Solo YouTuber AI Stack: Script, Visuals, Voice, Thumbnails

This stack is for a solo YouTuber publishing 2-4 videos per month on a $187/month tools budget.

Hands-on time per 10-minute video: 4-5 hours, down from a conventional 12-15 hours. Five tools cover script, visuals, voiceover, editing, and thumbnails.

System Verdict

Pick this stack for 2-4 mid-length videos per month with generated voiceover and AI-assisted visuals. Claude scripts, Midjourney visualizes, ElevenLabs voices, Descript assembles, Ideogram thumbnails.

Skip it for photoreal-video niches (Midjourney is stylized), human-voice niches (ASMR, meditation, high-end corporate), or daily-publishing cadence.

Total cost: $187/month. Replaces roughly $800-1,800 of freelance work per video.

Key Facts

The short version

• Claude Opus 4.7 writes the script and hook variations. Rewrite the opener and CTA by hand. Saves ~30 min.
• Midjourney V8 Alpha generates 8-12 concept images. Pick 3-4 for B-roll or reference. Saves ~45 min.
• ElevenLabs generates a cloned-voice voiceover. Layer with music and sound design. Saves ~60 min.
• Descript assembles video, auto-captions at ~98% accuracy, handles basic color grading. Saves ~30 min.
• Ideogram creates thumbnail variations. Pick one, tweak text in Descript. Saves ~20 min.
• Net time saved per video: ~3 hours. Monthly stack cost: $187.

The stack

Claude Opus 4.7 ($20/mo)

Owns script generation and hook ideation.

Inputs: video topic, target audience, desired length (800-1,200 words for a 10-minute video). The 1M-token context window accepts a full channel transcript history as voice reference.

Output is ~70% usable. Always rewrite the first 30 seconds and the CTA by hand; AI hooks feel generic. Claude Pro at $20/mo is sufficient; Max 5x at $100/mo is overkill for this workflow.

Midjourney V8 Alpha ($30/mo)

Generates visual concepts and B-roll reference images.

Prompt with scene descriptions pulled from the script (“overhead shot of a laptop with code on screen, warm lighting, 4K”). V8 Alpha is faster than V7 and handles text-in-image better, though complex typography still fails.

Typical usage: 8-12 images per video, 3-4 selected. V8 Alpha is on alpha.midjourney.com. Standard $30/mo plan covers ~200-250 fast GPU minutes per video.

ElevenLabs ($99/mo)

Owns voiceover generation.

Train a Creator voice clone on 30 minutes of clean narration. Paste script, set speed to 0.95x (slightly slower than natural speech), generate in one pass. Output is ~95% broadcast-ready.

The $99/mo plan includes unlimited generations and voice cloning. This is the single biggest time-saver in the stack.

Descript ($24/mo)

Assembles the video.

Import the ElevenLabs voiceover, drag Midjourney images and stock footage onto the timeline. Auto-captions land at ~98% accuracy. Auto-reframe crops footage for vertical clips.

Standard plan at $24/mo covers a solo workflow. Pro at $34/mo adds collaboration; not needed for solo.

Ideogram ($14/mo)

Owns thumbnail generation.

Prompt with topic plus visual style (“bold red text, shocked face, 3D effect”). Ideogram returns 4-6 options in seconds. April 2026 text rendering reads clearly at 168x94 pixels.

$14/mo plan includes 100 monthly generations. Typical usage: 30-40 per video.

The workflow, step by step

1. Script generation (Claude, 20 min). Open Claude Pro. Paste video topic and target keyword. Prompt: “Write a 1,000-word YouTube script for [topic]. Hook in the first 30 seconds. Tone: conversational, no jargon. Include 3 call-to-action variations at the end.” Copy output to a Google Doc. Rewrite paragraph 1 and the CTA by hand.

2. Visual concept ideation (Midjourney, 15 min). Extract 4-5 key scenes from the script. For each scene, write a Midjourney prompt: “/imagine [scene description], cinematic, 4K, warm lighting, no text.” Generate 2-3 images per scene.

