Key takeaways
- Almost every number a creator needs fits on one page. This is that page — specs, limits, thresholds, benchmarks and the tool stack, with the reasoning compressed out and links to the long versions where you need it back.
- Three specs decide more than the rest combined: 1280×720 minimum thumbnail at 16:9, a 100-character title that is really a 48-character title, and the first 24–48 hours where YouTube runs its small test.
- The tool market has eight jobs and about forty products. Nobody needs forty. A working stack is one tool per job, and three of the eight jobs have a free option good enough to keep permanently.
- Read the credit line, not the price line. A $23/month editor that grants 25 AI credits is a more expensive tool than a $9/month generator with 700, because the unit of a credit differs — one buys an operation, the other buys a finished image.
- Research is the cheap half of channel growth and is almost entirely free. Packaging is the expensive half. Budget accordingly: most creators buy the reverse.
- When a video underperforms, the fix is decided by which of three numbers is low — impressions, CTR, or average view duration. Part 9 has the diagnostic table.
Most YouTube advice is a story with a number buried in it. That is fine the first time you read it and useless the fourteenth, when you already know the argument and just need to remember whether the tag field caps at 400 characters or 500.
So this is the opposite: a reference. Every spec, limit, threshold and benchmark that comes up in normal channel operation, plus an honest map of the tool market — what each category is for, what the leading products cost in 2026, and which jobs you can do for nothing indefinitely. No narrative arc. Skim it, bookmark it, come back for the row you need.
Where a number needs an argument behind it, there is a link to the full guide. The 2026 thumbnail playbook, the title guide and the trending ideas guide are the three long pieces this page compresses.
How to use this page
Parts 1–3 are the reference half: specs, benchmarks, and the checklist you run before every publish. Parts 4–7 are the tools half: what to buy, what to never buy, and three complete stacks at three budgets. Parts 8–10 are operations — the weekly cadence, the diagnostic table, and a one-page recap you can screenshot.
Part 1: The spec sheet
Everything YouTube will accept, reject or silently truncate. Where YouTube publishes a hard limit, it is stated as one; where the real limit is a display limit rather than a technical one, that is called out, because those are the ones that quietly cost you clicks.
Video and upload
| Spec | Value | Notes |
|---|---|---|
| Max length | 12 hours | Requires a verified account. Unverified caps at 15 minutes. |
| Max file size | 256 GB | Whichever limit you hit first, length or size. |
| Container / codec | MP4, H.264 + AAC-LC | The safest combination. YouTube accepts many more; this one re-encodes cleanest. |
| Aspect ratio | 16:9 | Anything else gets pillarboxed by the player. 9:16 routes to Shorts. |
| Standard resolutions | 1080p = 1920×1080 · 1440p = 2560×1440 · 4K = 3840×2160 | 1440p and above get a better codec on playback, which is why some creators upload 1440p. |
| Recommended bitrate | ~8 Mbps at 1080p24–30 · ~12 Mbps at 1080p60 · ~35–45 Mbps at 4K | YouTube re-encodes everything. Upload above the target, not at it. |
| Frame rate | Match your source — 24, 25, 30, 48, 50 or 60 | Never convert frame rate on export. Interlaced footage should be deinterlaced first. |
| Processing | Low resolutions first, 4K last | Publishing the instant upload finishes means early viewers may see 360p. Wait, or schedule. |
Thumbnails
| Spec | Value | Notes |
|---|---|---|
| Recommended size | 1280×720 minimum | Width must be at least 640px. 16:9 is the only ratio that fills the frame. |
| Max file size | 50 MB | Raised from 2 MB in late 2025, alongside 4K thumbnail support. |
| Formats | JPG, PNG, GIF, WEBP | GIF uploads as a still. PNG for flat colour and text, JPG for photographic. |
| The size that actually matters | 168×94 | Mobile feed. If it fails here, the other numbers are irrelevant. |
| Duration badge safe zone | Bottom-right, ~15% of width | Keep faces and text out of it. It is drawn over your image, not beside it. |
| Requirement | Verified account | Without verification you pick one of three auto-generated frames. |
| A/B testing | Up to 3 variants | Test & Compare resolves on watch time per impression, not raw CTR. |
Platform limits verified against YouTube Help in August 2026. The 4K/50 MB thumbnail change was announced 29 October 2025. YouTube ships product changes without much notice — confirm anything load-bearing against current documentation.
