NotebookLM
Research in. Video out.
Web + iOS + Android. Video Overviews on mobile since Jan 2026.
notebooklm.google.comMake video people actually watch
This isn’t a trend or a tool stack. It’s a shift in what computing is, how we use it, and what we can make. Three lenses make everything else click.
Text me your work
Made something in class? Send it. 720-254-7893
Scan the QR or text me directly. I’ll play the good stuff on the big screen so the room sees what you made.
AI changed three things at once
AI as OS. The foundation changed. AI isn’t an app you open; it’s the substrate everything runs on. Your workflow, your editor, your camera, your voice. Build on it, or build against it.
AI as UX. The interface changed. You don’t hunt through menus anymore; you tell the system what you want. Underlord listens. HeyGen’s agent listens. NotebookLM listens. Natural language is the new mouse.
AI as creative multiplier. The output scale changed. One recording becomes twelve languages by lunch. One avatar becomes infinite videos. One person becomes many. AI doesn’t replace the creator. It multiplies them.
Where each tool lives inside the frame
Riverside. OS + UX. Podcast-grade infrastructure with a conversational studio.
Descript. UX. You talk to Underlord; it edits.
HeyGen. All three. Infrastructure, conversation, and you at scale.
NotebookLM + Nano Banana. Creative multiplier. Content at volume — videos from research, thumbnails in seconds.
The arc of today
Four outcomes, four tools.
Record. Broadcast-quality video anywhere, solo or with guests. (Riverside)
Edit. At the speed of thought, by asking an AI agent. (Descript)
Become. Scale yourself into every language and channel. (HeyGen)
Generate. Turn research into finished video. Thumbnails in seconds. No camera required. (NotebookLM + Nano Banana)
Everything today is live. No reels, no canned clips. Let’s build.
Stay in touch
Subscribe to The Talentless Hack
Steve’s Substack. Weekly AI build notes, experiments, and what’s worth paying attention to. This is also where future classes and the Talentless community will get announced first.
Record once. Use everywhere.
Every podcast, webinar, customer interview, and piece of thought leadership starts with a recording you don’t have to do twice. If you want to ship more video, you need a studio you can open with one click, anywhere, that doesn’t fall apart when your wifi does.
Before you press record: seven basics
Most AI video is bad because the basics are wrong. Get these right and everything else works.
1Good sound. Mic close to your mouth.
2Good lighting. Face the light.
3Good background. Clean or intentional.
4Walk. Move. Don’t be a statue.
5Shoot from above. Tilt the phone down.
6Smile. The mic hears it.
7Fix your hair. Self-check before you record.
iPhone settings for highest quality (set once, forget)
Open Settings → Camera on your phone. These flip your default capture from “fine” to “master file.”
Record Video. 4K at 60fps for smooth motion. 4K at 30fps if storage is tight.
HDR Video on. Dolby Vision HDR. Bigger files, way more dynamic range.
Formats → High Efficiency. HEVC saves space. Switch to “Most Compatible” only if your editor balks at HEVC.
Grid on. Rule of thirds, instantly. No excuses for centered talking heads.
Mirror Front Camera on. Selfies record un-flipped, the way people actually see you.
Record Stereo Sound on. Two mics beats one, even built-in.
Right before you tap record
Wipe the lens. Your shirt works. Smudges kill sharpness.
Rear camera when you can. Sharper sensor, better glass than the front.
Tap and hold the subject. Locks focus + exposure. The yellow “AE/AF Lock” banner is your friend.
Stable surface. Tripod, mount, or prop the phone on a stack of books. Never handheld for keepers.
Plug in a mic. AirPods Pro, a $30 lav, anything wired — all beat the built-in mic at distance.
Focus mode on. Notifications kill takes. Silence the phone.
Plug in power. 4K + HDR drains the battery. For long sessions, stay on the cable.
Don’t pinch-zoom. Walk closer instead. Digital zoom destroys quality.
The tool: Riverside
Records locally on every device.
Podcast-grade: separate track per guest, multiple participants.
Text-based editing right in Riverside. No timeline.
Auto-highlight detection. Riverside finds the clips.
Built-in teleprompter. Read your script while looking at the camera.
Full mobile app.