3. Voiceover (ElevenLabs, 5 min). Paste final script. Select Creator voice, set speed to 0.95x, generate. Download MP3. Wall-clock is mostly the generation wait.

4. Thumbnail concepts (Ideogram, 10 min). Prompt: “YouTube thumbnail for [topic]. Bold text, high contrast, emotional reaction. 1200x675px.” Generate 4-6 variations. Download the strongest.

5. Video assembly (Descript, 25 min). Create a new Descript project. Import ElevenLabs voiceover as primary track. Drag in Midjourney images plus stock footage (Pexels or Unsplash for filler B-roll). Descript auto-generates captions.

6. Caption correction and timing (Descript, 15 min). Read auto-captions. Correct 2-3 errors, usually proper nouns or technical terms. Adjust timing on key phrases to match visual cuts. Add speaker labels if needed.

7. Thumbnail finalization (Descript, 10 min). Import Ideogram thumbnail into Descript’s text overlay tool. Add final title text. Adjust font size and position for small-size legibility. Export PNG.

8. Export and upload (15 min). Export video from Descript at 1080p, 30fps. Upload to YouTube with auto-captions as the subtitle file. Attach thumbnail. Write the description by hand; AI-generated descriptions are too generic.

Where it breaks

Claude hooks are generic. Opus 4.7 writes competent openers but lacks the specificity that stops a viewer mid-scroll. Skipping the manual rewrite on the first 30 seconds drops CTR by 15-20%.

Midjourney V8 Alpha cannot render logos or branded graphics reliably. Workaround: generate the concept image and add text in Figma or Descript.

ElevenLabs Creator voice does not emphasize key phrases naturally. Re-generating a single sentence with an SSML emphasis marker adds 5-10 minutes per video.

Descript auto-reframe crops out details in corners on vertical clips. Always preview vertical exports before publishing to Shorts.

Monthly cost

Prices verified 2026-04-17 via vendor pricing pages.

For comparison: a freelance video editor runs $500-1,500 per video. A voiceover artist runs $100-300 per video. This stack replaces ~$800-1,800 of freelance spend per video for $187/month.

Who this is for

Copy this stack for 2-4 videos per month, solo operator, looking to cut editing time in half without sacrificing quality. Baseline comfort with pacing, color grading, and caption placement is required; the tools handle generation, not creative direction.

Skip it for photoreal video needs (Midjourney generates stylized images, not stock footage), human-voice niches (ASMR, meditation, high-end corporate), or daily-publishing cadence.

FAQ

Can the free tiers cover this workflow? Partially, and not at a sustainable cadence. Claude Free caps at 30-100 daily messages depending on complexity. Midjourney has no free tier. ElevenLabs Free caps at 10,000 characters/month (~one short voiceover). Descript Free caps at 3 hours/month. Ceiling hits after one video. Paid tiers are necessary.

OpenAI frontier models instead of Claude for scripts? Both work equally well for a single script. Claude’s advantage is the 1M-token context window for feeding full channel history as voice reference. For a one-off script, they are interchangeable.

How is the Creator voice trained? Upload 30 minutes of clean audio (no background noise, consistent microphone). ElevenLabs trains the voice model in ~24 hours. Training cost is bundled in the $99/mo plan. Quality lands at ~90%+ match to the source voice given clean input.

Can the entire workflow run fully automated? Partially. Zapier or Make can chain Claude to Midjourney to ElevenLabs to Descript, but creative control drops at each stage. Keep a human in the loop for script approval and thumbnail selection. Full automation saves ~30 minutes at a quality cost that is not worth the tradeoff at this cadence.

System Notes

This page documents an operational stack verified by the aipedia.wiki editorial pipeline. Last verified 2026-04-17.

Related

• Tools: Claude · Midjourney · ElevenLabs · Descript · Ideogram
• Workflows: Podcast Automation Stack · Newsletter Stack · Research Assistant Stack