Lesson 1 got you a niche. This lesson gets you a video.
The bottleneck for most faceless creators isn't motivation — it's the production gap. They know what niche to target, they know the algorithm wants volume, but they're staring at a blank project file not knowing which tool to open first.
This lesson closes that gap. You'll walk out with a complete production pipeline: four phases, five tools, and a workflow that takes you from blank document to uploaded video in under 60 minutes. Not 3 hours. Not a weekend project. 60 minutes.
This is the exact workflow used for AI Tools and Finance niches — both high-volume, high-RPM categories. If you're in a different niche, the tools stay the same. The script prompts change slightly, but the pipeline structure is identical.
Before diving into each phase, here's the full timeline. Knowing the structure upfront makes the difference between treating this like a creative process (slow) and treating it like a production system (fast).
Claude prompt → raw script → edit for hook and pacing. 600–900 words max.
Paste script into ElevenLabs → generate → export MP3. One pass, no retakes.
Download matching stock clips from Pexels. Generate custom AI scenes in Runway for gaps.
CapCut: drag audio, lay clips, auto-captions, export 1080p. Upload to YouTube.
That's it. The whole pipeline is 60 minutes because every tool in it has one job, does it fast, and hands off cleanly to the next. There's no creative ambiguity — just execution steps.
The script is the backbone. Everything else — your voice, your visuals, your edit — derives from it. Get the script wrong and no amount of production polish saves the video.
The script prompt below is the one I use for every video. It's not a template you adapt. It's a specification you fill in. The three variables are: your niche, your video topic, and the target length (always 600–900 words).
Hook quality is everything. The first 30 seconds of your video determine whether the algorithm shows it. If Claude gives you a weak hook ("Many people wonder…" or "Today we're going to…"), regenerate just the opening paragraph with this follow-up: "The hook is too generic. Rewrite the first paragraph with a specific dollar figure, counterintuitive claim, or concrete surprising fact. No rhetorical questions."
After Claude outputs the script, read it once out loud. Fix anything that sounds unnatural when spoken. Pay attention to:
Don't over-edit. The goal isn't prose perfection — it's clean audio when spoken at 1.1x TTS speed. If it sounds fine spoken aloud, it ships.
ElevenLabs is the standard for faceless YouTube voice. The output quality difference between ElevenLabs and the free alternatives (Google TTS, Amazon Polly) is not subtle — it's the difference between content that sounds professional and content that sounds like a 2019 podcast intro.
| Niche | Recommended Voice | Settings |
|---|---|---|
| Finance / Business | Adam or George | Stability 0.65, Clarity 0.80, Speed 1.05 |
| AI Tools / Tech | Charlie or Daniel | Stability 0.55, Clarity 0.85, Speed 1.10 |
| Health / Wellness | Rachel or Bella | Stability 0.70, Clarity 0.75, Speed 1.00 |
| Motivation / Self-help | Ethan or Liam | Stability 0.60, Clarity 0.80, Speed 1.05 |
| History / Education | Thomas or Matthew | Stability 0.75, Clarity 0.82, Speed 0.98 |
The settings matter. Stability controls how consistent the voice sounds across the full script. Set it too high and the voice sounds robotic on emphasis words. Set it too low and the cadence varies unpredictably. The values above are tuned for long-form narration — don't adjust them unless the output is clearly wrong.
Generate the full script in one pass. Don't regenerate sentence by sentence — the micro-variations in tone are what make the voice sound natural. If one phrase sounds slightly off, accept it. Listeners don't notice what you notice. The only reason to regenerate is if a word is mispronounced entirely, and even then, you can fix it in CapCut's auto-caption edit or add a pronunciation note in brackets.
Free plan limit: ElevenLabs free tier generates 10,000 characters/month. A 700-word script is roughly 4,200 characters. That's 2 videos per month on the free plan. For volume production, the Starter plan ($5/mo, 30K chars) covers about 7 videos. Creator plan ($22/mo, 100K chars) covers the full 3x/week schedule. Budget this in — voice generation is not an optional cost for professional output.
Faceless YouTube is a visual medium. The audio carries the information — the visuals carry the viewer. A video with great audio and mediocre visuals gets abandoned at 30 seconds. The visual workflow below takes 20 minutes because there's a system: stock clips first, AI generation only for gaps.
Pexels and Pixabay are the two best free sources. For every major topic in your script, download 3–5 clips. You won't use all of them — you're building a library to pull from during assembly. Search terms should match the visual concept, not the script words.
Search strategy: Don't search "artificial intelligence" on Pexels — you'll get generic blue robot hands. Search for the output of what you're describing. Talking about productivity tools? Search "person typing laptop coffee shop" or "focused work morning." The visuals that hold attention are human and contextual, not abstract.
Runway Gen-3 fills the gaps where stock footage doesn't have what you need — specific scenarios, product-style close-ups, or anything that needs to match a particular visual identity. Gen-3 is the current standard; Kling and Sora are the alternatives. All three produce usable output for B-roll purposes in under 3 minutes per clip.