Full detail lives in the thumbnail size guide, and the design rules behind them in the playbook. If you only remember one line from this table: design at 1280×720, judge at 168×94.
Title, description, tags and chapters
| Field | Hard limit | Practical limit | What to do with it |
|---|---|---|---|
| Title | 100 characters | ~48–60 before truncation | Front-load the promise. Everything after character 60 is for search engines, not humans. |
| Description | 5,000 characters | First ~120 shown above the fold | One sentence that restates the promise, then chapters, then links, then boilerplate. |
| Tags | 500 characters total | Marginal impact | Spend two minutes, not twenty. Useful mainly for misspellings of your own brand. |
| Hashtags | 15 max | First 3 display above the title | Exceed 15 and YouTube ignores all of them. Three is the working number. |
| Chapters | No cap | Minimum 3, first at 00:00, each ≥10s | Break any of those three rules and chapters silently do not render. |
| End screens | Up to 4 elements | Last 5–20 seconds | Video must be at least 25 seconds long to have one. |
| Cards | 5 per video | Rarely clicked | Useful for corrections and one mid-roll suggestion. Not a growth lever. |
| Playlists | 5,000 videos | — | Playlist pages rank in search on their own. Underused. |
The gap between the title's 100-character limit and its ~48-character display is the single most expensive misunderstanding on this page. The full title guide covers how traffic source changes the job, ten patterns that work, and the back-catalogue rewrite.
Shorts
| Spec | Value | Notes |
|---|---|---|
| Max length | 3 minutes | Extended from 60 seconds in late 2024. |
| Aspect ratio | 9:16 vertical | Square or 1:1 gets treated as a Short if under the length cap. |
| Resolution | 1080×1920 | Higher is fine. Text should sit clear of the bottom ~20% where the UI overlays sit. |
| Thumbnail | Selectable frame | The Shorts feed autoplays — thumbnails matter for your channel page and search, not the feed. |
| The first second | Everything | The Shorts equivalent of a thumbnail is frame one. Treat it with the same seriousness. |
Eligibility gates
| Unlock | Requirement |
|---|---|
| Custom thumbnails | Verified account (phone number) |
| Videos over 15 minutes | Verified account |
| External links in cards / end screens | YouTube Partner Programme membership |
| YPP — fan funding tier | 500 subscribers, 3 public uploads in 90 days, and either 3,000 valid public watch hours in 12 months or 3M valid public Shorts views in 90 days |
| YPP — ad revenue tier | 1,000 subscribers, and either 4,000 valid public watch hours in 12 months or 10M valid public Shorts views in 90 days |
| Thumbnail A/B testing | Rolled out broadly, but availability still varies by channel |
Monetisation thresholds are the current published figures for most eligible countries and are subject to regional variation and periodic change. Check YouTube's Partner Programme page before planning around them.
Part 2: The numbers worth knowing
Benchmarks are dangerous, because a number that is healthy for a 200-subscriber channel doing search traffic is alarming for a 200,000-subscriber channel doing browse traffic. Use these as ranges to locate yourself in, not targets to hit.