In-class exercise: Join my studio
Scan the QR or tap the link. We’re going to put the whole room on camera at once. A wall of faces, all recording locally on their own phones.
Steve’s Riverside studio
Tap to join the studio live
riverside.com/studio/steve-mudds-studio
Bring headphones if you have them. Trust me on this.
Settings for a room of 30 people
Headphones on if you have them. Breaks the feedback loop. Speakers + mics in the same room without headphones = echo chaos.
Mics muted by default. Steve unmutes one person at a time. Hosts: use “Mute all participants” + “Mute on entry.”
Recording: High Quality. Each phone captures its own track locally. Cross-talk in the room won’t pollute the recording.
Echo cancellation + noise suppression on. Riverside Studio Settings → Audio. Default but worth confirming.
Be an influencer for 30 seconds
Pick a hook. Open Riverside on your phone. Hit record. Run with it for 30 seconds on a topic you actually care about. Text the clip to Steve at 720-254-7893. We’ll play the best ones on the big screen.
Don’t overthink it. Confidence is the costume.
“Breaking news…”
“Stop scrolling.”
“You are not going to believe this.”
“POV:…”
“Hot take:…”
“Wait until the end.”
“Storytime.”
“I just figured out…”
“Three things nobody tells you about…”
“If you’re a [founder / marketer / parent], stop what you’re doing.”
“The thing nobody is talking about…”
“I tried this for 30 days. Here’s what happened.”
“PSA:…”
“Real talk.”
“If I had to start over today…”
“Here’s the truth about…”
Monday exercise
Record a 2-minute “about me” in Riverside mobile. Text-edit one filler word out. Post it.
Edit at the speed of thought
You don’t have time to scrub timelines anymore. The fastest edit is the one you don’t do with your hands. You just tell an AI agent what you want. Three hours of editing becomes three minutes. That changes how much you can ship.
Start in Riverside — the editing you already have
You already have your clip from the last section. Before we switch tools, here’s what Riverside can do without ever leaving it.
Text-based editing. Open the transcript, highlight a word, delete. The video cuts to match. No timeline required.
Auto-detected highlights. Riverside scans the recording and surfaces the clip-worthy moments. You pick the one that pops.
AI Producer. Set Pace removes long pauses. Smooth Speech kills filler words. One pass, done.
Short vertical clips. Export ready-to-post verticals for TikTok, Reels, and Shorts — straight from Riverside.
That covers 80% of the edits most people actually need. For the other 20% — Studio Sound, eye contact correction, chapters, translation, and multi-step automation — you graduate to Descript.
The tool: Descript + Underlord
Underlord is a natural-language AI agent inside your editor. You talk to it. All the old features still exist. You just don’t have to hunt for them.
Ask it anything, for example:
“Remove all the filler words.”
“Apply Studio Sound.”
“Add subtle background music.”
“Fix my eye contact.”
“Shorten all the word gaps.”
“Make a 60-second highlight reel.”
“Draft show notes, a LinkedIn post, and a blog post.”
“Translate this to Spanish.”
Monday exercise
Import last week’s Zoom recording. Ask Underlord: “Remove filler words, apply Studio Sound, and draft a 60-second highlight reel with a LinkedIn caption.” Post it.
Be in more places than you can fly
One person can only film so much. But your audience is everywhere: in every language, on every platform, and they want to hear from you personally. You can’t record a thousand versions of yourself. You can build one, and scale it.
Attendee-only discount
Use code
WUIM72OJ on HeyGen. $29 USD off, covers a full month of the Creator plan.The tool: HeyGen
Today’s technology sponsor. Four avatar types, lip sync, voice basics, and an AI Video Agent that ships video for you.
The four avatar types
APhoto Avatar. One photo → talking head. Fastest, easiest, lowest fidelity.
BInstant Avatar. Short mobile clip → motion avatar. What most people should make.
CStudio / Premium. 30+ min of clean footage → brand-grade avatar.
DCharacter / Mascot. Not a real person. Synthetic spokesperson.
Lip sync translation
One English clip → twelve languages by lunch. Accurate mouth movement, native voice. A single clip can travel further than a week of filming.
Voice basics
Voice cloning. Record a short clean sample; your cloned voice powers your avatar.
