Sushil Mishra, REALTOR® | Founder, The Real Tech | Last updated: July 2026
If you can upload a document, you can now make a video podcast. This guide walks through a 6-tool workflow that converts your notes into a fully edited, two-host video podcast — with an AI version of your own face and voice as one of the hosts. Total hands-on time: about 90 minutes per episode.
Most Ontario Realtors know they should be putting out video content. Fewer have the time, the co-host, or the camera confidence to actually do it. This workflow solves all three at once — and once it's set up, you can produce an episode a week without ever turning on a camera.
Sushil Mishra | Founder, The Real Tech | REALTOR® , Catch The Key Inc. www.CatchTheKey.Ca
How do you turn written notes into an AI podcast? (NotebookLM)
Google's NotebookLM is the starting point. Upload your source material — market notes, listing details, articles, anything — and it generates a genuinely natural-sounding conversation between two AI hosts, complete with banter and back-and-forth.
Here's exactly how:
Create a new notebook and upload your documents, links, or notes.
Open the Notebook Guide in the bottom-right corner.
Generate an Audio Overview (Deep Dive Podcast). NotebookLM builds a two-host discussion — one male voice, one female voice — based entirely on what you gave it.
Download the finished
.wavor.mp3file once it's done.
This step alone gives you a podcast-ready script and audio track, without writing a single word of dialogue yourself.
How do you separate two AI voices from one audio file?
NotebookLM hands you one combined track with both hosts talking. Before you can put your own face and voice on one of them, you need the two voices split apart. A speaker-isolation tool (SpeakerSplit or a similar audio separation tool) does this automatically:
Upload the NotebookLM audio file.
Let the tool separate Host A (the voice you'll replace with your clone) from Host B (the AI co-host you'll keep).
Download both isolated tracks.
You now have two clean, individual audio files instead of one blended conversation.
How do you clone your own voice for a podcast? (ElevenLabs)
This is where it becomes your podcast, not a generic AI one. ElevenLabs lets you build a voice clone from a short recording and then have it "speak" someone else's script — including Host A's isolated NotebookLM track.
In the Voice Lab, upload a clear 1–2 minute recording of your own voice to create an instant or professional voice clone.
Go to the Dubbing Studio or standard voice generation tab.
Upload Host A's isolated track and map it to your new voice clone.
Generate and download the result. Host A now says exactly what the original AI host said — but in your actual voice.
If you'd rather try an alternative, fish.audio does emotionally expressive voice cloning too and is worth comparing on cost and quality.
How do you get an AI video avatar that looks like you? (Syllaby)
Now the audio needs a face. Syllaby turns a still photo (or a pre-built digital twin) into a talking avatar synced to your uploaded audio.
Upload a high-quality photo of yourself, or use a pre-made digital twin, for Host A's avatar.
Choose a matching AI avatar for Host B.
Upload your cloned voice track (from Step 3) to animate Host A.
Upload the isolated Host B track (from Step 2) to animate Host B.
Render and download both video files separately.
You now have two individual talking-head videos — yours, and your AI co-host's.
How do you assemble two separate avatar videos into one podcast?
Line up the two rendered clips so their timing matches the conversation flow. Whether you arrange them side-by-side (split screen) or back-and-forth (whoever's talking is on screen), the goal is simple: the timelines need to sync so the "conversation" reads naturally, with no dead air or talk-over gaps.
This is a manual assembly step — no tool required yet, just a rough sequencing pass before final editing.
What's the best free tool to finish and polish the podcast video? (CapCut)
CapCut (desktop or web, free) is where the raw assembly becomes a publishable video:
Import both clips into a split-screen or alternating podcast layout.
Turn on Auto Captions for dynamic, engaging subtitles.
Cut dead air and awkward silences between exchanges.
Add light b-roll, a subtle studio backdrop, or background elements.
Export at 9:16 for Reels/Shorts/TikTok, or 16:9 for YouTube.
That's a complete episode — script, two-voice audio, two-host video, captions, and export — without a single hour spent in a recording studio.
The Real Tech Notes-to-Podcast Framework (at a glance)
Step | Tool | What it does | Rough time |
|---|---|---|---|
1. Script + audio | NotebookLM | Turns notes into a two-host discussion | 15 min |
2. Voice separation | SpeakerSplit (or similar) | Splits the two hosts into separate tracks | 10 min |
3. Voice clone | ElevenLabs | Puts your voice on Host A's lines | 15 min |
4. Avatar video | Syllaby | Animates two talking-head avatars | 20 min |
5. Assembly | Manual sync | Lines up both videos on one timeline | 10 min |
6. Final polish | CapCut | Captions, cuts, b-roll, export | 20 min |
Total: roughly 90 minutes per episode, most of it waiting on renders rather than active editing.
Want to go further? Three tools to extend the workflow
Once you have a finished episode, three more tools make it work harder for you:
Buzzsprout — host the audio-only version as a proper podcast feed, distributed to Spotify and Apple Podcasts, so this becomes a real weekly show, not just social clips.
repurpose.io — automatically reformats and reposts your finished episode across YouTube, Instagram, and TikTok, so one recording session covers a week of content.
VidiQ — if YouTube is part of your distribution, VidiQ helps with titles, tags, and thumbnails so the episode actually gets found.
None of these are required to finish an episode — they're for agents ready to turn one video into an ongoing content system.
FAQ
Do I need any video or editing experience to do this? No. Every tool in this workflow is upload-and-generate. The only manual step is lining up two video files in Step 5, which takes about 10 minutes.
Can I use my real voice instead of NotebookLM's AI hosts? Yes — you can skip the cloning step and record yourself directly if you prefer. The voice-clone approach is for agents who want a polished, consistent host voice without recording every episode from scratch.
How much does this whole workflow cost? Each tool has a free or low-cost starting tier (NotebookLM is free, CapCut is free). ElevenLabs and Syllaby have paid plans once you go past their free usage limits — budget roughly $20–$50/month combined if you're producing weekly.
What should my source notes actually contain? Anything specific: a market update, a neighbourhood breakdown, a client FAQ list, or even your own listing notes. The more specific the source, the more specific — and more citable — the generated conversation will be.
Can I do this for content in other languages? Yes, most of these tools (NotebookLM, ElevenLabs, Syllaby) support multiple languages, which is useful if part of your audience prefers Spanish, purtuguese, Hindi, Punjabi, or Urdu content.
Want this done for you instead?
If this looks useful but you'd rather not spend three evenings testing six tools yourself - that's exactly the kind of thing I help Ontario Realtors with. I can set up this entire workflow for your brand, your voice, and your listings, or just build you one finished episode to see what it looks like.
Reach out anytime:
Email: [email protected]
Book a call: 30 Minutes Meeting
Related reading from The Real Tech: NotebookLM: The Free AI Tool Ontario Agents Are Sleeping On and How Real Estate Agents Use AI to Remember Every Client Detail.
Sushil Mishra is a licensed REALTOR® with Catch The Key Inc. / REMAX West Realty Inc., based in Etobicoke, and founder of The Real Tech — an AI and PropTech education platform for Ontario real estate professionals. He holds a Google AI Professional Certificate and the Gemini Certified Educator designation.