| Metric | Typical range | What it actually tells you |
|---|---|---|
| Impressions CTR | 2–10% for most channels and videos | Whether the package earned the click. Below 2% is a packaging problem almost every time. |
| CTR by traffic source | Search often 8–15% · Browse often 2–6% | Never compare a search video's CTR to a browse video's. Different jobs. |
| Average percentage viewed | ~30–50% on long-form | Whether the video kept the promise. Falls with video length, so read it against duration. |
| Average view duration | Highly variable | The number YouTube optimises against. More honest than percentage on long videos. |
| First-30-second retention | Aim above 70% | The steepest part of every curve. If you lose here, nothing downstream matters. |
| Returning-viewer share | Grows slowly, then fast | The real health metric. Views can be bought with a trend; returns cannot. |
| Subscribers per 1,000 views | ~1–10 | How well the video converts strangers. Wildly niche-dependent. |
| The initial test | First 24–48 hours | YouTube shows the video to a small group, then decides. The window is short and quiet. |
| A/B test sample | ~1,000–5,000 impressions per variant | Under this, you are reading noise. Give it about two weeks. |
The one benchmark that is not a range
A thumbnail and title that overpromise now lose the test they were invented to win. YouTube's A/B tool picks winners on watch time per impression, so a variant can take more clicks and still lose. The Test & Compare guide covers the mechanics.
Part 3: The pre-publish checklist
Twelve items, roughly five minutes, run in this order. Most channels that publish inconsistently are not lacking talent; they are lacking a checklist and re-deciding everything from scratch every week.
- Title under 60 visible characters, promise in the first 40.
- Thumbnail checked at 168×94 — not at the size you designed it.
- Three thumbnail variants ready if you have Test & Compare.
- Nothing important in the bottom-right where the duration badge lands.
- Thumbnail and title do not repeat each other. They are two halves of one sentence, not the same half twice.
- First 120 characters of the description restate the promise in plain words.
- Chapters — three or more, first at 00:00, each at least ten seconds.
- End screen placed in the last 5–20 seconds, pointing at one specific next video.
- Playlist assigned, and it is a playlist a stranger would plausibly binge.
- Captions uploaded or auto-captions reviewed for anything embarrassing.
- Processing complete to your top resolution before it goes public.
- Hook re-watched at 1× with sound on, as a viewer, one final time.
Item 12 is the one everyone skips and the one that catches the most. You have watched the first thirty seconds four hundred times in an editor and stopped being able to hear it.
Part 4: The tool stack, by job
There are roughly eight jobs in running a channel. The tool market presents itself as forty products because vendors want to own several jobs at once, but the buying decision is always the same: pick one tool per job, prefer the free one where it is genuinely good enough, and pay only where the paid version removes real hours.
| Job | What it buys you | Free option good enough? | Worth paying for? |
|---|---|---|---|
| Research & keywords | Which topics have demand | ✓ Yes | ◐ Convenience only |
| Packaging | Thumbnails and titles that earn the click | ◐ Partly | ✓ The best-value spend |
| Video editing | The video itself | ✓ Yes | ◐ Speed and collaboration |
| Audio | Sound that does not drive viewers away | ✓ Yes | ◐ Music licensing |
| Captions & repurposing | Accessibility, plus Shorts from long-form | ◐ Partly | ◐ At volume |
| Graphics & stock | B-roll, overlays, footage you didn't shoot | ✓ Yes | ✗ Rarely early |
| Analytics | Knowing what happened and why | ✓ Yes — Studio is the best tool here | ✗ No |
| Workflow & team | Shipping on schedule with other people | ✓ Yes | ◐ Once you have a team |
Read the third column carefully. Five of eight jobs have a free option that is genuinely good enough to keep forever. That is the actual state of the 2026 market, and it is why a well-constructed stack costs far less than the sum of what gets recommended to you.
Job 1 — Research and keywords
Free stack: YouTube autocomplete for phrasing, Google Trends set to YouTube Search for direction, YouTube Studio's own Research tab for demand inside your niche, and an outlier scan — sorting similar-sized channels by views-to-subscriber ratio — for proof that a topic travels.
Paid options: vidIQ (free tier, paid from around $7.50/mo up to roughly $79/mo) and TubeBuddy (free tier, paid from around $5/mo to $30+/mo) are the two dominant suites. Both are good. Neither sells you data you cannot otherwise reach — they sell it to you faster and in one place.