Background noise. Clean room + good mic during cloning = dramatically better avatar voice.
Voice mirroring. Match your avatar’s voice to yours, or to a brand voice you’ve set.
AI Video Agent: the closer
Tell your avatar what to make a video about. The agent writes the script, generates the delivery, ships the clip. You review and send.
Bonus: generating a Photo Avatar from a prompt
Don’t have a usable photo of yourself? HeyGen also lets you create a Photo Avatar from a generated image. The prompt is everything. Be specific about role, wardrobe, lighting, camera, expression.
Template[ROLE / ARCHETYPE] in their [AGE], [action to camera].
Wardrobe: [what they’re wearing, styling notes].
Setting: [background, props, depth of field].
Lighting: [direction, quality, mood].
Camera: [angle, framing, lens feel].
Expression: [emotion + micro-detail].
Style: [reference aesthetic — editorial, cinematic, documentary].
Avoid: [what NOT to produce — stock photography, uncanny plastic skin, stiff corporate posture].
Example: A confident podcast presenterA confident podcast presenter in their mid-30s, speaking directly to camera.
Wardrobe: dark charcoal sweater over a plain white t-shirt, minimal styling, one small piece of jewelry.
Setting: modern podcast studio, softly blurred, warm ambient light, a studio microphone faintly visible at the foreground edge.
Lighting: key light from upper left, soft rim light from the right, cinematic and flattering, shadows soft but defined.
Camera: eye-level, medium close-up, shallow depth of field, 50mm lens feel.
Expression: warm, engaged, a little wry, mouth slightly open as if mid-sentence.
Style: editorial photography, photoreal, high fidelity, natural skin texture.
Avoid: stock photography, plastic skin, stiff corporate posture, generic AI aesthetic.
Monday exercise
Take a selfie. Create a Photo Avatar. Type a 30-second welcome script. Ship it in two languages.
Be trusted
Synthetic video only works if the people watching trust you. Every shortcut you take with someone else’s face, voice, or likeness is debt, and it comes due. Five rules keep you on the right side of that line.
The five non-negotiables
1Never create an avatar without consent.
2Understand the rights for real people’s likenesses.
3Never distribute team avatars without explicit approval.
4Disclose synthetic content. Note approvals in the disclosure.
5Be culturally sensitive when casting avatars and voices.
The frame
Let yourself be creative. Be human.
Before you go: tell me how it went
Two-question survey
What’s one thing I can do differently? What do you want to learn next time?
Open the survey
Sixty seconds. Honest answers shape the next class.
Generate content at scale
Two creative multipliers. NotebookLM turns research into finished video with no camera required. Nano Banana turns any image into a thumbnail in seconds. One person can now produce what an agency would have charged for.
The tools: NotebookLM → Google Vids, and Nano Banana
Two sides of the same coin. NotebookLM makes the video. Nano Banana makes the thumbnail that gets it watched.
What NotebookLM actually does — in three moves
1A walled garden of information. Drop in your sources — PDFs, docs, URLs, transcripts, your own writing. NotebookLM only knows what you give it. No outside hallucination.
2The ability to ask questions. Query the garden in natural language. Every answer is grounded in your sources, with inline citations. It’s research with receipts.
3The ability to create. Turn the garden into outputs: briefing docs, study guides, audio overviews, and — the one we’re using today — Video Overviews.
Use case 1: Video from the ether
Cinematic. Drop sources into NotebookLM. Studio → Video Overview → cinematic style. Play the result.
Explainer, any language. Generate the same video in Spanish, Mandarin, whatever the audience speaks. Native voiceover, matched pacing.
Branded cinematic prompt
Default cinematic output is generic. Pass a style directive in the Customize field. Spell out voice, visual language, do’s and don’ts.
TemplateCreate a cinematic Video Overview for [BRAND / TOPIC].
Voice & tone: [3-5 words. bold, confident, warm, smart, irreverent.]
Visual language:
• Color palette: [hot pink + deep navy + cream]
• Imagery style: [editorial photography, neon urban scenes, film-grain]
• Typography feel: [bold sans-serif, whitespace]
• Avoid: stock photography, corporate clip art, pastel gradients, generic AI look.
