AI-driven YouTube Video Summarizer
BibiGPTAI.toWatch
Blog Home

Podcast Summary Skill: BibiGPT Covers 9 Platforms in One Command (2026)

bibigpt-skill is the only Agent Skill supporting 9 podcast platforms — Spotify, Apple Podcasts, Xiaoyuzhou, Ximalaya, and more. One command generates timestamped structured summaries.

BibiGPT Team

Podcast Summary Skill: BibiGPT Covers 9 Platforms in One Command (2026)

Table of Contents

The core pain point for heavy podcast listeners isn't finding good shows — it's forgetting what you heard, juggling fragmented platforms, and inability to batch-process episodes. When you subscribe to shows across Spotify, Apple Podcasts, Xiaoyuzhou, and Ximalaya simultaneously, accumulating 10+ hours of audio per week, your brain simply cannot keep up. bibigpt-skill is the only Agent Skill in the ecosystem that supports 9 podcast platforms, turning any podcast link into a timestamped structured summary with a single bibi command — making your AI Agent a podcast knowledge manager.

For basic installation and setup, see the Claude Code Skills Guide: bibigpt-skill AI Video Summary.

Why Podcast Users Need a Cross-Platform Skill

AI Subtitle Extraction Preview

Let's build GPT: from scratch, in code, spelled out

Let's build GPT: from scratch, in code, spelled out

Andrej Karpathy walks through building a tiny GPT in PyTorch — tokenizer, attention, transformer block, training loop.

0:00Opens with ChatGPT demos and reminds the audience that under the hood it is a next-token predictor — nothing more.
1:30Sets up the agenda: tokenisation, bigram baseline, self-attention, transformer block, training loop, and a tour of how the toy model maps to the real one.
4:00Loads the tinyshakespeare corpus (~1MB of plain text) and inspects the first few hundred characters so the dataset feels concrete before any modelling starts.
8:00Builds simple `encode` / `decode` functions that map characters ↔ integers, contrasting with BPE used by production GPT.
11:00Splits the data 90/10 into train/val and explains why language models train on overlapping context windows rather than disjoint chunks.
14:00Implements `get_batch` to sample random offsets for input/target tensors of shape (B, T), which the rest of the lecture will reuse.
18:00Wraps `nn.Embedding` so each token id directly produces logits over the next token. Computes cross-entropy loss against the targets.
21:00Runs an autoregressive `generate` loop using `torch.multinomial`; the output is gibberish but proves the plumbing works.
24:00Trains for a few thousand steps with AdamW; loss drops from ~4.7 to ~2.5 — a useful baseline before adding any attention.
27:00Version 1: explicit Python `for` loops averaging previous timesteps — clear but slow.
31:00Version 2: replace the loop with a lower-triangular matrix multiplication so the same average runs in one tensor op.
35:00Version 3: replace the uniform weights with `softmax(masked scores)` — the exact operation a self-attention head will compute.
40:00Each token emits a query (“what am I looking for”) and a key (“what do I contain”). Their dot product becomes the affinity score.
44:00Scales the scores by `1/√d_k` to keep the variance under control before softmax — the famous scaled dot-product detail.
48:00Drops the head into the model; the loss improves further and generations start showing word-like clusters.
52:00Concatenates several smaller heads instead of one big head — the same compute, more expressive.
56:00Adds a position-wise feed-forward layer (Linear → ReLU → Linear) so each token can transform its representation in isolation.
1:01:00Wraps both inside a `Block` class — the canonical transformer block layout.
1:06:00Residual streams give gradients an unobstructed path back through the network — essential once depth grows past a few blocks.
1:10:00LayerNorm (the modern pre-norm variant) keeps activations well-conditioned and lets you train with larger learning rates.
1:15:00Reorganises the block into the standard `pre-norm` recipe — exactly what production GPT-style models use today.
1:20:00Bumps embedding dim, number of heads, and number of blocks; switches to GPU and adds dropout.
1:24:00Trains the bigger model for ~5,000 steps; validation loss drops noticeably and quality follows.
1:30:00Samples 500 tokens — the output reads like a passable, if nonsensical, Shakespearean monologue.
1:36:00Distinguishes encoder vs decoder transformers; what we built is decoder-only, which is the GPT family.
1:41:00Explains the OpenAI three-stage recipe: pretraining → supervised fine-tuning on conversations → reinforcement learning from human feedback.
1:47:00Closes by encouraging viewers to keep tinkering — the architecture is small enough to fit in a notebook, but the same building blocks scale to GPT-4.