Verdict: start free. Buy one only when you can name the hour it saves you each week. The full free method is in the trending ideas guide.
Job 2 — Packaging
This is the job where the money should go, and it is the one most creators fund last. Packaging is the only part of your video every single person in your impressions sees. A 10× better edit changes how the people who clicked feel; a 10× better package changes how many people there are.
The category splits three ways, and the split matters more than the price:
| Type | Examples | What you supply | Time per thumbnail |
|---|---|---|---|
| Editor | Photoshop, Photopea, GIMP | Every decision, plus the skill | 20–45 min |
| Template maker | Canva, Adobe Express, Picsart | Layout choices and assets | 10–20 min |
| Generator | Thumblore | A description and a face | 1–3 min for a set of variants |
| Tool | Free tier | Typical paid price | AI allowance |
|---|---|---|---|
| Thumblore | Yes — 100 credits | $9 / $24 / $45 per month | 700 / 2,000 / 4,200 credits per month |
| Canva Pro | Yes, limited | $18/mo rolling · $144/yr | ~500 credits/mo shared across all Magic Studio tools |
| Adobe Photoshop | No — 7-day trial | $22.99/mo annual · $34.49/mo rolling | 25 generative credits/mo on standard plans |
| Adobe Express | Yes | $9.99/mo Premium | 250 credits/mo (Firefly Pro: 4,000 at $19.99/mo) |
| Picsart Pro | Yes, limited | $15/mo rolling · $84/yr | Tiered AI tools, +$5 per extra seat |
| Photopea | Yes, ad-supported | ~$5/mo to remove ads | None |
| Freelance designer | No | ~$15–$60 per thumbnail | n/a — 24–72h turnaround |
Prices checked August 2026 against vendor pricing pages. Tiers are restructured frequently and vary by promotion, region and billing period. Freelance figures are marketplace ranges, not quotes.
Verdict: if you already know Photoshop, keep it — skill you have is cheaper than skill you have to buy. If you do not, the decision is between a template maker and a generator, and it comes down to whether you need three variants per video. Fifteen minutes per thumbnail is tolerable; forty-five minutes for a test set, every week, is what quietly ends A/B testing on most channels. The full comparison runs the annual maths.
Job 3 — Video editing
| Tool | Price | Best for |
|---|---|---|
| DaVinci Resolve | Free · Studio ~$295 one-time | The strongest free editor in existence. Colour work is class-leading. |
| CapCut | Free tier · Pro ~$8–13/mo | Fast social edits, captions, mobile-first workflows. |
| Descript | Free tier · from ~$12–19/mo | Talking-head content — you edit the transcript, it edits the video. |
| Adobe Premiere Pro | ~$22.99/mo annual | Industry standard, teams, deep integration with the Adobe stack. |
| Final Cut Pro | ~$299 one-time (Mac) | Mac-only, fast, no subscription. Pays back inside a year and a half. |
Verdict: Resolve free is enough to build a career on. Move only when a specific limitation is costing you hours — collaboration, or transcript-based editing if you talk to camera for a living.
Job 4 — Audio
Viewers forgive bad video and leave over bad audio. This is the cheapest quality upgrade available to any channel, and most of it is free.
- YouTube Audio Library — free, cleared for monetised use, and the safest music source that exists for a YouTube channel.
- Uppbeat — free tier with attribution; paid tier around $7/mo removes it.
- Epidemic Sound — roughly $10–15/mo personal. Worth it when your music is part of your identity.
- Audacity — free, and still the fastest way to clean up a noisy recording.
- Adobe Podcast Enhance — free tier, and genuinely striking on bad room audio.
- ElevenLabs — free tier, paid from around $5/mo. Relevant to faceless channels; disclose synthetic voice where required.