Structure:
• Hook in the first 5 seconds.
• 3 main beats, one idea each.
• Close with one action the viewer can take today.
Voiceover: [warm, smart, slightly wry, senior creative director voice — not an announcer.]
Length: 60–90 seconds.
Example: Talentless AI cinematicCreate a cinematic Video Overview for Talentless AI, a studio where art meets artificial intelligence.
Voice & tone: confident, a little irreverent, warm, smart, Texan-punk.
Visual language:
• Color palette: hot pink, deep navy, cream
• Imagery style: high-contrast editorial photography, neon-lit urban scenes, film-grain
• Typography feel: heavy sans-serif, lots of whitespace
• Avoid: stock photography, corporate clip art, pastel gradients, generic AI aesthetics.
Structure:
• Hook: ask whether the audience is still making video the old way.
• Beat 1: record once, use everywhere.
• Beat 2: edit at the speed of thought.
• Beat 3: scale yourself into every language and channel.
• Close: invite the viewer to build something this week, not someday.
Voiceover: warm, smart, slightly wry, senior creative director voice, not an announcer.
Length: 75 seconds.
On-brand isn’t perfect-brand. NotebookLM lands in the neighborhood. For public-facing work, finish in Google Vids or hand to a designer. For internal and early drafts, it’s usable as-is.
NotebookLM — Presentation to Video
You already have a deck. NotebookLM → Google Slides → Google Vids, with AI voiceover. That deck becomes a narrated video in minutes.
Nano Banana — Thumbnails, fast
Nano Banana is Google’s image model inside Gemini. Feed it a still frame from your video, describe the thumbnail, iterate. Ten thumbnails in two minutes. Pick the one that pops.
Actual Nano Banana output. Talentless AI Masterclass thumbnail. Navy + orange, cinema camera + wireframe avatar, sticker-style hook.
Nano Banana thumbnail promptTake this photo of [me / my avatar / my guest] and turn it into a YouTube thumbnail.
Style: high-contrast editorial photograph, bold rim light, film-grain texture.
Background: [deep navy / hot pink / whatever matches your brand].
Expression: exaggerated. Surprised, laughing, pointing. Not neutral.
Text overlay: “[5-7 word hook]”. Heavy sans-serif, huge, in hot pink.
Frame: rule of thirds. Face on the left, text on the right.
Make three versions with different expressions and color treatments.
Monday exercise
Take 3–5 documents from a current project. Drop them into NotebookLM. Generate a Video Overview. Screenshot a strong frame. Open Nano Banana, generate three thumbnails using the prompt above. Ship the best one.
Google Vids
Turn slides into narrated video with AI voiceover.
Part of Google Workspace.
vids.google.comNano Banana
Thumbnails and image edits, fast.
Google Gemini image model. Web + iOS + Android. Iterate thumbnails in seconds.
gemini.google.comA studio where art meets artificial intelligence
Talentless AI helps brands, creators, and communicators turn AI from a buzzword into a working production stack. We work where culture, code, and storytelling collide. Austin-based. Built to ship work, not slide decks.
What I do
I’m Steve Mudd — founder of Talentless AI. Veteran of IBM Watson and Ogilvy. I help organizations transform their marketing using generative AI, synthetic media, and bold storytelling. I run classes like this one, advise brands on how to deploy AI without losing their voice, and host the AI After Hours podcast.
Speaker + class leader. Half-day masterclasses (like today), keynotes, in-house team workshops.
Strategic AI advisor. Synthetic media playbooks, AI marketing transformation, creative-tech audits.
Creative technologist. I build the pipelines I teach. If it’s on this page, I’ve shipped with it.
HeyGen Community Host. Years on the platform, plugged into the roadmap.
What we do
Three intersecting practices. Each one feeds the others.
1Platform Partnerships + AI. We install the systems, the thinking, and the rituals that scale brand content globally. Stack design, training, governance, brand-compliant pipelines.
2Brand Content Lab. We cocreate with bold brands and creators. Synthetic media, multilingual launches, AI-native campaigns, IP development.
3Channels & Originals. We make our own weird synthetic content. Unresolved Signals (AI-native UAP documentary podcast — see below). FutureKeepers (global sustainability platform with Danny Kennedy). Backlot74 (an AI micro-series studio system). AI Applied Live (multi-city event series).
