Translating any podcast episode for about a dollar: the architecture, and the one rule that shaped it

Description and use

A workflow that translates podcast episodes with a transcript tag, using managed MT and TTS services, delivering a single‑voice audio file for about a dollar with a 93‑second turnaround and upfront cost quoting.

Problem

A user wants to share a non‑English podcast episode with a friend who doesn’t speak English, but a summary or transcript alone is insufficient for listening.

Outcome

The system delivers a translated podcast audio file for about a dollar, with a 93‑second turnaround, and guarantees no surprise costs by quoting the price before any paid API calls.

Workflow

• Detect <podcast:transcript> in RSS feed
• If missing, refuse the request
• Fetch the podcaster’s published transcript
• Translate the transcript per character using a managed MT service
• Synthesize the translated text per character with a managed neural TTS voice
• Stitch the TTS audio into a single playable file
• Store the file in the requester’s household library

Evidence

The obvious architecture is: transcribe the audio with speech‑to‑text, translate that, narrate the result. I deliberately did not do that...

Source

One episode of a (faintly ridiculous) British royal‑gossip podcast that happened to be in my test feed — 8,570 characters of transcript, English to German — landed at 90 cents.

Source

A 9‑minute episode, translated, narrated, and assembled into a playable file, in 93 seconds.

Source