The one paid line that is nearly always worth it: licensed music, but only once Content ID claims are actually costing you revenue. Before that, the Audio Library is fine.
Job 5 — Captions and repurposing
| Tool | Price | Best for |
|---|---|---|
| YouTube auto-captions | Free | Baseline accessibility. Review them — they are good, not correct. |
| CapCut / Descript | Free tiers | Burned-in animated captions for Shorts. |
| Opus Clip | Free tier · ~$9–29/mo | Cutting long-form into Shorts candidates automatically. |
| Submagic | ~$16–30/mo | High-volume short-form caption styling. |
Verdict: free until you are publishing Shorts several times a week. Automated clipping is a genuine time-saver at volume and a distraction below it — the clips it finds still need a human to judge whether the first second works.
Job 6 — Graphics and stock
- Pexels, Unsplash, Pixabay — free photo and video, no attribution required. Check each licence for your use.
- Google Fonts — free, and safe for monetised channels, which is not true of every font marketplace.
- Envato Elements — around $16.50/mo for unlimited downloads across footage, music, templates and fonts. The best value bundle if you need three of those four.
- Figma — free tier, and a perfectly capable channel-branding tool for banners and end cards.
Job 7 — Analytics
YouTube Studio is the best analytics product in this market, it is free, and no third-party tool has access to data it does not. The three screens that matter: Reach for impressions and CTR, Engagement for retention curves, and Audience for returning viewers.
Google Trends (free) and Social Blade (free tier) round it out. Anything charging for YouTube analytics is selling presentation, not information.
Job 8 — Workflow
Notion or a Google Sheet, free, forever, until you have three or more people touching a video. The tool is not the bottleneck here — a written definition of "done" is. Most creators do not need project management software; they need a checklist like Part 3.
Part 5: The stack that costs nothing
It is entirely possible to run a serious channel at $0 in software. This is what that looks like, and it is not a compromise stack — several of these are the best product in their category.
| Job | Free tool | What you give up |
|---|---|---|
| Research | YouTube autocomplete, Google Trends, Studio Research tab | Speed, not access |
| Editing | DaVinci Resolve | Almost nothing |
| Audio | YouTube Audio Library, Audacity | Music distinctiveness |
| Captions | YouTube auto-captions, CapCut free | Styling control |
| Stock | Pexels, Unsplash, Google Fonts | Exclusivity |
| Analytics | YouTube Studio | Nothing |
| Workflow | Google Sheets | Nothing |
| Packaging | Thumblore free tier, plus the utilities below | Volume — the free tier is a trial, not a plan |
We also keep a set of small browser utilities free and permanently uncapped, because they are the jobs that come up mid-workflow and should not require a subscription or a login:
- Thumbnail preview — see your thumbnail at real feed sizes, including 168×94, before you upload it.
- Thumbnail downloader — pull competitor thumbnails at full resolution for a swipe file.
- Image compressor — get under a size limit without visible quality loss.
- Image converter — PNG, JPG and WEBP in both directions.
- Social image resizer — one image, correct dimensions for every platform.
- Colour palette extractor — lift a palette from any image and reuse it across a channel.
No account, no watermark, no daily cap on any of them. They are utilities, not a funnel.
Part 6: Three complete stacks, three budgets
| Job | $0/month | ~$30/month | ~$100/month |
|---|---|---|---|
| Research | Free signals | Free signals | vidIQ or TubeBuddy paid tier |
| Packaging | Thumblore free tier | Thumblore Starter — $9 | Thumblore Pro — $24 |
| Editing | DaVinci Resolve | DaVinci Resolve | Premiere Pro or Resolve Studio |
| Audio | Audio Library | Uppbeat — ~$7 | Epidemic Sound — ~$15 |
| Captions | Auto-captions | CapCut Pro — ~$10 | Opus Clip — ~$29 |
| Stock | Pexels, Unsplash | Pexels, Unsplash | Envato Elements — ~$16.50 |
| Analytics | Studio | Studio | Studio |
| Total | $0 | ~$26 | ~$107 |
Notice what does not change across the three columns: analytics, and the fact that research stays free until the top tier. The money moves into packaging, sound and time — which is the correct order, because those are the three things a viewer experiences directly.
