Turn Any YouTube Channel Into a Searchable AI Knowledge Base
YouTube channels are dense information archives. A university lecture series, a technical conference, a prolific educator with hundreds of videos — the content is there, but it's locked in a format you can only consume sequentially. You can't search across a hundred videos for every time a speaker discussed a specific concept, compare how their thinking evolved on a topic over three years, or ask an AI assistant a question and have it retrieve the relevant moment from a 6-hour lecture.
A semantic knowledge base solves this. The workflow: extract transcripts from the channel's videos, convert them to embeddings, store them in a vector database, and build a search interface that retrieves the most relevant passages for any natural language query.
Why YouTube Is a High-Value Source for Knowledge Bases
YouTube hosts content that doesn't exist anywhere else in accessible form. Conference keynotes with no published proceedings, interview series where experts share knowledge they never wrote down, educational courses from institutions that don't publish transcripts. The transcript is the text layer that makes this content processable.
The value of a knowledge base over individual video search is precision. YouTube's own search finds videos whose titles or descriptions match your query. A semantic knowledge base finds the specific 90-second passage in a 3-hour lecture where a concept was explained — and links you directly to that timestamp. The granularity is different by an order of magnitude.
Step 1: Build the Playlist
INDXR.AI accepts playlist URLs, not channel URLs directly. There are two practical paths:
Path A: Use an existing playlist. Many channels organize content into playlists by series or topic. If the channel has thematic playlists — a lecture series, a conference year, a topic archive — use those URLs directly.
Path B: Create your own playlist. Log into YouTube, open the channel, and use the "Save to playlist" option on each video. You can create a playlist of up to 500 videos from any public channel. For channels with hundreds of videos, batch by year or topic rather than extracting everything at once — this makes the knowledge base easier to update incrementally.
Step 2: Extract Transcripts with INDXR.AI
Paste the playlist URL into INDXR.AI's Playlist tab. The pre-extraction scan shows every video's duration and whether you've already processed it. For a channel knowledge base, AI Transcription produces punctuated, accurately capitalized text — the quality difference matters when these transcripts become your retrieval corpus.
Credit cost for a typical knowledge base project:
A channel with 50 videos averaging 30 minutes each, all processed with AI Transcription:
- 50 videos × 30 minutes × 1 credit/minute = 1,500 credits
- At Plus pricing (€0.025/credit): €37.50
- RAG JSON export: 1,500 minutes ÷ 10 minutes per credit = 150 more credits
- Total: 1,650 credits = ~€41.25
This is a one-time extraction cost. The knowledge base persists indefinitely; adding new videos costs only the per-video transcription.
Step 3: Understand the RAG JSON Output
Each video's RAG JSON file contains 60-second chunks (by default) with everything a vector database needs:
{
"metadata": {
"video_id": "kBdfcR-8hEY",
"title": "Justice: What's The Right Thing To Do? Episode 01 …",
"duration_seconds": 3282,
"extracted_at": "2026-08-27T12:43:09.393Z",
"chunking_config": {
"chunk_size_seconds": 60,
"overlap_seconds": 9,
"overlap_strategy": "segment_boundary",
"total_chunks": 60
},
"channel": "Harvard University",
"language": "en",
"extraction_method": "youtube_captions"
},
"chunks": [
{
"chunk_index": 0,
"chunk_id": "kBdfcR-8hEY_chunk_000",
"text": "This is a course about Justice and we begin with a story suppose you're the driver of a trolley car, and your trolley car is hurdling down the track at sixty miles an hour …",
"start_time": 4.2,
"end_time": 65.08,
"deep_link": "https://youtu.be/kBdfcR-8hEY?t=4",
"token_count_estimate": 128,
"metadata": {
"video_id": "kBdfcR-8hEY",
"title": "Justice: What's The Right Thing To Do? Episode 01 …",
"channel": "Harvard University",
"chunk_index": 0,
"start_time": 4.2,
"end_time": 65.08,
"language": "en",
"total_chunks": 60
}
}
]
}The deep_link field is the key feature for a knowledge base: when your AI system retrieves a chunk and uses it to answer a question, it can cite the exact video and timestamp rather than just the video title.
