KI-gestütztes YouTube-Produktionssystem

As a professional YouTube production manager, here is a comprehensive AI-assisted workflow designed to produce high-quality videos faster, with systems for every stage from ideation to repurposing. --- ### Phase 1: Topic Research & Strategy (The "What") **Goal:** Identify high-probability video topics with proven search demand and audience interest. **AI Tools:** VidIQ, TubeBuddy, ChatGPT, Google Trends, AnswerThePublic. **Workflow:** 1. **Seed Generation:** Use ChatGPT to generate 20-30 video ideas based on your niche, using prompts like: "Generate 20 YouTube video ideas for [NICHE] that solve common pain points for beginners." 2. **Validation:** Feed these ideas into VidIQ or TubeBuddy. Filter for topics with a **Keyword Score** above 60 and a **Competition** score below 40. 3. **Title & Angle Optimization:** For the top 3 ideas, use ChatGPT to generate 5 title variations each. Use the prompt: "Create 5 click-worthy YouTube titles for a video about [TOPIC]. Use power words, numbers, and curiosity gaps." 4. **Outline Creation:** Ask ChatGPT to create a detailed, 10-point video outline with timestamps, ensuring a logical flow and a strong hook in the first 30 seconds. **Quality-Control Checklist:** - [ ] Topic has proven search volume (not just personal interest). - [ ] Title includes a primary keyword and a curiosity/benefit driver. - [ ] Outline has a clear "Payoff" (what will the viewer learn by the end?). - [ ] Hook is defined and compelling. --- ### Phase 2: Scripting & Storyboarding (The "Blueprint") **Goal:** Transform the outline into a tight, engaging script and visual plan. **AI Tools:** ChatGPT, Claude, Notion (for template management), Jasper. **Workflow:** 1. **Draft Script:** Use ChatGPT to expand the outline into a full script. Use the prompt: "Write a conversational, energetic YouTube script for the following outline: [PASTE OUTLINE]. Use short sentences, active voice, and include [NUMBER] verbal transitions." 2. **Human Edit (Critical):** You must edit this draft. Inject your personality, add personal anecdotes, and remove any robotic phrasing. AI is a first draft, not a final product. 3. **B-Roll & Visual Cues:** In the script, use brackets `[BRoll: Close-up of product]` to indicate where B-roll, screen recordings, or stock footage should go. 4. **Storyboard:** Use a tool like Canva or Milanote to create a simple visual storyboard for complex sections, ensuring the visuals match the narrative. **Quality-Control Checklist:** - [ ] Script is 100% in your voice (not AI-sounding). - [ ] Hook is within the first 15 seconds. - [ ] Every 60-90 seconds has a visual change or B-roll cut. - [ ] Script includes a clear Call-to-Action (CTA) for like, subscribe, and comment. --- ### Phase 3: Voiceover & Filming (The "Capture") **Goal:** Capture high-quality audio and video assets efficiently. **AI Tools:** ElevenLabs, Adobe Podcast Enhance, Descript, OBS Studio. **Workflow:** 1. **Voiceover (Option A - AI):** If using AI voice, use ElevenLabs. Select a voice that matches your brand. Generate the audio in segments (per paragraph) for easier editing. Use the "Voice Cloning" feature for a consistent brand voice. 2. **Voiceover (Option B - Human):** Record your voiceover in a quiet room. Use a dynamic microphone (e.g., Shure SM7B). Run the raw audio through **Adobe Podcast Enhance** to remove room tone and background noise automatically. 