Part 7: Five tool traps
Trap 1 — Reading the price line instead of the credit line
Photoshop's 25 generative credits a month, against roughly 10 credits per generative fill, is about two operations before you wait for a reset — on a $22.99/mo subscription. Canva Pro's ~500 credits are shared across every Magic Studio feature, so iterating three to five times per concept drains them faster than the number implies. Adobe Express gives 250 at $9.99.
This is not a scandal. Inference costs money and every vendor meters it, us included. The point is that a credit is not a unit. In an editor it buys one operation inside a design you still have to finish; in a generator it buys a finished image. Compare what a credit produces, never how many there are.
Trap 2 — Paying for research before paying for packaging
Research tools are the most heavily marketed category in this market and the one with the best free alternatives. Packaging is the least marketed and the one where free options run out fastest. The recommendation flow and the value flow point in opposite directions, and most creators follow the recommendation flow.
Trap 3 — The annual discount on a tool you have used twice
Annual billing typically saves 20–35%, and it converts a reversible decision into a locked one. Pay monthly for the first two months of anything. If it survives eight weeks of real use, take the discount then.
Trap 4 — Stack overlap
Canva, Photoshop, Picsart and Express all do substantially the same job. CapCut, Descript and Premiere overlap heavily. Audit quarterly: list your subscriptions against the eight jobs in Part 4 and cancel anything that is the second entry in a row.
Trap 5 — Buying a tool to avoid a decision
No product will tell you what your channel is about. Tools compress the time between a decision and its result — they do not make the decision. If you are three tools deep and still publishing irregularly, the missing thing is a format, not a subscription.
Part 8: The weekly cadence
| When | What | Time |
|---|---|---|
| Before you script | Write the title and brief the thumbnail. If you cannot package it, do not make it. | 15 min |
| Before you film | Confirm the first 30 seconds delivers what the package promised. | 10 min |
| After the edit | Produce three thumbnail variants. | 5–45 min, depending on your stack |
| At publish | Run the twelve-item checklist in Part 3. | 5 min |
| 48 hours later | Check impressions, CTR and first-30-second retention. Change nothing yet. | 5 min |
| Day 14 | Read the A/B result. Note which variant won and why. | 10 min |
| Monthly | Rewrite the packaging on your best-performing evergreen video. | 30 min |
| Quarterly | Audit subscriptions against the eight jobs. Cancel duplicates. | 20 min |
The first row is the one that changes channels. Packaging before production means you never again finish a video you cannot sell — and the videos you abandon at the title stage are cheaper than the ones you abandon after editing.
Part 9: The diagnostic table
When a video underperforms, resist the urge to fix everything. Exactly one of three numbers is low, and which one it is determines the entire response.
| Symptom | What it means | Fix | Do not |
|---|---|---|---|
| Low impressions | YouTube does not know who to show it to, or tried and stopped | Topic and audience match. Move toward search-intent titles; stay adjacent to your existing content. | Redesign the thumbnail — it was barely shown |
| High impressions, low CTR | Shown and not clicked — a packaging failure | New thumbnail and title. Test three variants. This is the most fixable failure on YouTube. | Re-edit the video |
| Good CTR, low retention | The package overpromised, or the hook is slow | Cut the first 30 seconds hard. Deliver on the thumbnail within 15 seconds. | Make the thumbnail louder — that widens the gap |
| Good retention, no growth | The content works; nobody is finding it | Packaging and topic selection. You have a distribution problem, not a quality problem. | Add more production value |
| Views but no subscribers | No reason to come back | Format consistency and visual identity. Make the next video recognisably yours. | Ask harder in the outro |
| CTR falling over months | Your style has become invisible, or the niche moved | Compare your last ten thumbnails against the current top ten in your niche. | Change everything at once — you lose the signal |
The rule behind the table
Impressions are a topic problem. CTR is a packaging problem. Retention is a promise problem. Almost every "my video flopped" question resolves into one of those three, and treating the wrong one is why so much effort produces so little movement.
