Creator Studio

Channels

Analytics

Views

0

Watch Time (Min)

0

Avg View Duration

0:00

Ave View Percent

0%

Subs Gained

+0

Subs Lost

-0

Likes

0

Dislikes

0

Comments

0

Shares

0

Playlist Adds

0

Est Revenue

$0.00

Loading analytics...

Content

Loading videos...

1

Upload Defaults

Set standard default metadata that applies across your channel videos.

Saved

Title 0/100

Description 0/5000

Visibility

Tags 0/500

1

Create Directory

Create folder directory for video assets.

Folder Sequence

Numerical

Save Folder

Asset Folder(s)

2

Generate Folders

Generate folders in directory.

Generate Folders status will appear here...

1

Scan Channel

AI will scan channel information to understand what it is about.

Input

Description Keywords Audience Videos

2

Scan Competitors

Choose up to 6 competitor channels. AI will scan channel information from each.

Input

Description Keywords Videos

3

Scan Trends

Trending

Keywords

YouTube

Fetch the top viral "Trending Now" searches or related Keywords on Google Trends.

Loading searches...

Location

Time

Language

Category

Query

Location

Time

Source

Category

Language

Timezone

Location

Category

Max Results

Scan Trends status will appear here...

4

Search Sources

Bing

URL

File

Scan trending news, a URL, or a file for articles to select for writing script.

Input

Trends

Search Sources status will appear here...

5

Generate Ideas

Chat with AI to help with idea generation.

Saved!

Input

Channel Competitors Trends Sources

Prompt

6

Generate Titles

AI will generate 10 video ideas based on channel, competitor, and source information.

Input

Channel Competitors Trends Sources Ideas

Format

Default Listicle How, What, When, Why, Where, Who News Headline

Generate Titles status will appear here...

1

Upload Reference

Upload reference script. AI will scan it to use as template for writing new script. (optional)

2

Generate Script

AI will write new script based on selected video idea and reference script. (if available)

Input

Channel Title Sources Ideas Reference

Words

Min

Max

0 words

Generate Script status will appear here...

4

Generate Data

AI will scan finished script to produce video metadata.

×

Generate Data status will appear here...

5

Save Script

Choose version of finished script to save.

1

Scan Channel

AI will scan channel information to understand what it is about.

Input

Description Keywords Audience Videos

2

Upload Script

Upload script that will be used as reference for image search.

3

Search Images

Google

Library of Congress

AI will scan script and search for 4 images per sentence. Select best image per sentence once scan is complete.

Size

Search Images status will appear here...

4

Create AI Images

Colorize

AI will create images based on searched images and settings.

Cost: $0.00

Source Folder

References Folder

Prompt

Quality

Background

Image Size

Resolution

x

Output Format

Number of Images: 1

Concurrency Limit: 1

Create AI Images status will appear here...

5

Process Images

This will process selected images to desired size and format.

Image Size

Resolution

x

Gaussian Blur

Brightness: 25%

Output Format

Quality: 95%

Process Images status will appear here...

1

Upload Data

Upload video metadata. This will be used as reference for chapters.

2

Upload Script

AI will use script as reference while editing voiceover.

3

Create AI Voiceover

AI will create voiceover based on script. (optional)

Scene

Sample Context

Audio Profile

Voice

Director's Note: Style

Director's Note: Pace

Director's Note: Accent

Temperature 1.00

Voice

Speed 1.00

Slower Faster

Stability 0.50

More variable More stable

Similarity 0.75

Low High

Style Exaggeration 0.00

None Exaggerated

Output Format

Language Override

Use Speaker Boost

Preview Voice Text

Loading credits...

0 / 5,000 characters

Preview Audio

0:00 / 0:00

Warm Up Text

Pronunciation Rules

Add words and their phonetic replacements.

→

AI voiceover status will appear here...

4

Upload Voiceover

Upload raw voiceover audio file for processing.

5

Process Voiceover

AI will process voiceover to add pauses according to script and create timing file.

Beginning Pause (ms)

End Pause (ms)

Sentence Pause (ms)

Paragraph Pause (ms)

Attack Buffer (ms)

Tail Buffer (ms)

Low Pass Filter (Hz)

Filter Strength (Passes)

Process Voiceover status will appear here...

1

Scan Channel

AI will scan channel information to understand what it is about.

Input

Description Keywords Audience Videos

2

Upload Script

Upload script that will be used as reference for video search.