Step 4: Build the Vector Index
The following example uses ChromaDB (local, no infrastructure required) and OpenAI embeddings. The same pattern works with Pinecone, Weaviate, or Qdrant for production deployments.
import json
import glob
import chromadb
from openai import OpenAI
client = OpenAI()
chroma_client = chromadb.PersistentClient(path="./knowledge_base")
collection = chroma_client.get_or_create_collection(
name="youtube_channel",
metadata={"hnsw:space": "cosine"}
)
def embed_texts(texts):
response = client.embeddings.create(
input=texts,
model="text-embedding-3-small"
)
return [item.embedding for item in response.data]
# Process all RAG JSON files
for filepath in glob.glob("./transcripts/*.json"):
with open(filepath) as f:
data = json.load(f)
chunks = data["chunks"]
if not chunks:
continue
texts = [chunk["text"] for chunk in chunks]
ids = [chunk["chunk_id"] for chunk in chunks]
metadatas = [chunk["metadata"] for chunk in chunks]
# Add deep_link to metadata for citation
for i, chunk in enumerate(chunks):
metadatas[i]["deep_link"] = chunk["deep_link"]
# Embed in batches of 100
for i in range(0, len(texts), 100):
batch_texts = texts[i:i+100]
batch_embeddings = embed_texts(batch_texts)
collection.add(
embeddings=batch_embeddings,
documents=batch_texts,
metadatas=metadatas[i:i+100],
ids=ids[i:i+100]
)
print(f"Indexed {len(chunks)} chunks from: {data['video']['title']}")
print(f"Total chunks indexed: {collection.count()}")Step 5: Query with Natural Language
def search_knowledge_base(query, n_results=5):
query_embedding = client.embeddings.create(
input=[query],
model="text-embedding-3-small"
).data[0].embedding
results = collection.query(
query_embeddings=[query_embedding],
n_results=n_results
)
print(f"\nQuery: {query}\n")
for i, (doc, metadata) in enumerate(zip(
results["documents"][0],
results["metadatas"][0]
)):
print(f"Result {i+1}:")
print(f" Video: {metadata['title']}")
print(f" Link: {metadata['deep_link']}")
print(f" Text: {doc[:200]}...")
print()
search_knowledge_base("trolley problem and utilitarian ethics")
search_knowledge_base("how does Rawls define justice")Step 6: Add an LLM Response Layer
def answer_question(question, n_chunks=4):
query_embedding = client.embeddings.create(
input=[question],
model="text-embedding-3-small"
).data[0].embedding
results = collection.query(
query_embeddings=[query_embedding],
n_results=n_chunks
)
context_parts = []
for doc, metadata in zip(results["documents"][0], results["metadatas"][0]):
source = f"{metadata['title']}"
context_parts.append(f"[Source: {source}]\n{doc}")
context = "\n\n".join(context_parts)
response = client.chat.completions.create(
model="gpt-4o-mini",
messages=[
{
"role": "system",
"content": "Answer questions based on the provided transcript excerpts. Always cite the video title and timestamp for claims you make."
},
{"role": "user", "content": f"Context:\n{context}\n\nQuestion: {question}"}
]
)
return response.choices[0].message.content
answer = answer_question("Explain the trolley problem and its implications for moral philosophy")
print(answer)Keeping the Knowledge Base Current
For channels that publish regularly, INDXR.AI's duplicate detection means re-running a playlist extraction only processes new videos — existing ones are skipped and not charged. A simple update workflow: extract the channel playlist weekly or monthly, new transcripts appear automatically and are ready to embed.
For the full chunking research behind the 60-second default chunk size, see How to Chunk YouTube Transcripts for RAG. For credit packages, see pricing.