3. **Filming (Talking Head):** Use a 4K camera (or high-end phone) with a 50mm lens look. Set up a 3-point lighting system (Key, Fill, Backlight). Record in 4K, 24fps for a cinematic look. 4. **Screen Recording:** Use OBS Studio or ScreenFlow. Record at 4K resolution, 60fps for smooth cursor movement. Clean up your desktop and zoom into 150% for better readability. **Quality-Control Checklist:** - [ ] Audio is clean, no background hum or echo. - [ ] Video is sharp, well-lit, and in focus. - [ ] Screen recordings are high-resolution and cursor is visible. - [ ] All files are named consistently (e.g., `VO_01`, `BRoll_Product_01`). --- ### Phase 4: Editing (The "Assembly") **Goal:** Assemble the video into a polished, engaging final product. **AI Tools:** Descript, Adobe Premiere Pro (with AI features), CapCut, RunwayML. **Workflow:** 1. **Auto-Transcription & Rough Cut:** Import your voiceover into Descript. It transcribes the audio. Delete any filler words ("um", "ah") directly from the text, and the video/audio edits itself. 2. **Visual Assembly:** Import the edited audio into Premiere Pro (or CapCut). Lay down B-roll, screen recordings, and talking head footage on the timeline according to your script cues. 3. **AI-Assisted Effects:** Use RunwayML for AI-powered background removal or object removal if needed. Use Premiere Pro's "Auto Reframe" to create vertical versions for Shorts later. 4. **Audio Mixing:** Use AI tools like Auphonic to automatically level audio and reduce noise. Add a background music track at -18dB to -20dB (under the voiceover). 5. **Captions:** Use CapCut or Premiere Pro's auto-captioning feature. Style them with a bold, readable font (e.g., Montserrat) with a subtle background for contrast. **Quality-Control Checklist:** - [ ] Video is under the target length (usually 8-12 minutes for long-form). - [ ] Pacing is fast - no dead air or long pauses. - [ ] Captions are 100% accurate and synced. - [ ] Music volume does not overpower the voiceover. - [ ] Visuals match the spoken content. --- ### Phase 5: Thumbnail & Title (The "Click") **Goal:** Create a high-CTR thumbnail and title combination. **AI Tools:** Canva (with Magic Design), Photoshop (Generative Fill), ChatGPT. **Workflow:** 1. **Thumbnail Draft:** Use Canva's Magic Design. Upload a high-quality frame of your face (with an expressive emotion) and a relevant product/object. Let AI generate 3-5 layout options. 2. **Manual Refinement:** In Photoshop, use Generative Fill to expand backgrounds or remove unwanted objects. Add 3-4 words of large, bold text (e.g., "STOP DOING THIS"). Use high-contrast colors (yellow, red, green). 3. **Title Finalization:** Use ChatGPT to generate 5 final title options based on your script. Choose the one that balances curiosity, keyword, and clarity. 4. **A/B Testing:** Use YouTube's "Test & Compare" feature to test 2-3 thumbnails against each other for the first 48 hours. **Quality-Control Checklist:** - [ ] Thumbnail is readable at 150% zoom (mobile size). - [ ] Thumbnail has a single focal point (face or object). - [ ] Text on thumbnail is 3 words or less. - [ ] Title and thumbnail tell a cohesive story. --- ### Phase 6: Publishing & SEO (The "Launch") **Goal:** Optimize metadata for search and algorithm discovery. **AI Tools:** TubeBuddy, VidIQ, ChatGPT. **Workflow:** 1. **Description:** Use ChatGPT to generate a 3-paragraph description. Paragraph 1: Summary with primary keyword. Paragraph 2: Detailed breakdown with timestamps. Paragraph 3: Links, socials, and "Subscribe" CTA. 2. **Tags:** Use VidIQ's "Tag Inspector" to find 10-15 high-volume, low-competition tags. Include 1-2 broad tags and 1-2 exact-match tags. 3. **End Screens & Cards:** Add an end screen (20 seconds) with a link to a related video and a "Subscribe" button. Add 2-3 info cards during the video (e.g., at the 2-minute mark, link to a previous video). 