For a typical 5-minute video, you need 8–12 distinct B-roll clips. Stock usually covers 6–8 of those. Runway fills the rest. Plan for 3–4 Runway generations per video.
| Tool | Cost | Best For | Generation Time |
|---|---|---|---|
| Pexels Free | $0 | Real-world stock: offices, people, cityscapes | Instant download |
| Pixabay Free | $0 | Nature, abstract, backgrounds | Instant download |
| Runway Gen-3 Paid | ~$12/mo basic | Custom AI scenes stock can't provide | 1–3 min per clip |
| Kling AI Paid | ~$10/mo basic | Runway alternative, often faster | 30–90 sec per clip |
You don't need both Runway and Kling. Pick one based on your budget and stick with it. Consistency in your workflow matters more than the marginal quality difference between the two.
CapCut is the assembly tool. It's free, runs in-browser, handles auto-captions, and exports 1080p. It replaced $300/year subscriptions for most faceless creators because it does the three things that actually matter: timeline editing, captions, and export — all without a learning curve.
In CapCut, follow this exact order. Changing the order introduces re-sync headaches.
Drag your ElevenLabs MP3 to the timeline. This is your master track — everything else aligns to it.
Match clips to the audio topic — not word for word, but topic by topic. Finance discussion → office/trading visuals. Tool demonstration → screen-style close-ups.
CapCut's auto-caption tool is accurate enough for edited TTS audio. Review once, fix any mispronounced words, set style to match your niche (minimal for Finance, bold for Motivation).
Instrumental only. Set at 8–12% volume — just enough to fill silence without competing with the voice. CapCut's built-in library has clean royalty-free options.
Export at 1080p 30fps. Title and description can be drafted by Claude in 2 minutes using a title prompt — don't spend more than that on metadata.
Batching is where real speed comes from. Run all 5 scripts through Phase 1 on Monday. Generate all 5 voice files on Tuesday. Download all stock + AI footage on Wednesday. Edit all 5 on Thursday. Upload Friday–Sunday. Five videos. One week. Under 6 total hours if you're tight about it.
Here's every tool in the pipeline with its monthly cost. The $0 path produces 2 videos per month on the free tier limits. The $39/month stack supports full 3x/week production.
| Phase | Tool | Purpose | Free Plan | Paid |
|---|---|---|---|---|
| Script | Claude (Anthropic) | Script generation | Yes (limited) | $20/mo Pro |
| Voice | ElevenLabs | Text-to-speech | 10K chars/mo | $5/mo Starter |
| Visuals | Pexels | Free stock footage | Unlimited | — |
| Visuals | Runway Gen-3 | AI video generation | Limited credits | $12/mo Basic |
| Edit | CapCut | Edit + captions + export | Unlimited | — |
| Thumbnails | Canva | Thumbnail design | Yes (templates) | $2/mo Pro |
The math on paid tools: At $12 RPM in the AI Tools niche, 1,000 views generates $12. A single video that hits 2,000 views/month covers your entire $37/month tool stack. At 3 videos/week × 1,000 views each = 12,000 views/month = $144 AdSense. The tool cost is a rounding error. The question is never "can I afford the tools" — it's "can I hit 1,000 views per video."
Thumbnails aren't in the 60-minute count because they don't block upload. Create them in parallel or after — YouTube lets you update thumbnails post-publish without resetting video performance.
For faceless channels, the effective thumbnail formula is: one bold text claim + one strong visual. No faces, obviously. Finance channels use clean financial charts or cash imagery. AI Tools channels use interface screenshots or robot/tech visuals. Motivation channels use abstract silhouettes or dramatic landscapes.
Always A/B test two thumbnails per video. Change only one variable — the text or the image, not both. Check click-through rate (CTR) at 500 impressions and keep the winner. A 1% improvement in CTR compounds massively across a catalog of 50+ videos.
The first 10 videos exist to teach the algorithm your audience. They're not for views. They're not for subscribers. They're for categorization.
During this phase: stay strictly on-niche. If you're an AI Tools channel, every video is an AI Tools video. No detours into productivity, no off-topic experiments. The algorithm needs 10+ videos in the same niche before it understands who to recommend you to.
| Video # | Focus | Goal |
|---|---|---|
| 1–3 | Core niche topics | Algorithm categorization (not views) |
| 4–6 | Higher-volume search terms | First organic search traffic |
| 7–10 | Mix of search + browse topics | Identify which format your audience prefers |
| 11+ | Double down on what's working | Build on confirmed hits, not guesses |
At video 10, look at watch time percentages and CTR. The videos with the highest watch time are the ones the algorithm will amplify. Make more of those. The rest you keep as baseline content — don't delete anything.
Lessons 1–2 are free. Lessons 3–12 cover everything else: the affiliate stack, digital product launch, algorithm optimization, monetization stacking, and the system that scales it.
The exact spreadsheet, prompt library, and 30-day posting schedule we use to launch faceless channels from zero. Free, forever.