Thumbnail Style Guide: 10 Rules + Contrast Ratios

What you paste, what you get

You paste: one paragraph about your channel or series (topic, audience, tone, typical video types or three example titles, and any colors to use or avoid).

You get: a one-page thumbnail style guide with exactly 10 numbered, reusable rules (3-color a proposed 3-color palette with hex codes and contrast ratios computed on those three colors only (not your photos), text length and size limits, one focal subject, a 3x3 composition grid with the bottom-right corner kept clear, an element budget, a series marker, what may vary and what never changes, a do-not list and a phone test), applied to three example titles on paper, cross-checked by four review lenses, plus a 5-question checklist. It plans; it does not draw or look at images. Validation and review lenses are self-checked by the same model; spot-check the numbers.

$5 one-time download. Sign in, pay once and download the SKILL.md, then run it in your own AI agent. Built for Claude Code; other agents that read SKILL.md files (such as Codex CLI or OpenClaw) should also work. No network access or extra tools needed. No subscription from us: you use your own agent and its plan.

GlowHalo is not affiliated with or endorsed by Anthropic, OpenAI or OpenClaw; the names are used only to say which agents the file is written for.

Sample output

Sample input (an invented example)
Channel: "Pocket Garden Lab", a faceless channel about growing herbs and small vegetables on apartment balconies, for beginners who have no yard. Tone: calm, friendly, practical. Video types: step-by-step how-tos, short comparisons of containers, and monthly "what to plant now" lists. Example titles: "How to Grow Basil in a Small Pot", "Plastic vs Clay Pots: Which Holds Water Longer?", "5 Herbs to Plant on a Shady Balcony". No faces, I film plants and hands only. I like greens and warm cream, and want to avoid red and neon colors.

Real, unedited output from running this Skill on a sample problem:

Planning only: this cannot see images, analytics or how any platform shows or ranks videos - nothing here predicts views, clicks or rankings.

Pocket Garden Lab (balcony herbs and small vegetables for beginners) · calm, friendly, practical · 10 rules · 3 palette colors · 3 test videos · SFW

## Essentials

- Topic: growing herbs and small vegetables on apartment balconies `(given)`
- Audience: beginners who have no yard `(given)`
- Tone: calm, friendly, practical `(given)`
- Video types: step-by-step how-tos, short container comparisons, monthly "what to plant now" lists `(given)`
- Example titles: three titles provided `(given)`
- Faces: none; plants and hands only `(given)`
- Colors to use: greens and warm cream `(given)`
- Colors to avoid: red and neon `(given)`
- Series name: "Pocket Garden Lab" `(given)`
- Title and overlay language: English `(assumed)`
- Font: a generic bold sans-serif `(assumed)`
- Character counts include spaces `(assumed)`

## Anchors and test videos

- **Audience:** beginners with a balcony and no yard.
- **Tone:** calm, friendly, practical. No shouting visuals and no clutter.
- **Video types:** how-tos, container comparisons, monthly planting lists.
- **Faces:** none. The subject is a plant, a pot or hands with a plant.
- **Colors:** use greens and warm cream. Avoid red and neon.
- **Series name:** Pocket Garden Lab.
- **Working limits:**
  - thumbnail 16:9, phone size about 168x94 px
  - text at most 4 words and 20 characters
  - capital height at least 12% of frame height per line (about 11 px)
  - text contrast at least 4.5:1, subject contrast at least 3:1
  - at most 3 elements
  - bottom-right corner kept clear for the length badge
- **Status of these limits:** planning heuristics, not platform rules. Check real previews.

