Podlog

Spec sheet

Purpose

Podlog is a self-hosted web app for audio transcription, search, and metadata analysis.Read 9 Oct 2026 · github.com

Integrations

Podlog supports Telegram and email notifications and an optional Telegram bot for search, questions, transcripts, queue checks, and adding sources.Read 4 Oct 2026 · github.com

Privacy

The project describes itself as local-first, with no accounts, cloud, or telemetry, while offering optional Fireworks remote inference.Read 9 Oct 2026 · github.com

License

Podlog uses the O'Saasy License, which permits use and distribution but bars offering the software as a hosted service that directly competes with the licensor.Read 9 Oct 2026 · github.com

Maker

The repository credits the project to @brlauuu; the linked GitHub profile identifies the maker as Đorđe.Read 9 Oct 2026 · github.com

Analytics

The Meta-Analysis dashboard reports speaker minutes, word counts, and host-versus-guest talking-time differences across feeds.Read 4 Oct 2026 · github.com

Ask AI

Ask AI provides streamed, citation-backed answers from the transcript library, using local Ollama by default or optional Fireworks inference.Read 9 Oct 2026 · github.com

Audio input

It can ingest episodes from RSS feeds or accept audio uploads, including MP3, M4A, WAV, OGG, FLAC, OPUS, AAC, WMA, WEBM, and MP4 files.Read 9 Oct 2026 · github.com

Deployment limits

Published images support linux/amd64 only; other architectures require building from source.Read 9 Oct 2026 · github.com

Deployment requirements

Installation requires Docker Compose V2 and a Hugging Face access token; users must accept the pyannote model license for diarization.Read 4 Oct 2026 · github.com

Exports

Users can download search results or transcripts as Markdown or plain text, or use a print-friendly view.Read 9 Oct 2026 · github.com

Exports and playback

Users can play audio from transcript timestamps and export transcripts or search results as Markdown or plain text.Read 4 Oct 2026 · github.com

Hardware limits

The hardware guide lists 4-core x86-64 and 8 GB RAM as minimums, says GPU is not required, and notes processing is CPU-only and sequential.Read 4 Oct 2026 · github.com

Local model storage

The README says pulling all three local Ask AI models takes about 12 GB, and the first run downloads about 3 GB of Whisper and pyannote model weights.Read 9 Oct 2026 · github.com

Notifications

Podlog supports Telegram and email alerts for episode completion or failure, with optional daily or weekly digests.Read 9 Oct 2026 · github.com

Platform constraint

Published container images support linux/amd64; other architectures require building from source.Read 4 Oct 2026 · github.com

Requirements

The documented prerequisites include Docker with Compose V2, a free Hugging Face account and access token, and accepting the pyannote model license for diarization.Read 9 Oct 2026 · github.com

Search

Hybrid search combines full-text keyword search with pgvector semantic similarity.Read 9 Oct 2026 · github.com

Speaker labels

Local pyannote diarization assigns speaker labels, and the UI lets users rename or merge speakers.Read 9 Oct 2026 · github.com

Telegram bot

The optional Telegram bot can search, ask questions, fetch transcripts, check the queue, and add sources; it responds only to accounts on the configured allowlist.Read 9 Oct 2026 · github.com

Transcription

Speech-to-text uses WhisperX with large-v3-turbo by default, configurable down to tiny for low-RAM machines.Read 9 Oct 2026 · github.com

Transcription and speakers

It uses WhisperX for transcription and supports local speaker diarization, with an optional paid pyannote.ai cloud model.Read 4 Oct 2026 · github.com

Where Specifiction places it

• Podcast Transcription Software#22 of 257.3