI specs, or an MCP server, making it easy to add podcast knowledge to Claude Code, Cursor, Windsurf, and other agent environments. Pricing is usage-based: $15 for 100 transcripts, $50 for 500, and $160 for 2,000, with no subscription or overage charges. Each transcript costs roughly a dime, replacing what the company estimates as a dollar’s worth of typical transcription infrastructure. Errors are not charged, and credits are debited only when a transcript is successfully delivered. The service has processed over 2,700 transcripts as of May 2026, primarily by developers building podcast summarizers, knowledge bases, and search tools.
- Category
- Dev Tools
- Pricing
- Paid
Problem solved
Developers building AI agents, podcast search tools, or content summarizers face a significant obstacle: podcast episodes are locked in audio format, making their insights inaccessible to text-based systems. Manually setting up a transcription pipeline requires downloading audio, converting formats, running speech-to-text models like Whisper, bolting on speaker diarization, chunking for LLMs, and managing storage — a time‑consuming infrastructure burden that distracts from product development. Spoken solves this by providing a single API call that returns any podcast episode as clean, speaker‑labeled Markdown, ready to drop into an LLM or knowledge base, so developers can skip the pipeline and ship their product faster.
Use cases
- Building a podcast summarizer that automatically generates key takeaways from episodes
- Adding every episode of a favorite podcast to a personal knowledge base for later retrieval and chat with an AI assistant
- Creating a RAG pipeline that indexes podcast transcripts for semantic search
- Integrating podcast content into an AI agent so it can reference specific episodes during conversations
- Developing a tool that monitors new episodes of subscribed podcasts and feeds transcripts into a daily briefing
- Automating the extraction of quotes, topics, or guest information from a large back‑catalogue of episodes
Audience
- Developers
- ML Engineers
- Startups
Pros
- One API call replaces an entire transcription pipeline (downloading, diarization, formatting)
- Clean Markdown output with real speaker names, not generic labels like ‘Speaker 1’
- No infrastructure to manage — works via simple GET requests with instant key delivery
- Rich agent integration options: npx skills, OpenAPI spec, MCP server, or plain HTTP calls
- Transparent, pay‑per‑use pricing with no subscription and credits that never expire
Cons
- Speaker name resolution may fall back to ‘Host’ or ‘Guest’ when real names cannot be determined from context
- Dependency on a third‑party API introduces external availability and rate‑limit risks
- Only supports podcast episodes, not arbitrary audio files or live streams
Official website: https://spoken.md/
Canonical LambdaBase page: https://www.lambdabase.com/tools/spoken