4. **Scheduling:** Publish at your audience's peak time (check YouTube Analytics > Audience > When your viewers are online). Schedule the video for that time. **Quality-Control Checklist:** - [ ] Description includes primary keyword in first 150 characters. - [ ] Timestamps are accurate. - [ ] End screen elements are not covering important content. - [ ] Video is set to "Public" (not "Unlisted" by mistake). --- ### Phase 7: Analytics Review (The "Feedback Loop") **Goal:** Understand performance and iterate for the next video. **AI Tools:** YouTube Analytics, VidIQ, ChatGPT. **Workflow:** 1. **48-Hour Check:** After 48 hours, review the "Engagement" tab. Look at **Average View Duration (AVD)** and **Click-Through Rate (CTR)** . 2. **Diagnosis:** If CTR is below 4%, the thumbnail/title is the problem. If AVD is below 40%, the content or pacing is the problem. 3. **AI Analysis:** Copy your video transcript and key analytics into ChatGPT. Prompt: "Analyze this transcript and the following metrics [PASTE METRICS]. Identify the top 3 reasons for viewer drop-off and suggest 3 specific improvements for the next video." 4. **Documentation:** Log your findings in a Notion database. Track what worked and what didn't for future reference. **Quality-Control Checklist:** - [ ] CTR > 5% (if not, redesign thumbnail). - [ ] AVD > 50% (if not, improve hook and pacing). - [ ] Top 3 drop-off points identified and understood. --- ### Phase 8: Content Repurposing (The "Amplification") **Goal:** Turn one long-form video into multiple short-form assets. **AI Tools:** OpusClip, CapCut, Repurpose.io, ChatGPT. **Workflow:** 1. **Clip Extraction:** Use **OpusClip** to automatically find the most engaging 30-60 second segments from your long-form video. It uses AI to detect "viral moments" based on speech patterns and visual changes. 2. **Manual Selection:** Review the AI-selected clips. Choose 3-5 that have a self-contained "hook, story, and payoff." 3. **Vertical Formatting:** Use CapCut's "Auto Captions" and "Auto Reframe" to convert the 16:9 clip to 9:16. Add large, bold captions (80% of viewers watch without sound). 4. **Platform-Specific Tweaks:** - **Shorts:** Add a trending sound (use the "Add Sound" feature in YouTube Shorts). - **TikTok:** Add a "Stitch" or "Duet" prompt in the caption. - **Reels:** Use Instagram's "Remix" feature to encourage engagement. 5. **Scheduling:** Use **Repurpose.io** to automatically publish the clips to all platforms at optimal times. **Quality-Control Checklist:** - [ ] Clip has a clear hook in the first 2 seconds. - [ ] Captions are readable on a phone screen. - [ ] Clip is under 60 seconds (ideally 30-45s). - [ ] Each platform has a unique caption (not copy-pasted). --- ### Recommended Tool Stack (Summary) | Stage | Tool | Purpose | | :--- | :--- | :--- | | Research | VidIQ / TubeBuddy | Keyword & competition analysis | | Scripting | ChatGPT / Claude | Drafting, outlining, title generation | | Voiceover | ElevenLabs / Adobe Enhance | AI voice or audio cleanup | | Filming | OBS Studio / 4K Camera | Screen capture / talking head | | Editing | Descript / Premiere Pro | Transcription, assembly, captions | | Thumbnail | Canva / Photoshop | Design and AI image expansion | | Publishing | TubeBuddy / YouTube Studio | Tags, scheduling, A/B testing | | Analytics | YouTube Analytics / ChatGPT | Performance diagnosis and iteration | | Repurposing | OpusClip / CapCut / Repurpose.io | Clip extraction, vertical formatting, auto-publishing | --- ### Final Production Manager's Note The key to this workflow is **human-in-the-loop**. AI handles the heavy lifting (research, transcription, rough cuts, captioning, clip finding), but the creative decisions - the personality, the humor, the emotional connection - must come from you. Use AI to save 60-70% of your time, but reinvest that time into making the final 30% of the video truly exceptional.