**Test videos**
- **V1:** "How to Grow Basil in a Small Pot" (how-to)
- **V2:** "Plastic vs Clay Pots: Which Holds Water Longer?" (comparison)
- **V3:** "5 Herbs to Plant on a Shady Balcony" (list)

## Style guide

Grid cells are numbered in reading order: 1 2 3 / 4 5 6 / 7 8 9. Cell 9 is the bottom-right corner.

| Rule | What it says | Numbers |
|---|---|---|
| R01 | Palette: background Warm Cream #F5EEDC, text Deep Green #1F3D2B, accent Leaf Green #4A7C3A. No red, no neon. | 3 colors; L = 0.8573 / 0.0380 / 0.1618 |
| R02 | Text is always Deep Green #1F3D2B on the Warm Cream #F5EEDC background. Text never sits on a photo. | 10.31:1 (≥ 4.5:1) |
| R03* | Text: at most 4 words and 20 characters including spaces; at most 2 lines and at most 12 characters per line; ALL CAPS; bold (weight 700+) sans-serif; cap height at least 12% of frame height per line; no outline. | ≤4 words, ≤20 chars, ≤2 lines, ≤12 chars/line, cap ≥12% (~11 px at 168x94) |
| R04 | Focal subject: one plant, pot or hands-with-plant, cut out or shot against plain cream, filling at least 45% of frame height. Planned subject contrast uses Leaf Green #4A7C3A as the stand-in leaf color. Real photo colors are not covered by this number. | ≥45% height (~42 px); 4.28:1 (≥ 3:1) |
| R05 | Grid: cells 1-2 = text slot; cell 3 = series marker; cells 4, 5, 7, 8 = subject; cell 6 = empty breathing space; cell 9 = kept clear for the length badge. | text slot ~112x31 px; subject area ~112x63 px |
| R06* | Element budget: exactly 3 elements - 1 subject, 1 text block, 1 series marker. A comparison pair shot side by side in one photo counts as one subject. | 3 of max 3 |
| R07* | Series marker: a solid Leaf Green #4A7C3A rounded tab in cell 3, 20% of frame width × 12% of frame height. It sits in the same place at the same size every time, with no text or icon inside. | ~34x11 px at 168x94; 4.28:1 vs cream |
| R08 | May vary: the subject, the text words, and the subject's crop within cells 4, 5, 7, 8. Never varies: palette, text slot (cells 1-2), marker, grid, font, case. | 3 variable, 6 fixed |
| R09* | Do not: use text over 4 words or 20 characters; use red or neon; put text over a photo or busy background; add outlines or drop shadows on text; add a fourth element (arrows, emoji, badges); put text on Leaf Green. | text on Leaf Green = 2.41:1 (fails 4.5:1) |
| R10 | Phone test: view the finished thumbnail at about 168x94 px. Can you read the text and name the subject in one second? Does it still look like the series next to your last 3 thumbnails? | 168x94 px, 1 second, last 3 |

## Test on 3 videos

| Video | Overlay text (words/characters) | Elements | How it keeps the series look |
|---|---|---|---|
| V1 | "SMALL POT / BASIL" (3 words / 15 chars; lines 9 + 5) | basil in small pot, text, marker = 3 | Basil fills cells 4, 5, 7, 8 on cream; Deep Green caps in cells 1-2; Leaf Green tab in cell 3; cell 9 clear |
| V2 | "PLASTIC VS / CLAY" (3 words / 15 chars; lines 10 + 4) | plastic and clay pots side by side in one shot, text, marker = 3 | Same slots; only the pair photo and the words change |
| V3 | "5 SHADE / HERBS" (3 words / 13 chars; lines 7 + 5) | herb group in one container, text, marker = 3 | Same palette, slot, tab and clear corner; subject and words change |

