Muvon Overview
Write for a named viewer with a real reason to watch. Follow supplied facts and proof, then the author's documented voice, then platform conventions. Use content-voice for general editing; this skill adds spoken delivery and YouTube packaging.
Produce a package with an identifiable author and an honest next step. Drafting doesn't authorize publication, replies, or ad operations. Keep source notes separate from speech; don't fill missing evidence with invented experience.
Mental model
Work backward from the viewer's task and the evidence you can show. Choose packaging before scripting, then revise it if the footage cannot keep the promise (directional; Galloway). Treat retention graphs as diagnostic evidence for that upload. Don't prescribe universal opening deadlines, re-hook intervals, or a guaranteed format-transfer effect (directional).
Views, attention, and commercial outcomes answer different questions. Metricool observed rising views alongside falling viewing duration and interaction measures across both formats (measured, Metricool, n=799,718 videos, 2026-07). Keep format and traffic-source comparisons separate; don't infer qualified demand from starts alone (directional).
Open reference/evidence.md when choosing cadence, interpreting benchmarks, estimating revenue, or using a case study. Open reference/delivery.md when preparing assets, optional formats, disclosures, or monetization-dependent copy. Open reference/launch.md when planning a launch campaign or creator collaboration.
Format and packaging decisions
Use these craft bands as editing choices; they are not platform limits or measured optimum lengths (directional).
| Viewer situation | Format and length band | Structure and pacing |
|---|---|---|
| Casual discovery; result is immediately visible | Short; brief, self-contained demonstration | Show the outcome immediately, then the decisive action; cut setup that doesn't affect understanding |
| Buyer solving a named workflow | Long-form; as long as the complete task requires | Show the finished artifact, establish prerequisites, perform the job, retain failure conditions |
| Dedicated learner evaluating tradeoffs | Long-form; extended walkthrough | Keep readable screens and thinking time; chapter meaningful stages instead of adding arbitrary cuts |
| Comparing products | Long-form; matched demonstrations through a bounded verdict | Hold conditions constant, disclose omissions, show where each option fits |
| Audience needs to participate | Live session or Premiere if available; session-length | Schedule around a concrete discussion and staff moderation; prepare the replay package |
| Existing viewers need an update | Channel post or poll if available; compact prose | State the change or ask a decision-relevant question; attach useful proof |
Square/vertical uploads up to three minutes classify as Shorts under the current date/channel rules; longer-than-one-minute Shorts with active copyright claims are blocked globally (official, Shorts classification, checked 2026-09). Check the delivery reference before scripting around music. Finish the promise at its useful length; compare viewed-versus-swiped-away analytics with similar Shorts, not a universal cutoff (official, YouTube Shorts analytics, 2026-09; editing choice is directional).
The Shorts player update includes Clear Screen, accelerated playback and mute controls; hearts replace thumbs-up, with “Not Interested” and “Don't recommend this channel” replacing dislike (official, YouTube Shorts experience, 2026-06). Make the meaning legible in the footage and captions even with sound muted or interface overlays hidden; test accelerated playback for comprehension (directional).
Match packaging to the entry point
| Intended entry point | Title | Thumbnail and opening |
|---|---|---|
| Search | Name the exact task using the buyer's language | Show the output and required context; answer promptly |
| Browse/Home | Make the situation understandable without channel history | Show concrete stakes; pay off the visual promise immediately |
| Suggested beside a related video | Offer the next useful question in that viewing context | Make the connection clear without copying the adjacent creator |
| Existing audience | Name the development they have reason to care about | Supply enough context for a newcomer; avoid unexplained episode lore |
These packaging choices are craft hypotheses, not traffic-source ranking formulas (directional). Front-load the essential promise and preview truncation on the intended screens. Don't force a fixed character band (directional).
Choose a readable thumbnail concept that complements the title. Use the fewest words needed, including none; a relevant natural expression is optional (directional). The breakout-thumbnail sample included text-free images and rarely used exaggerated expressions (measured, vidIQ thumbnails, n=500 thumbnails across 30 niches, 2026-07). Preview at small size and against light and dark interfaces (directional).
Where available, native A/B testing supports up to three titles, thumbnails, or combinations. Results use watch-time share and the labels “Winner,” “Performed Same,” and “Inconclusive” (official, YouTube A/B testing, 2026-09). Keep variants honest and meaningfully different; don't declare victory from a tiny CTR change or stop a test to manufacture a winner (directional).