3

Upload Timing

Upload timing file that will be used as reference for video search clip lengths. (optional)

4

Search Videos

YouTube

Internet Archive

C-SPAN

AI will scan script and search for 4 videos per sentence. Select best video per sentence once scan is complete.

Video Search status will appear here...

5

Create AI Videos

AI will create video based on script. (optional)

Scene

Sample Context

Video Profile

Style

Director's Note: Lighting

Director's Note: Camera Pace

Director's Note: Lens

Creativity 1.00

1

Scan Channel

AI will scan channel information to understand what it is about.

Input

Description Keywords Audience Videos

2

Upload Script

AI will use script as reference for creating music.

3

Upload Timing

AI will use voiceover timing as reference for creating music.

4

Create AI Music

AI will create music based off settings.

Save Folder

Instrumental

Length

AI will create unique music tracks for each paragraph of voiceover.

Prompt

Lyrics

AI music status will appear here...

Audio Preview

0:00

Saved Tracks

1

Upload End Screen

Upload image for video end screen.

2

Upload Data

Upload video metadata. This will be used for text graphics reference.

3

Upload Script

AI will use script as reference while editing video.

4

Upload Voiceover

Upload final voiceover audio file.

5

Upload Images

Choose folder that contains images for video edit.

6

Upload Music

Choose folder that contains music for video edit. (optional)

7

Edit Video

This will edit video using script, images, and voiceover.

Graphics

Title Headers Quotes Date

Edit Video status will appear here...

1

Upload Data

Upload video metadata. This will be used for reference.

2

Upload References

Select up to 9 reference images for the AI to emulate. (optional)

3

Search Images

Google

URL

File

AI will scan data and search for 4 images for both thumbnail background and foreground. Select best images once scan is complete. (optional)

Size

Background:

Foreground:

Background:

Foreground:

Search Images status will appear here...

4

Generate Titles

AI will generate 10 thumbnail titles based on video title. (optional)

Generate Titles status will appear here...

5

Generate Thumbnail

AI will generate thumbnail based on selections above.

Input

Title Data

Resolution

Quality

Generate Thumbnail status will appear here...

6

Save Thumbnail

Choose where to save thumbnail.

1

Upload Media

Select or verify the rendered video file and custom thumbnail image for this upload.

Upload Video

No video selected

No video selected

Upload Thumbnail

No thumbnail selected

No thumbnail selected

2

Upload Data

Configure video title, description, and tags.

Title (required) 0/100

Description

Hashtags Chapters Defaults

0/5000

Tags

Defaults

0/500

3

Apply Settings

Configure publishing visibility, playlists, category, audience compliance, disclosures, licensing, and discussion settings.

Visibility

Add to Playlist

Category

Publish Date

Publish Time (Local)

Audience (COPPA Compliance)

No, it's not made for kids Yes, it's made for kids

Paid Promotion

No paid promotion Yes, contains paid promotion

Altered or Synthetic Content (AI Disclosure)

No Yes

Video & Audio Language

License

Recording Date

None

Distribution & Engagement

Publish to subscriptions feed and notify subscribers Allow embedding on external websites Show how many viewers like this video

Subtitles

Upload auto-generated Whisper subtitles (.srt)

Comments

Post first pinned discussion comment

4

Insert Ads

Playwright Headless

Automatically insert mid-roll ad breaks at natural paragraph boundaries in YouTube Studio.

Spacing:

Timestamps: 0 breaks

No ad cues loaded. Render a video or select files to preview.

No manual timestamps added yet. Type a timestamp above and press Enter.

5

Insert End Screen

Playwright Headless

Configure and size YouTube End Screen elements to promote your next video during the outro.

Start Time:

End Time:

Duration: --s

Interactive 16:9 Canvas: Drag & resize elements (YouTube Studio layout)

Snap to Grid Snap to Element Show Grid

Select a video in Step 1 to preview your live outro frames

OUTRO

00:00

--:-- / --:--

                                                        X: --% | Y: --% | W: --% | H: --%
                                                    

Elements: 2/4 elements

6

Publish Video

Execute complete YouTube publishing pipeline with automated asset uploads, metadata, ads, and end screen.

0% 0 MB / 0 MB • 0 MB/s • ETA: 0s

Upload status will appear here...

Add Channel

Enter a name for this prompt

Confirm

Are you sure?

Delete Channel

Are you sure you want to remove this channel?

Notification

AI Status

Checking API keys and discovering supported models...

Loading models...