## Validation

1. **Count:** 10 rules, in order R01-R10. 3 palette colors [#F5EEDC, #1F3D2B, #4A7C3A]. 3 test videos [V1, V2, V3]. PASS.
2. **[CON]:** All contrast numbers are computed from WCAG relative luminance.
   - **Linearising:** sRGB values c/255 are linearised with ((c+0.055)/1.055)^2.4.
   - **#F5EEDC:**
     - R 0.9608→0.9131, G 0.9333→0.8550, B 0.8627→0.7157
     - L = 0.2126·0.9131 + 0.7152·0.8550 + 0.0722·0.7157 = 0.8573
   - **#1F3D2B:**
     - R 0.1216→0.0137, G 0.2392→0.0467, B 0.1686→0.0242
     - L = 0.0029 + 0.0334 + 0.0017 = 0.0380
   - **#4A7C3A:**
     - R 0.2902→0.0685, G 0.4863→0.2016, B 0.2275→0.0423
     - L = 0.0146 + 0.1441 + 0.0031 = 0.1618
   - **Text #1F3D2B / background #F5EEDC:** (0.8573+0.05)/(0.0380+0.05) = 0.9073/0.0880 = 10.31:1. PASS (≥4.5).
   - **Subject stand-in and marker #4A7C3A / background #F5EEDC:** 0.9073/(0.1618+0.05) = 0.9073/0.2118 = 4.28:1. PASS (≥3).
   - **Text #1F3D2B on #4A7C3A:** 0.2118/0.0880 = 2.41:1. This would fail 4.5:1, but no rule places text on accent; R09 forbids it. PASS.
3. **[TXT]:**
   - R03 limits: ≤4 words, ≤20 characters, ≤2 lines, ≤12 characters per line.
   - Cap height: 12% of 94 px = 11.3 px.
   - Fit in the text slot: 2 lines need about 30% of the frame height, which is within the 33% text row.
   - V1 "SMALL POT BASIL": 3 words / 15 chars.
   - V2 "PLASTIC VS CLAY": 3 words / 15 chars.
   - V3 "5 SHADE HERBS": 3 words / 13 chars.
   - All lines are ≤12 chars.
   - PASS.
4. **[LAY]:** text in cells 1-2, marker in 3, subject in 4, 5, 7, 8, cell 6 empty, cell 9 clear. No cell is used twice. PASS.
5. **[BUD]:** V1 3, V2 3, V3 3, marker included. PASS.
6. **[SER]:** each test video keeps the palette, text slot, marker and grid; only the subject and words change (see test table). PASS.
7. **[SAFE]:** no named creator, brand or character; no deception rule; SFW. PASS.
8. **[NOIMG]:** no rule claims the Skill saw or measured an image. R04 states that photo colors are not covered. PASS.

**Changes in Phase 3:** no changes needed. Phase 4 changes are listed under Changes applied, and checks 1-8 were re-run on the final guide.

### Validation table

| Check | Result | Numbers |
|---|---|---|
| Count | PASS | 10 rules; 3 colors; 3 videos |
| [CON] | PASS | text 10.31:1; subject/marker 4.28:1; text-on-accent 2.41:1 forbidden by R09 |
| [TXT] | PASS | V1 3/15, V2 3/15, V3 3/13; max line 10 chars; cap ~11 px |
| [LAY] | PASS | text 1-2, marker 3, subject 4/5/7/8, empty 6, clear 9 |
| [BUD] | PASS | V1 3, V2 3, V3 3 |
| [SER] | PASS | palette, slot, marker, grid fixed in all 3 videos |
| [SAFE] | PASS | no names, no deception, SFW |
| [NOIMG] | PASS | no image claims |

## Review lenses

### Findings per lens

**Lens 1 - Thumb-Size Reader**
- **1a [R03]:** The draft says "at most 4 words and 20 characters including spaces; at most 2 lines". A 20-character line does not fit the ~112 px slot at ~11 px bold caps (about 8-9 px per character, so about 12-13 characters). Proposed fix: add a per-line limit of 12 characters.

**Lens 2 - Contrast Inspector**
- **2a [R09]:** The draft do-not list does not mention text on Leaf Green. Deep Green on Leaf Green computes to 2.41:1, below 4.5:1, and the marker tab invites putting a label on it. Proposed fix: add "put text on Leaf Green" to R09.
- **2b [R04]:** R04 says the subject contrast "uses Leaf Green #4A7C3A as the stand-in leaf color". Pale subjects such as clay pots in V2 or light-colored herbs may sit close to cream. Proposed fix: require a 2 px Deep Green outline around pale subjects.