Part 10: The one-page version
If you remember nothing else from this page, remember these fifteen lines.
- Thumbnail 1280×720, 16:9, under 50 MB — but judge it at 168×94.
- Keep faces and text out of the bottom-right duration badge.
- Title caps at 100 characters and displays about 48. Promise first.
- Three chapters minimum, first at 00:00, ten seconds each.
- First 120 characters of the description are the only ones most people see.
- Tags are worth two minutes, not twenty.
- Shorts run to 3 minutes at 9:16. Frame one is the thumbnail.
- CTR of 2–10% is normal. Compare against traffic source, never against other channels.
- Aim above 70% retention at 30 seconds.
- The initial test is the first 24–48 hours, and it is quiet.
- A/B tests resolve on watch time per impression, not clicks. Overpromising loses.
- Give each variant 1,000–5,000 impressions and two weeks.
- Research is free. Packaging is not. Budget accordingly.
- One tool per job. Five of the eight jobs have a free winner.
- Package before you produce. If you cannot sell it, do not make it.
Where Thumblore fits in this stack
We have been deliberately even-handed above, so it is worth being direct about our own position rather than dressing it up as neutral advice.
Thumblore does exactly one of the eight jobs: packaging. It is not a research suite, not an editor, and it will not tell you what to make. It exists because of a specific, boring bottleneck that shows up in almost every channel's workflow — YouTube's A/B test accepts three thumbnails, and producing three thumbnails in an editor takes most of an evening, so the test gets skipped and the most fixable metric on the platform never gets fixed.
- Job it does
- Packaging — turns a video description into finished, correctly-sized thumbnails.
- Output
- Four variations per request, so a three-variant test set costs one request rather than an evening.
- Your face
- Upload a photo once and save it as an avatar, then reuse it for consistency across a channel. AI avatars available if you would rather not appear.
- Exports
- Correct dimensions for YouTube every time — no size guide, no export dialog, no 2 MB surprise.
- Free tier
- 100 credits on signup, no credit card. Enough to judge it against the thumbnail you were going to use anyway.
- What it does not do
- Research, editing, analytics, or your topic selection. Use the free tools in Parts 1, 5 and 7 for those.
The honest comparison is not against Canva's sticker price — $18/mo against $9/mo is close enough not to matter. It is against your evenings. At one video a week with three variants each, that is 156 thumbnails a year: roughly 13 hours in a generator against 65–78 hours in an editor. Whether that trade is worth it depends entirely on what an hour of yours is worth, which is a number only you have.
The cheapest experiment on this page
Take the next video you were going to publish anyway. Generate a set of variants on the free tier, put them beside the thumbnail you had planned, and run all three through Test & Compare. Let watch time per impression settle it. It costs nothing and it does not care whose tool wins.
Frequently asked questions
What is the correct YouTube thumbnail size in 2026?
1280×720 as the recommended minimum, 16:9 aspect ratio, at least 640 pixels wide, under 50 MB, in JPG, PNG, GIF or WEBP. The file size limit was raised from 2 MB in late 2025 alongside support for 4K thumbnails. The size that actually decides your click-through rate, though, is 168×94 — what it shrinks to in a mobile feed.
How long can a YouTube title be?
100 characters is the hard limit. Roughly 48–60 characters display before truncation, depending on surface and device, so treat everything past 60 as being written for search rather than for a human.
What is a good click-through rate on YouTube?
Most channels and videos sit between 2% and 10%. Search traffic often runs higher than browse traffic on the same channel, so compare a video against your own videos from the same traffic source rather than against a published average.
What tools do I actually need to start a YouTube channel?