Script anatomy and publishing package
Cold-open on the promised task or evidence. Explain only the context needed for the next action. Bridge from the hook directly into the demonstration; replace a background reset with context over relevant footage. Change visuals or introduce a new payoff when the subject needs it, then inspect the resulting retention curve (directional).
Deliver a package containing title options and their distinct promises, thumbnail asset brief, spoken script with separate visual directions, description, caption/subtitle copy, and the intended next step. Include a pinned-comment draft, optional channel-post teaser, chapter labels, end-screen cue, and reply plan where available (directional).
Use this description anatomy as a template (directional):
| Position | Content |
|---|---|
| Opening | Exact problem and what the viewer will see; put material limitations near the claim |
| Action | Plain-language CTA and verified destination; state affiliation beside the relevant promotion |
| Navigation | Timestamps from the final edit with descriptive chapter names |
| Evidence | Source links, conditions, corrections, credits and relevant disclosures |
| Optional metadata | Relevant hashtags only; preview their rendered placement rather than promising where they appear |
Make the pinned comment useful on its own: clarify a limitation, answer the likely setup question, or explain the next step. Use descriptive tags only where useful; don't keyword-stuff descriptions or promise transcript-indexing gains. Correct captions for names and technical terms. Invite viewers to a verified next video only when it helps continue their task; check the editor's end-screen timing. Otherwise, end on the result or material limitation without a CTA (directional).
Launch and proof posts
For campaign planning, record audience, buying situation, promise, proof inventory with sources and reuse-permission status for each asset or quotation, desired action/destination, stage, and disclosures. Verify availability and price conditions if claimed; name the speaker and reply owner. Use voice samples when available; missing samples or optional fields don't block supported copy. No CTA is valid, and unsupported outcomes stay out of the script.
Use task-complete length bands, not invented minute targets (directional):
| Shape | Native anatomy | Length band |
|---|---|---|
| Launch day | Recognizable problem → working product → who can use it now → limitation → action | Compact demo; extend for necessary setup |
| Demo or artifact | Show output → reproduce the decisive steps → inspect the result | Short for an isolated action; walkthrough for a full workflow |
| Customer proof | Baseline and conditions → intervention → bounded outcome → what remains unresolved | Enough room for the evidence and exclusions |
| Founder decision | Observed constraint → decision → visible artifact → cost of the choice | Brief update or extended explanation when tradeoffs need it |
| Objection answer | State the actual objection → show the relevant test → say who should decline | Complete answer without unrelated feature tour |
| Recap or lessons | Revisit the original promise → show what happened → correct assumptions → next useful step | Compact synthesis supported by source material |
Apply the strip-test: with the product name removed, the upload should still teach a useful task or show meaningful evidence (directional). For B2B, start with actual support questions and buyer-intent searches; selected founder cases support testing narrow tutorials, without proving a universal conversion formula (directional; SellOnTube).
Route long-form CTAs through the description and an available pinned comment; for Shorts, check whether the account offers a related-video or profile route and verify the destination from the viewer's interface. Don't rely on typed URL text being clickable. Use a relevant next-video or series path when further evaluation is needed (directional). Keep source, medium and campaign naming consistent; distinguish placements with utm_content (official, Google Analytics URL builder, 2026-09).
For eligible U.S. creators, Amazon products can be tagged in Shorts, long-form and live streams through Shopping (official, YouTube Amazon affiliate announcement, 2026-08). Disclose the commercial relationship and check account eligibility before writing a native-shopping CTA (directional).
Collaborate with relevant creators and voluntary, openly affiliated employees; give them proof constraints without scripting fake reactions. YouTube permits ordinary like/comment/subscribe requests, but prohibits deceptive engagement manipulation, repetitive promotional spam, and synthetic mass-production that floods the platform (official, YouTube Spam Policy, 2026-09). Editorially, ask for an action only when it fits the viewer's next step.
Distinguish own-channel updates from seeding another creator's comments, live chat or Community. Read the conversation and host rules; give a standalone answer. Link the launch only when directly useful and promotion is permitted; disclose affiliation. Otherwise omit it. Don't coordinate duplicate comments or fake independent reactions (directional).