Featured project: Unresolved Signals
A Talentless AI Production
UNRESOLVED
SIGNALS
An AI-powered documentary investigation into the global UAP record. We cross-reference government archives, military reports, and declassified documents from dozens of countries to trace the oldest open question in human history.
9
Episodes live
95
Primary sources cited
27
Countries in the record
This is what we ship with the same stack you just learned. AI-native documentary, end to end.
How we work
NOT a tools company. We use tools — we don’t sell them.
NOT a platform chasing volume. We ship work, not metrics.
AI as accelerator, not shortcut. We bridge the gap between “AI slop” and cinema.
Very. Very. Texas. Austin sandbox, global ambition.
Want to work together?
Email steve.mudd@talentless.ai, text 720-254-7893, or scan the Substack QR on the Welcome tab to stay close to what we’re building next.
The cheat sheet
Everything in one scrollable card. Screenshot this tab. Pin it. Use it on Monday.
The seven basics
1Good sound. Mic close.
2Good lighting. Face the light.
3Good background. Clean or intentional.
4Walk. Move. Don’t be a statue.
5Shoot from above. Tilt down.
6Smile. The mic hears it.
7Fix your hair. Self-check.
The tools in one line each
Record — Riverside. Studio in your pocket. Records locally. Text-edit in-app.
Edit — Descript + Underlord. Tell it what to do. “Remove filler. Apply Studio Sound. Add music.”
Become — HeyGen. Photo / Instant / Studio / Mascot avatars. Lip sync to any language. AI Video Agent. Code WUIM72OJ for $29 off.
Generate video — NotebookLM. Sources in. Cinematic or explainer video out. Any language.
Generate thumbnail — Nano Banana. Ten thumbnails in two minutes. Iterate expressions and colors.
Voice — ElevenLabs. Clone separately. Pipe into HeyGen for better fidelity than default TTS.
The five ethics rules
1Consent always.
2Know the rights of real likenesses.
3Approval before distribution.
4Disclose AI content.
5Cultural sensitivity.
The go-to prompts
Descript, every clip: “Remove all the filler words, apply Studio Sound, and shorten the word gaps.”
NotebookLM, on-brand cinematic: “Create a cinematic Video Overview for [BRAND]. Voice: [tone]. Colors: [palette]. Avoid: stock.”
HeyGen avatar from scratch: “[Role] in their [age]. Wardrobe. Lighting. Camera. Expression. Avoid stock + plastic skin.”
Nano Banana thumbnail: “Make a YouTube thumbnail. Bold color, high contrast, text overlay, 3 variants.”
The full toolkit — every QR you need
Scan or tap to install / open each tool.
Riverside
Record broadcast-quality video, solo or with guests.
Web + iOS + Android. Records locally on every device.
riverside.fmDescript
Edit by asking. Underlord is the AI agent.
Desktop / laptop / browser. No mobile app.
www.descript.comNotebookLM
Research in. Video out.
Web + iOS + Android. Video Overviews on mobile since Jan 2026.
notebooklm.google.comGoogle Vids
Turn slides into narrated video with AI voiceover.
Part of Google Workspace.
vids.google.comHeyGen
Scale yourself. Avatars, lip sync, AI video agent.
Web + iOS + Android. Instant Avatar from a short mobile clip. Use code WUIM72OJ for $29 off.
www.heygen.com/Nano Banana
Thumbnails and image edits, fast.
Google Gemini image model. Web + iOS + Android. Iterate thumbnails in seconds.
gemini.google.com#ProTips worth keeping
Do better. Lighting and framing matter.
Don’t worry about being perfect. You won’t be.
Record locally, always. Wifi dies; your file doesn’t.
Mic first. Camera second. Light third.
If you can talk, you can edit.
Studio Sound is magic. Use it on every clip.
Research to finished video in the time it takes to read your sources.
Old slides aren’t dead. They’re source material for your next video.
Smile when you clone. The mic picks up emotional nuance.
Write natively in the target language with AI.
For video cloning, be sure you like your look.
Clone your voice separately.
Exaggerate your performance. Smile.
Let yourself be creative. Be human.