Google Status Anthropic Status OpenAI Status ElevenLabs Status Fal Status

Design Custom Voice

Voice Prompt

Sample Text

Preview Audio

0:00 / 0:00

Save Voice As

Generate Thumbnail

Steps:

Asset Collection: The system collects the target Title provided in the UI, any chosen Reference Images (Background, Foreground, or General), and extracts reference context from the active Data file based on your selected setting (either the Video Title or the entire Data file).
Model Preparation: The system securely packages these image and text assets and routes them to the currently active LLM model.
AI Prompt Injection: Depending on the selected references and context settings, the system injects the following exact instruction blocks into the model's prompt (you can edit and save your custom prompt here):

If Background or Foreground images ARE selected:

If NO Background or Foreground images are selected:

Additional Context (Based on Selected Input):

If Title is selected: "For reference the actual full video is titled: '[Video Title extracted from Data file]'"

If Data is selected: "For reference here is the full video data:
[Full Data file content]"
Native Generation: The active model processes the prompt instructions natively using its multimodal image generation capabilities.
Delivery: The generated thumbnail variations are returned directly to the UI gallery for preview, download, or regeneration.

Generate Titles

Steps:

Asset Collection: The system collects the active Video Title extracted from the loaded Channel Data file.
Model Preparation: The system securely packages the Video Title and routes it to the currently active LLM model.
AI Prompt Injection: The system injects the following exact instruction block into the model's prompt (you can edit and save your custom prompt here):

System Instructions:
Native Generation: The active model processes the prompt instructions and returns the array of 10 structured titles formatted in JSON.
Delivery: The generated titles are formatted as draggable items and populated in the UI gallery for preview and insertion into the thumbnail.

Search Images

Steps:

Script Parsing: The system splits your uploaded script into individual sentences (protecting abbreviations, titles, and numbers).
AI Search Query Generation: The active model formulates targeted search terms for the active image source toggle (you can customize and save instructions for Google and Library of Congress independently):

Google Query Generation Instructions:

Library of Congress Query Generation Instructions:
Image Retrieval & Master Scans: Downloads 20 candidate images per sentence. For Library of Congress, requests master archival scans and searches across the digital collections.
AI Vision Quality Control: Evaluates image candidates to choose and rank the 4 best matching visual assets per sentence.
Selection, Flags, Crop, Grayscale & Colorize: Review candidate images for each sentence. Select your preferred candidate image, click the Flag button to mark sentences you want to revisit, enable Crop to configure framing with the interactive ratio badge (cycles: 16:9, 1:1, 9:16, Smart AI face/saliency detection, or Custom freeform crop), enable Grayscale to preview and export pure black & white imagery in Process Images, and enable Colorize to mark black & white photos for colorization in the Create AI Images function.

Process Images

Steps:

Image Selection & Output Destination: Collects all selected candidate images (or their colorized variants), Crop settings (16:9, 1:1, 9:16, Smart, Custom), and Grayscale selections from the Search Images outputs, then prompts you to choose a destination save folder on your computer.
Grayscale Pre-Processing & Custom Crop: If Grayscale was selected on an image, it is converted to pure black-and-white first. If Crop was enabled, it is framed according to the chosen mode: automated OpenCV face/saliency detection for Smart, or the exact bounding box for 16:9, 1:1, 9:16, and Custom.
Canvas Composition & Export: Centers and scales each image onto your selected canvas resolution (e.g. 1080p FHD, 4K UHD, 9:16 Portrait) with optional Gaussian blur background padding, brightness adjustment, format selection (JPEG/PNG), and quality compression.

Create AI Images

Steps:

Source & References Selection: Choose a Source Folder (or let it auto-detect from Search Images) and an optional References Folder for style transfer, consistent character art, or compositional guidance.
Automated Colorize Mode vs. Standard Generation:
- Colorize Mode: When any image is selected for Colorize in Search Images, the Colorize toggle enables automatically. In this mode, Create AI Images isolates and processes only those designated candidate images.
- Standard Generation Mode: When Colorize is not active, the system generates new AI images for all images in the source folder according to your custom prompt and reference settings.
Black & White Pre-Processing: In Colorize mode, before each image is uploaded to the AI, it is automatically converted into a pure 100% grayscale (black-and-white) photo to eliminate existing color cast and ensure authentic color reconstruction.
AI Model Processing & Instructions: Images are processed concurrently up to your specified Concurrency Limit using the chosen AI model with the following customizable instruction sets:

AI Colorization Instructions:

AI Image Generation Instructions:
Live UI Sync & Dynamic File Updates: Colorized images are saved to your designated output directory with a c suffix (e.g., 0-2_3 c.jpg) and mirrored back to the source folder. The Search Images page updates immediately in real-time, loading the colorized preview, setting the clickable filename link to the colorized image, showing the Colorized label with the round green checkmark, and allowing instant before/after toggling.