Want to summarize your own videos?

BibiGPT supports YouTube, Bilibili, TikTok and 30+ platforms with one-click AI summaries

Try BibiGPT Free

The podcast ecosystem is defined by fragmentation. Chinese users rely on Xiaoyuzhou and Ximalaya, international users prefer Spotify and Apple Podcasts, and niche academic podcasts may only publish on ListenHub or Pod.link. It's extremely common for one person to use 3-5 podcast apps simultaneously.

Yet existing AI tools almost universally fall short:

  • Most AI summarization tools only handle YouTube videos
  • General-purpose Agents cannot parse podcast URLs or extract audio
  • Manually pasting links episode by episode is painfully slow

bibigpt-skill, as an AI Agent capability component, solves this at the infrastructure level: it gives Agents (like Claude Code or OpenClaw) the ability to "listen to podcasts" across 9 platforms — no tool-switching required.

The 9 Podcast Platforms bibigpt-skill Supports

PlatformDomainHighlight
Xiaoyuzhouxiaoyuzhoufm.comChina's largest independent podcast platform
Ximalayaximalaya.comChina's largest audio platform (audiobooks + podcasts)
Spotifyspotify.comWorld's largest streaming platform
Apple Podcastspodcasts.apple.comBuilt-in podcast platform for Apple ecosystem
NetEase Cloud Music Podcastmusic.163.comPodcast channel within NetEase Cloud Music
Pod.linkpod.linkPodcast aggregation and redirect platform
Podcastaddictpca.stMost popular podcast client on Android
Google Podcastspodcasts.google.comGoogle's podcast platform
ListenHublistenhub.aiAI-native podcast discovery platform

These 9 platforms cover 95%+ of global podcast listening scenarios. No other Agent Skill in the ecosystem can match this cross-platform coverage of both Chinese and international podcast platforms.

One Command: From Podcast Link to Structured Summary

After installing the BibiGPT desktop app, run npx skills add JimmyLv/bibigpt-skill to give your Agent podcast summarization capabilities.

# Xiaoyuzhou (China)
bibi https://www.xiaoyuzhoufm.com/episode/xxxxxxxxxx

# Ximalaya (China)
bibi https://www.ximalaya.com/sound/xxxxxxxxxx

# Spotify
bibi https://open.spotify.com/episode/xxxxxxxxxx

# Apple Podcasts
bibi https://podcasts.apple.com/podcast/xxxxxxxxxx

# NetEase Cloud Music
bibi https://music.163.com/program?id=xxxxxxxxxx

# Pod.link
bibi https://pod.link/episode/xxxxxxxxxx

# Podcastaddict
bibi https://pca.st/episode/xxxxxxxxxx

# ListenHub
bibi https://listenhub.ai/episode/xxxxxxxxxx

bibi CLI help interfacebibi CLI help interface

Output includes:

  1. Timestamped segment summaries — precise to the second, with jump-to links
  2. Key insight extraction — 3-5 core arguments distilled
  3. Full transcript text — searchable and quotable
  4. Structured Markdown — ready to import into Notion/Obsidian

BibiGPT serves 1M+ users with over 5M+ AI summaries across 30+ platforms, and podcasts are among the fastest-growing categories.

Use Case 1: Automated Daily Podcast Intake for Academic Researchers

User profile: PhD student in cognitive science, subscribing to 8 academic podcasts in English and Chinese.