**Lens 3 - Series Consistency Scout**
- **3a [R07]:** The draft says "a solid Leaf Green #4A7C3A rounded tab in cell 3, 20% of frame width × 12% of frame height". It does not say the tab is identical every time, so it could drift into labels or icons. Proposed fix: add "same place at the same size every time, with no text or icon inside".

**Lens 4 - Fit & Claims Auditor**
- **4a [R06]:** The draft says "1 subject". The comparison video type (V2) needs two pots, which could be read as two subjects and break the budget. Proposed fix: state that a comparison pair shot side by side in one photo counts as one subject.
- **4b [R03]:** "ALL CAPS" reads slightly louder than a calm tone. However, bold caps at ~11 px are a legibility choice, and the words stay short and plain. No change proposed.

### Cross-check

- `[1a, R03]` Lens 2: no challenge / Lens 3: no challenge / Lens 4: no challenge
- `[2a, R09]` Lens 1: corroborate - "text on the 34x11 px tab would also be too small to read at phone size" / Lens 3: no challenge / Lens 4: no challenge
- `[2b, R04]` Lens 1: challenge - "a 2 px outline at a ~1280 px-wide original shrinks to about 0.26 px at 168 px wide and is invisible at phone size; it also adds a style R09 keeps off text" / Lens 3: no challenge / Lens 4: corroborate - "V2 clay pots are a stated video type, so pale subjects will recur"
- `[3a, R07]` Lens 1: no challenge / Lens 2: no challenge / Lens 4: no challenge
- `[4a, R06]` Lens 1: no challenge / Lens 2: no challenge / Lens 3: corroborate - "comparisons shot as one pair photo keep the same subject area as the other types"
- `[4b, R03]` Lens 1: no challenge / Lens 2: no challenge / Lens 3: no challenge

### Verdict

| Finding | Tag | Action |
|---|---|---|
| 1a [R03] | CONFIRMED | Add the 12-characters-per-line limit. |
| 2a [R09] | CORROBORATED | Add the no-text-on-Leaf-Green prohibition. |
| 2b [R04] | DISPUTED | See both sides below; R04 is left as written. |
| 3a [R07] | CONFIRMED | Fix the marker's place, size and content. |
| 4a [R06] | CORROBORATED | Count a comparison pair in one photo as one subject. |
| 4b [R03] | NO CHANGE | — |

On 2b, the two sides are:
- **For an outline:** pale subjects such as clay pots will recur and may blend into cream.
- **Against:** a 2 px outline is invisible at phone size and adds a style element.

An alternative to weigh is choosing a darker pot or angle when shooting. The user decides.

4 rules changed [R03, R06, R07, R09]. 6 findings [1a, 2a, 2b, 3a, 4a, 4b].

### Changes applied

- **R03 (1a)**
  - Before: "Text: at most 4 words and 20 characters including spaces; at most 2 lines; ALL CAPS; bold (weight 700+) sans-serif; cap height at least 12% of frame height per line; no outline."
  - After: "Text: at most 4 words and 20 characters including spaces; at most 2 lines and at most 12 characters per line; ALL CAPS; bold (weight 700+) sans-serif; cap height at least 12% of frame height per line; no outline."
  - The Numbers column also gained "≤12 chars/line".
- **R06 (4a)**
  - Before: "Element budget: exactly 3 elements - 1 subject, 1 text block, 1 series marker."
  - After: "Element budget: exactly 3 elements - 1 subject, 1 text block, 1 series marker. A comparison pair shot side by side in one photo counts as one subject."
- **R07 (3a)**
  - Before: "Series marker: a solid Leaf Green #4A7C3A rounded tab in cell 3, 20% of frame width × 12% of frame height."
  - After: "Series marker: a solid Leaf Green #4A7C3A rounded tab in cell 3, 20% of frame width × 12% of frame height. It sits in the same place at the same size every time, with no text or icon inside."
- **R09 (2a)**
  - Before: "Do not: use text over 4 words or 20 characters; use red or neon; put text over a photo or busy background; add outlines or drop shadows on text; add a fourth element (arrows, emoji, badges)."
  - After: "Do not: use text over 4 words or 20 characters; use red or neon; put text over a photo or busy background; add outlines or drop shadows on text; add a fourth element (arrows, emoji, badges); put text on Leaf Green."
  - The Numbers column also gained "text on Leaf Green = 2.41:1 (fails 4.5:1)".
- **Re-run:** Phase 3 was re-run on [R03, R06, R07, R09]. All checks pass (see Validation table).