Use this illustrative launch-week schedule only as a planning example, adjusting to proof readiness and staffing (directional):
| Relative day, illustrative | YouTube asset | Reply plan | Measure |
|---|---|---|---|
| T−7 | Channel-post poll or founder constraint update | Collect real objections | Relevant questions and audience fit |
| T−3 | Self-contained demo excerpt | Answer prerequisites; verify next-video path | Chose-to-view report and substantive comments |
| T0 | Launch walkthrough; Premiere only if staffed | Cover the first hour and following hours; correct broken paths | Watch time, retention at proof, traffic source, buyer questions |
| T+1 to T+3 | Objection answer or eligible live demonstration | Resolve unanswered setup questions | Useful follow-up viewing and resolved objections |
| T+7 | Bounded proof or recap; use a lessons update if outcomes aren't ready | Add corrections and revisit open questions | Continued viewing, attributable visits and qualified actions |
The staffing window is a craft choice. Metricool found 83% of interactions in the first ten days (measured, Metricool press release, n=799,718 videos, 2026-08). Keep coverage beyond launch day and preserve evergreen search usefulness (directional). Report destination conversions separately from YouTube-native metrics; don't credit a whole campaign's results to YouTube (directional).
Use founder delivery for firsthand decisions and brand-channel presentation for durable product instruction. Choose by documented voice and continuity of ownership; no verified founder-versus-brand distribution advantage is established here (directional).
Voice on this platform
Preserve the author's register, including natural contractions and fragments. Don't manufacture typos, fake edit notes, lowercase camouflage, or dropped articles. Use firsthand claims only when source records support that speaker's experience.
Write lines the speaker can say while doing the action. Read the script aloud, separate breaths at changes of thought, and replace abstract narration with concrete references to the visible screen or object. Keep useful hesitation only when it occurs naturally; don't write fake thinking pauses or inject filler (directional).
Remove generic spoken scaffolding such as “In this video we'll,” “without further ado,” and “let's dive in” when it delays the subject. End on the result, limitation, or useful invitation; don't append a recap followed by a subscribe request by habit (directional).
Review meaning across title, thumbnail, speech, description, and excerpts. A question, contrast, triad, polished sentence, or em-dash is not evidence of AI authorship. Cut padding and unsupported drama, not the author's register. Retain technical terms, units, negation, and the condition that makes a result true. Don't replace them with vague synonyms to sound conversational (directional).
Recut a Short from the actual explanation, preserving any qualifier that changes its claim. Rebuild the introduction and next step for its viewer; don't just crop a long-form script or import another platform's CTA. Use a fluent reviewer for localized speech; language research motivates this check without establishing a YouTube engagement gain (directional; MIDI).
Cadence and engagement
Choose a sustainable publishing pace that leaves time for evidence gathering and replies. Test modest Shorts support for a deeper video before increasing volume; the observed association with long-form views is correlational (measured, Metricool Shorts, n=799,718 videos, 2026-08). Use the evidence reference before quoting frequency benchmarks.
Assign a responder for launch hours and later coverage, including audience-active evenings or weekends when needed. Answer the actual question in ordinary prose; distinguish a confirmed limitation from a workaround. Remove scam links and abuse while preserving fair criticism (directional). Cross-platform moderation and reply studies motivate staffing, but don't establish a YouTube-specific reply uplift; see the launch reference.
What gets suppressed
Separate removal, monetization eligibility, and editorial weakness. The egregious-clickbait announcement described undelivered packaging promises and removal initially without strikes, with rollout starting in India (official, Google India, 2024-12). Don't extend that announcement's geographic scope.
Repetitive or mass-produced content was already ineligible for monetization when “inauthentic content” terminology was clarified; reused-content policy was unchanged (official, YouTube monetization policies, 2025-07). Avoid cosmetic remixes of others' work; add substantive original explanation and secure necessary rights (directional).
Disclose realistic generated or meaningfully altered media as required. Script/title/caption assistance and cloning the creator's own voice for voiceovers/dubs are listed exceptions; fabricated realistic events and AI-generated music require disclosure. The label alone doesn't limit audience or monetization eligibility (official, YouTube AI disclosure, checked 2026-09). A label doesn't validate the claim. See the delivery reference for automatic labels; don't infer an AI-narration penalty.