Pain point: 15+ hours of weekly podcast content is impossible to listen through entirely, but missing a research-relevant discussion is costly.

bibigpt-skill solution:

Agent scheduled task (runs daily at midnight):
1. Scan subscription feeds for new episodes
2. Run bibi command on each new episode
3. Filter by keywords ("neuroplasticity" "cognitive load" "working memory")
4. Full-transcript matched episodes with highlighted sections
5. Push to Notion database with citation tags

Result: Every morning, 2-3 pre-filtered, high-relevance summaries await in Notion. What used to take 3 hours now takes 10 minutes.

For details on building this kind of automation pipeline, see the OpenClaw + bibigpt-skill Complete AI Video Guide (Pillar article).

Use Case 2: Cross-Platform Podcast Content Curation

User profile: Content creator running a podcast recommendation account covering both Chinese and English shows.

Pain point: Tracking trending episodes across Xiaoyuzhou (Chinese), Spotify (English), and ListenHub (AI niche) simultaneously, then writing reviews manually — extremely time-consuming.

bibigpt-skill solution:

# Process multiple platforms in one session
bibi https://www.xiaoyuzhoufm.com/episode/xxx  # Xiaoyuzhou trending
bibi https://open.spotify.com/episode/yyy      # Spotify new release
bibi https://listenhub.ai/episode/zzz          # ListenHub AI podcast

The Agent auto-generates for each episode:

  • 300-word summary (social media ready)
  • Key quote extraction (with timestamp citations)
  • Content scoring suggestions (based on topic heat and depth)

Result: Each recommendation used to take 2 hours (listen + write). Now the first draft is done in 15 minutes — an 8x efficiency gain in curation output.

Use Case 3: Structuring Commute Micro-Learning

User profile: Professional with a 90-minute daily commute, using podcasts for continuous education.

Pain point: Listens to hours of podcasts weekly, but "listen and forget" — knowledge never sticks.

bibigpt-skill solution:

The Agent processes your "to-summarize" queue simultaneously as you commute:

  1. Before listening: Agent pre-generates a summary and table of contents so you can listen with questions
  2. During listening: Voice-mark interesting timestamps
  3. After listening: Agent extracts marked sections into note cards

Podcast transcription feature screenshotPodcast transcription feature screenshot

Combined with BibiGPT's flashcard feature, commute listening transforms into reviewable, exportable structured knowledge assets.

How bibigpt-skill Differs from Other Agent Skills

Dimensionbibigpt-skillOther AI Summary Tools
Podcast platforms9 (Chinese + international)0-2
Agent Skill integrationNative (one-line install)Not supported / DIY
Timestamped summariesYesPartial
Structured outputMarkdown / JSONPlain text
Speech recognitionMulti-engine (multilingual, speaker diarization)Single engine
Video + podcast unified30+ platforms, one entry pointPodcast-only or video-only
User-scale validation1M+ users, 5M+ summariesUndisclosed

Core differentiator: bibigpt-skill isn't just a podcast tool — it's the foundational component that gives AI Agents complete audio-visual comprehension. Podcasts are 9 of its 30+ supported platforms; the same bibi command also handles Bilibili, YouTube, Douyin, and more.

For deep-dive video summarization workflows, see the AI Podcast Summary Workflow Guide.

FAQ

Q1: Does bibigpt-skill work with paywalled podcasts (e.g., Ximalaya VIP content)?

A: For content requiring account authentication, bibigpt-skill cannot access it directly. However, you can download the paid episode as a local MP3 file and process it with bibi /path/to/file.mp3 — local audio files have no platform restrictions.

Q2: Is speech recognition accuracy consistent across all 9 platforms?

A: The core recognition engine is unified (BibiGPT's speech recognition service), so accuracy depends on audio quality rather than platform. Mandarin and English achieve 95%+ accuracy, with Cantonese, Japanese, and Korean also supported. For multi-speaker podcasts, we recommend the ElevenLabs Scribe engine for speaker diarization.

Q3: How do I install bibigpt-skill?

A: Install the BibiGPT desktop app, then run npx skills add JimmyLv/bibigpt-skill. First-time use requires bibi auth check for authorization. See the Claude Code Skills Guide for step-by-step instructions.

Give your AI Agent full cross-platform podcast summarization now:

BibiGPT Team