Update to 2.25.0
Update from PodNotes 2.24.0 to 2.25.0 to transcribe audio with ElevenLabs, Deepgram, OpenRouter, OpenAI, or an OpenAI-compatible server, and get clickable timestamps in audio transcripts. If you are updating from 2.21.1 or earlier, read Update to 2.22.0 first.
Why OpenAI transcription changed
Section titled “Why OpenAI transcription changed”OpenAI shuts down whisper-1 and gpt-4o-transcribe-diarize on 2027-02-26. PodNotes 2.25.0 moves OpenAI transcription to GPT Transcribe (gpt-transcribe).
GPT Transcribe returns plain text only, without timestamps and without speaker labels. ElevenLabs, Deepgram, and three OpenRouter models (Grok STT 2.0, ElevenLabs Scribe v2, and MAI-Transcribe 2) return both clickable timestamps and speaker labels. OpenRouter’s Voxtral Mini Transcribe 3 and Qwen3 ASR 1.7B, and an OpenAI-compatible server that returns word or segment times, add clickable timestamps without speaker labels.
How your 2.24.0 settings migrate
Section titled “How your 2.24.0 settings migrate”PodNotes maps your 2.24.0 transcript settings automatically on first load:
| Your 2.24.0 setup | Transcription provider in 2.25.0 | Model | Speaker labels | What changes |
|---|---|---|---|---|
| Speaker diarization on, Deepgram | Deepgram | Recommended (Nova-3) | On | Keeps using your Deepgram key. Audio transcripts now include clickable timestamps, and speakers are numbered 1, 2, 3 in the order they first speak. |
| Speaker diarization on, OpenAI (gpt-4o-transcribe-diarize) | OpenAI | Recommended (GPT Transcribe) | On | Keeps using your OpenAI key, but OpenAI no longer labels speakers. Each transcript saves as plain paragraphs and shows a notice naming the providers that label speakers. |
| Speaker diarization off, OpenAI key linked | OpenAI | Recommended (GPT Transcribe) | Off | Keeps using your OpenAI key with GPT Transcribe in place of whisper-1. Transcripts remain plain paragraphs without timestamps or speaker labels. |
| Speaker diarization off, no OpenAI key linked | ElevenLabs | Recommended (Scribe v3) | Off | Ready for ElevenLabs if you decide to add a key later. Turn on Speaker labels under Transcripts when you want speaker numbers. Feed transcripts still save for free without a key. |
Transcript path, Transcript template, and Speaker label format keep the values you set in 2.24.0. In settings under Transcripts, Speaker diarization is renamed to Speaker labels, and Diarization provider is replaced by Transcription provider and Model (plus Server address when you select OpenAI-compatible server).
Get speaker labels and timestamps after using OpenAI
Section titled “Get speaker labels and timestamps after using OpenAI”If you used OpenAI for speaker labels in 2.24.0, set up ElevenLabs, Deepgram, or a speaker-labeled OpenRouter model in Choose a transcription provider, or connect a local or hosted OpenAI-compatible server if you want timestamps without speaker labels. For speaker labels, check that Speaker labels is on:
- In Obsidian, open Settings → PodNotes.
- Under Transcripts, turn on Speaker labels if it is off.
If you prefer to stay on OpenAI (or use OpenAI-compatible server, Voxtral Mini Transcribe 3, or Qwen3 ASR 1.7B) without speaker labels, turn off Speaker labels under Transcripts so PodNotes stops showing the missing-labels notice after each transcription.
Devices still on 2.24.0
Section titled “Devices still on 2.24.0”PodNotes 2.25.0 keeps the same data.json storage format as 2.24.0 and preserves the 2.24.0 diarization fields alongside the new provider settings:
- A phone or second computer still running 2.24.0 keeps syncing your library and keeps using its 2.24.0 transcription setup without overwriting your 2.25.0 provider choice.
- Importing a 2.24.0 settings file carries over Deepgram when that file had Speaker diarization on with Deepgram, and otherwise keeps your library’s current Transcription provider.
- Update any 2.24.0 device that uses OpenAI before
2027-02-26, when OpenAI turns offwhisper-1andgpt-4o-transcribe-diarize.