Examples
These are illustrative drafting examples, not reports of real businesses. Brackets mark missing source material; such drafts are not publishable until filled and checked. The bad versions demonstrate packaging defects, not suggested copy.
Developer tool: recorded diagnostic workflow
Bad title: “This tool fixes every broken build.” Good title: “Find a missing environment variable with [tool].” Opening: “This build fails before the app starts. I'll show the missing variable in the log, then rerun it with [documented configuration]. This recording covers [environment]; other failures need their own diagnosis.” Thumbnail: the relevant log line, with secrets removed. Description: “Reproduce this failure using [fixture link]. Setup requirements: [verified requirements]. I maintain [tool].” Why it works: packaging promises the specific action the footage can show. Why it works: provenance and the environment limitation prevent a miracle-fix claim.
B2B SaaS: customer proof
Bad description: “Teams everywhere save hours with our platform.” Good title: “How [customer] handles duplicate renewal records.” Opening: “[Customer] supplied this export from [period]. The review queue changed from [before] to [after] under [conditions]. These records still needed a person's decision.” Description: “See the matching rules and exclusions in [approved case link]. Try the workflow with [destination]. [Affiliation disclosure].” Pinned comment: “The comparison excludes [excluded work]. The video shows why those records stayed in manual review.” Why it works: the proof shape ties the outcome to its denominator and conditions. Why it works: the CTA offers evaluation without turning the case into a general guarantee.
Consumer creator: product demonstration
Bad title: “The bag everyone needs.” Good title: “Packing [bag] around a rigid lunchbox.” Opening: “The lunchbox reaches this seam. I'll pack the supplied kit around it and show what doesn't fit. Durability needs a separate test.” Description: “Contents shown: [verified packing list]. [Gifted or sponsored disclosure]. Product details: [destination].” Thumbnail: the actual bag opening and lunchbox, without an exaggerated reaction. Why it works: a visible physical constraint gives the viewer a concrete buying question. Why it works: the test's limitation and commercial disclosure remain beside the claim.
Local service: objection answer
Bad title: “A perfect garden after a quick visit.” Good title: “What [service] can prune without entering the side passage.” Opening: “This gate blocks access to the back hedge. The site survey shows which branches are reachable from here and which need another access arrangement.” Description: “This example covers [approved site conditions]. Request a site assessment: [booking link]. A quote requires access confirmation.” Follow-up reply: “The video doesn't establish access for your property. Send the site details through the assessment form so the team can check.” Why it works: the objection answer teaches a bounded service decision. Why it works: the booking path preserves the condition instead of promising an unseen outcome.
Checklist
- Hook and first line deliver the packaging promise; the upload holds a single central idea.
- Format classification and current asset limits checked; thumbnail and opening previewed on intended surfaces.
- Description, verified CTA/link path, pinned comment and chapter labels prepared where applicable.
- Caption/subtitle copy checked; asset brief includes visual descriptions and alt text where supported; end-screen timing checked in the editor.
- Every specific is supplied or placeholder-marked; unresolved factual placeholders block publication.
- Proof inventory records sources, reuse permissions and restrictions for customer material, quotations and footage; publishing handoff identifies unresolved permissions.
- Meaning and author-voice review covers the whole package; captions and excerpts preserve qualifications without style quotas.
- Affiliation, sponsorship/paid-promotion controls and required synthetic-media labels checked.
- Optional channel-post teaser and first-hours reply plan ready; moderation and later follow-up assigned.
- Metric definitions and comparison cohorts recorded; test conclusions use the actual Studio result.
References
Dates on undated Help pages denote validation month. Benchmarks and case-study limitations are mapped to actions in the linked references.
- YouTube A/B testing, undated; checked 2026-09.
- YouTube Shorts analytics, undated; checked 2026-09.
- Shorts classification, undated; checked 2026-09.
- Shorts experience update, 2026-06-25.
- AI disclosure, undated; checked 2026-09.
- Monetization policies, clarification 2025-07-15.
- Spam Policy, undated; checked 2026-09.
- Clickbait announcement, 2024-12-18.
- Evidence sources, delivery sources, launch and voice sources, dated individually.
Re-validate when:
- Studio metrics, player controls, test outcomes, or upload limits change.
- Policy wording, geographic availability, monetization rules, or disclosure controls change.
- New vendor reports revise the sampled period or methodology.
Validated: 2026-09