Sign In

YouTube Thumbnail generator

Type

Workflows

Stats

72

0

Reviews

Published

Jan 26, 2026

Base Model

Flux.2 Klein 9B

Hash

AutoV2
A85ED313D8
default creator card background decoration
Krug_Lov's Avatar

Krug_Lov

The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.

IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.

🎬 Create cinematic YouTube thumbnails from any image — screenshot or photo.
Change genre, background and even facial expression while keeping the character consistent.

Don't judge me too harshly, this is my first experience in creating a node.

🎬 YouTube Thumbnail Studio

Cinematic prompt generator for image editing models

(Optimized for Qwen-Image-Edit, Flux.2 Klein )

🧠 What this node does

YouTube Thumbnail Studio generates precise, structured prompts for image-editing models, focused on:

🎭 controlled facial emotion changes

🎬 cinematic backgrounds by genre

🧍 character consistency (identity, pose, scale)

🎲 reproducible or random results via seed control

Designed and tuned primarily for Qwen-Image-Edit, possibly works with Flux Kontext.

Version 2, adapted for Flux.2 Klein 9B distilled, works much faster than Qwen Image Edit.

🎛️ Input Parameters

🎞️ genre

Select the cinematic genre of the scene (14+1)

Defines the overall visual language and background style.

Examples:

Action film

Adventure film

Horror film

Sci-Fi film

Retro Sci-Fi

Western

Musical

Claymation

➡ Used to select a curated cinematic background description.

🎭 emotion (30 pieces, but not all work perfectly.)

Facial expression to apply to the character

Controls the emotional state of the main subject’s face

(e.g. eyes, mouth tension, brows, micro-expressions).

Examples:

😊 happy

😠 angry

😬 awkward

😖 nervous

😈 playful mischief

➡ Emotion descriptions are internally expanded for better model understanding.

🎭 emotion_only

Emotion-only edit mode (advanced)

🔹 When enabled:

Only the facial expression is changed

Everything else is preserved as strictly as possible:

identity

pose

clothing

lighting

background

composition

Use this when you want a clean emotion swap without scene changes.

🧍 keep_character

Preserve character identity and pose

When enabled:

Keeps face, body, pose, scale, and clothing consistent

Allows background and lighting to change

Emotion change is subtle and controlled

Best for:

thumbnails

posters

character-centric edits

📝 include_text_elements

Add cinematic text elements

Adds:

a short movie-style title

optional tagline

small credit line

Use this for:

YouTube thumbnails

poster-style visuals

🎬 Background Control

🎞️ background_variant

Choose a specific background variant

random → random background from selected genre

0–29 → exact background index

Useful when:

you want consistency

you found a variant you like

🔒 lock_background

Lock the background between generations

When enabled:

The same background is reused

Emotion, seed, or other parameters can change

Great for:

A/B testing expressions

emotion comparisons

🔁 reroll_background

Force background reroll

Overrides background locking and forces a new selection.

🎲 Seed Control

🎲 seed_mode

Random or fixed seed

random → new result every run

fixed → reproducible output

🧬 fixed_seed

Manual seed value

Used only when seed_mode = fixed.

Range: 0 – 999999

✏️ Optional Overrides

🎨 custom_background_elements

Manual background description

Overrides genre background completely.

Example:

abandoned space station corridor with flickering red lights

📝 custom_title

Custom title text

Replaces auto-generated cinematic title.

📝 About Text & Titles (Important Limitation)

Qwen Image Edit (including 2511) is capable of understanding and generating text inside images, and in many cases it places titles or words correctly.

However, please note:

❌ Precise font selection is not supported

❌ Exact text positioning cannot be controlled

❌ Typography styles (kerning, line breaks, alignment) are not editable

❌ Consistent branding text is not reliable

This is a model limitation, not a node limitation.

🧠 extra_instruction

Additional prompt instructions

Appended at the end of the generated prompt.

Installation, Copy the file into: ComfyUI/custom_nodes/

The archive contains a node and a simple workflow.

Don't judge me too harshly, this is my first experience in creating a node.

☕ This node was fueled by coffee.

You can help refill it 🙂 buymeacoffee.com/Krug_Lov