Search Thumbnail Images

Steps:

Asset Collection: The system collects the active Video Topic extracted from the loaded Channel Data file.
Subject Extraction: The system prompts the active LLM to analyze the data file and identify 3 distinct core subjects for thumbnails (you can edit and save your custom prompt here):

Subject Extraction Instructions:
Query Generation: For each of the 3 subjects, the system prompts the LLM to generate precise, tangible Google Image search terms for both a Foreground image and a Background image:

Query Generation Instructions:

Background Asset Rule:

Foreground Asset Rule:
Image Sourcing: The system interfaces with the Google Custom Search API to retrieve the top 4 image results for every generated query, downloading a total of 12 Background and 12 Foreground image options.
Delivery: The downloaded images are processed, formatted, and loaded into the UI grid for review and selection.

Search Videos

Steps:

Script & Timing Parsing: The system splits your uploaded script into sentences (with honorifics/abbreviations protected). If a timing file is uploaded in Step 3, the exact spoken duration of each sentence is computed.
AI Search Query Generation: The active model formulates targeted search terms for the active video source toggle (you can customize and save instructions for each platform individually):

YouTube Query Generation Instructions:

Internet Archive Query Generation Instructions:

C-SPAN Query Generation Instructions:
Cascading Candidate Discovery: The search engine indexes 20 candidate videos tailored to the selected toggle (YouTube, Internet Archive, or C-SPAN) with a multi-tier query cascade that automatically extracts core subjects and eliminates noise.
Concurrent AI Video Watching: The top candidate videos are watched simultaneously in Google Cloud by Gemini to extract the single best clip matching the sentence duration:

Video Analysis & Clip Extraction Instructions:
Distinct Sourcing & Fallback: The engine selects 1 best clip from each of the 4 distinct candidate videos. If any candidate lacks suitable footage, it automatically analyzes candidates 5 through 20.
Parallel Stream Slicing: Slices the highest native resolution (up to 4K) directly from the stream using yt-dlp, FFmpeg, and sub-second keyframe precision.
Visual Grid Presentation: Probes resolution metadata and renders the 4 diverse video previews into the UI with clickable file links and radio selection controls.

Create AI Voiceover

Steps:

Script Processing: The system loads the target script and automatically chunks it into smaller, overlapping sections to bypass API maximum limits while maintaining context.
Configuration Mapping: The chosen voiceover settings (Scene, Sample Context, Audio Profile, Style, Pace, Accent, Voice, and Temperature) are compiled into a strict directive prompt for the active model (you can edit and save your custom prompt here):

System Instructions:
Native Audio Synthesis: The system uses the active model to natively generate high-fidelity, expressive speech audio for each script chunk based on the exact performance characteristics directed.
Chunk Assembly: The system stitches the generated audio chunks together, injecting short natural pauses between chunks and appending a brief tail of absolute silence to prevent clipping.
Delivery: The finished voiceover file is saved to the working directory and attached to the project.

Process Voiceover

Steps:

Master Script Parsing: The system processes the master text script and groups it into distinct sentence and paragraph tokens based on punctuation.
Voiceover Transcription: The system uses the active WhisperX small model to transcribe the raw voiceover audio and performs phoneme-level alignment for precise sub-second timestamps.
Timestamp Alignment: The parsed script tokens are aligned to the exact audio timestamps using Character Volume Matching to establish precise cut points.
Auto-Pacing & Slice Processing: The system extracts each sentence chunk individually and applies custom Start Padding (Attack buffer), an End Buffer (Tail safety net), an Anti-Pop shield (100ms fade in/out), and Per-Chunk Volume Normalization (-1.0 dBFS).
Assembly & Filtering: The system stitches the normalized chunks together while inserting precise digital silence (milliseconds) between sentences and paragraphs based on the user's settings. A high-quality Low-Pass Butterworth filter is applied to remove high-pitch AI artifacts.
Delivery: The finalized audio track and a JSON timing map file are saved to the project directory.