Four: a camera or screen recorder, an editor (DaVinci Resolve, free), a music source (YouTube Audio Library, free) and something to make thumbnails. Everything else in this guide is an optimisation you can add once you are publishing consistently.
Is vidIQ or TubeBuddy worth paying for?
Both are good products with genuinely useful free tiers. Their paid tiers buy speed and convenience rather than data you cannot otherwise reach — YouTube autocomplete, Google Trends and Studio's Research tab cover most of it free. Buy one when you can name the hour per week it saves you.
Can I run a YouTube channel with entirely free tools?
Yes, for editing, audio, captions, stock, analytics and research — five of the eight jobs have a free option that is genuinely best-in-class, and YouTube Studio is the strongest analytics product in the market. Packaging is where free tiers run out fastest, because it is the one job you repeat several times per video.
How many thumbnails should I make per video?
Three, because that is what YouTube's Test & Compare accepts. Give each variant roughly 1,000 to 5,000 impressions and about two weeks before drawing conclusions.
Does the A/B test winner get picked by click-through rate?
No. Test & Compare resolves on watch time per impression. A thumbnail can win more clicks and still lose the test, which is precisely why overpromising backfires now in a way it did not a few years ago.
What are the YouTube monetisation requirements?
Two tiers. Fan funding features: 500 subscribers, 3 public uploads in 90 days, and either 3,000 valid public watch hours in 12 months or 3 million valid public Shorts views in 90 days. Ad revenue: 1,000 subscribers, and either 4,000 valid public watch hours in 12 months or 10 million valid public Shorts views in 90 days. Thresholds vary by region and change periodically.
How long can YouTube Shorts be?
Up to 3 minutes, vertical at 9:16, ideally 1080×1920. Keep text clear of the bottom fifth where the interface overlays sit.
Why does my video have views but no subscribers?
Because nothing signalled that the next video would be like this one. That is a format and identity problem rather than an outro problem — consistent packaging, a recognisable visual style and a predictable format do far more for subscription rate than asking harder at the end.
What does Thumblore cost and is there a free version?
Signup gives 100 credits free with no card required. Paid plans are $9, $24 and $45 per month for 700, 2,000 and 4,200 credits — roughly 40, 120 and 240 thumbnails respectively.
Are AI-generated thumbnails allowed on YouTube?
Thumbnails must follow YouTube's Community Guidelines and must not mislead viewers about the video's content — true whether a human or a model made them. YouTube also has evolving disclosure requirements around realistic synthetic media. Check YouTube's current guidance rather than any tool's marketing, ours included.
Sources
Platform specifications and pricing were checked in August 2026. YouTube changes product behaviour without much notice and vendors restructure pricing and AI credit allocations frequently, so confirm current terms before acting on any figure here.
- YouTube Help — Add video thumbnails
- YouTube Help — Recommended upload encoding settings
- YouTube Help — A/B test titles and thumbnails
- YouTube Help — Upload videos longer than 15 minutes
- YouTube Help — YouTube Partner Programme overview
- YouTube Help — Add video chapters
- Google Trends
- Canva pricing
- Adobe — Generative credits FAQ
- vidIQ pricing
- TubeBuddy pricing
- DaVinci Resolve
The bottom line
Nothing on this page is secret. Every spec is in YouTube's documentation, every price is on a vendor's site, and every benchmark is publicly discussed. The reason a cheat sheet is useful anyway is that the information is scattered across forty pages written by people trying to sell you something, and the act of collecting it changes what you notice.
What you notice, once it is all on one page, is how lopsided the market is. The jobs with the most marketing behind them — research, analytics, keyword tools — have excellent free options. The job with the least marketing behind it is the only one every single person in your impressions experiences, and it is the one that runs out of free options first.
Fix that imbalance and most of the rest of this page becomes maintenance. Package before you produce, test three variants, read the right one of three numbers when something underperforms, and spend your software budget where the viewers actually are.