## Checklist

1. Is the text 4 words or fewer, 20 characters or fewer, and 12 characters or fewer per line? **R03**
2. Is the text Deep Green on cream, in cells 1-2, and nowhere on the photo or the green tab? **R02, R09**
3. Is there exactly one subject (or one side-by-side pair photo) filling at least 45% of the height, with nothing else added? **R04, R06**
4. Is the Leaf Green tab in cell 3 at the same size, and is the bottom-right corner empty? **R07, R05**
5. At about 168x94 px, can you read the text and name the subject in one second, and does it match your last 3 thumbnails? **R10**

**Still your call:** R04 [2b]. Decide whether pale subjects such as clay pots get an outline, a darker shooting angle, or nothing.

## Notes

- **Assumptions to confirm:**
  - English overlay text
  - a generic bold sans-serif font
  - character counts include spaces
  - a 45% minimum subject height
  - ALL CAPS (4b noted it reads slightly louder than "calm"; switch to sentence case if you prefer)
- **How to use:**
  - Apply the guide to your next 3 thumbnails.
  - View each one at about 168x94 px and next to your last 3.
  - Change one rule at a time.
- **Honest limit:** contrast numbers cover the three palette colors only, not your photos. Nothing here predicts views or clicks.
- **Next options:** I can write a guide for a second series, a stricter text rule, or a check of a thumbnail brief you already have.

Not affiliated with or endorsed by YouTube or Google.

Use cases

• Making a channel's thumbnails look like one family instead of a new design every week
• Giving a freelance designer or an AI image generator clear, checked rules to follow
• Catching text that is too long and colors that blur together at phone size before you make images
• Keeping a style consistent across a new series or a rebrand

FAQ

Does it generate or look at images? No. It plans rules from your description and checks the numbers. View real thumbnails at about 168x94 px yourself.

Will it raise my views or click-through rate? No promise either way. It checks legibility, contrast numbers and consistency of the rules only.

Are the size and contrast limits platform rules? No. They are planning heuristics. Check real previews on a phone and on a desktop list.

Is it SFW only? Yes, and it will not copy a named creator's look or add real brands.

Not affiliated with or endorsed by YouTube or Google; platform names are used only to describe what the plan is for.

How it works (in detail)

Thumbnail Style Guide for Your Channel: 10 Rules, Palette & Contrast turns a short description of a channel or series into one reusable thumbnail style guide, then checks it before you make any image. It plans and audits; it does not draw or inspect images.

Phase 0 - Intake. Needs the channel topic and audience, typical video types or three example titles, and the tone. Asks at most 3 questions, only if one is missing.

Phase 1 - Anchors and test videos. Audience, tone, colors, and the working limits (phone size about 168x94 px, 4 words and 20 characters of text, 12% capital height, 4.5:1 text contrast, 3 elements, clear bottom-right corner), plus three test videos.

Phase 2 - Ten rules. Palette, text color, text size and length, focal subject, composition grid, element budget, series marker, what may vary, do-not list, phone test.

Phase 3 - Mechanical validation. Count, contrast ratios computed on the proposed palette (palette colors only, not your photos), text limits per test video, layout cells, element budget, series test, safety, and a validation table.

Phase 4 - Four review lenses. Thumb-Size Reader, Contrast Inspector, Series Consistency Scout, Fit & Claims Auditor, applied in turn by one model. Each finding is cross-checked by the other three lenses and tagged CONFIRMED, CORROBORATED, DISPUTED or NO CHANGE; fixes are applied and listed.

Phase 5 - Checklist. Five yes/no questions tied to rule ids, plus any rule that is still your call. Strictly SFW. No promises about views or click-through rate.