Edit Video

Steps:

Script & Data Parsing: The system parses the master script and Data file, grouping paragraph tokens and extracting the finalized Video Title and Chapter Headers.
Audio-Visual Alignment: The system aligns the script sentences to the exact timestamps provided by the JSON timing map. If the timing map is missing, it dynamically loads the Whisper Base model to transcribe and re-align the voiceover track.
Timeline Assembly: Visual assets (backgrounds, foregrounds, and endscreens) are dynamically sequenced onto the timeline. Advanced keyframe animations (like ease-out quadratic zooming and 15-frame slide-ins) and visual transitions (crossfades, fade-ins) are applied.
Graphic Overlays: Custom text overlays (Main Title and Chapter Headers) are synthetically rendered using Pillow. The system generates dynamic drop shadows, rounded corner shapes, and applies left-to-right mask wipe animations before compositing them onto the video.
Audio Compositing: Any provided background music tracks undergo automated LUFS loudness normalization (LRA=11, TP=-1.5) and are concatenated, volume-matched (7%), and mixed seamlessly beneath the voiceover track.
Rendering & Delivery: The final composition is rendered using the libx264 codec in the bt709 color space and delivered as an MP4 file to the project directory.

Generate Script

Steps:

Context Assembly: The system compiles the active Channel Information document, Target Title, Ideas Info, Sources Info, and Reference Script into a massive context window for the AI.
Instruction Formatting: The AI is constrained by strict formatting, structural, and stylistic rules passed via the system prompt. The prompt dynamically adapts depending on which optional context documents you attach (you can edit and save your custom prompt here):

Base Instructions (Always Sent):

If Reference Script is attached:

If Sources Info is attached:

If Ideas Info is attached:
Script Generation: The system streams the script in real-time from the active model and renders it as markdown into the editor.

Scan Trends

Steps:

Mode & Parameter Configuration: Choose between two market research engines:
- Trending Mode: Scans real-time breakout searches breaking across Google Trends. Configure Location (e.g. Worldwide, United States, United Kingdom, Canada, Australia), Time window (Past 4 hours, Past 24 hours [default], Past 48 hours, Past 7 days), Language (hl), and Category (All, Business, Entertainment, Health, Sci/Tech, Sports, Top Stories).
- Keywords Mode: Analyzes interest over time and breakout sub-queries for specific target keywords. Configure Query, Location, Time range (Past Hour to 2004 to present), Source (Web, Image, News, Shopping, YouTube Search), Category, Language, and Timezone.
Quota & Rate Limit Verification: The system verifies your connected SerpApi account search balance (total_searches_left / searches_per_month) and displays your live monthly usage counter ({used} / {total} Searches).
SerpApi Google Trends Execution:
- Trending Now Search: Queries SerpApi's google_trends_trending_now engine to extract breakout topics, search volume metrics (e.g. 500K+ searches), active topic categories, and associated news headlines with pagination support (Load More).
- Keyword Timeseries & Related Queries: Queries SerpApi's google_trends engine (TIMESERIES & RELATED_QUERIES) to compute interest averages and breakout search queries.
Interactive Display & Selection: Renders trending topics, query scores, and related breakout keywords with interactive checkboxes/radios.
Project Sync & Multi-Step Downstream Delivery: Selecting any trend saves it to project state (selected_trend) and automatically powers downstream features:
- Search Sources (Step 4): Seeds Bing News RSS to retrieve real-time news articles related to the trending topic.
- Generate Ideas Chatbot (Step 5): Supplies the trending topic into the AI brainstormer for relevant video angles.
- Generate Titles (Step 6): Embeds the trending keyword into the title ranking engine to generate 10 high-CTR video concepts.

Search Sources

Steps:

Source Mode Selection: Configure the research source via the mode switches:
- Bing: Searches Bing News RSS for real-time coverage matching the selected trending keyword from Scan Trends.
- URL: Connects to any specific online news article, blog post, or web page.
- File: Loads a local document or script file directly from your machine.
Search Execution & Content Retrieval:
- Bing News Aggregation: Fetches multi-page RSS results (up to 25 articles), decodes true destination URLs, filters out video platforms (YouTube, TikTok, Vimeo), and sorts articles chronologically.
- URL Reader Extraction: Queries the target URL through Jina Reader to bypass web firewalls and extract article title, publisher, and timestamp.
- Local File Reading: Ingests the selected local file and extracts its filename and metadata.
AI Prompt Injection: When AI-powered research is used, the system injects the following base system instruction into the model (you can edit, reset, or set your custom prompt as official default here):

System Instructions:
Article Selection: Results populate the interactive sources list with publication dates, domain names, headlines, and outbound links. Select the desired article to download.
Extraction & Project Storage: Clicking Download parses the full article text via Jina Reader, strips ads/navigation boilerplate, and saves a clean text file as {Channel Number} Article {hash}.txt into the channel's Sources folder (e.g. 1 Sources) for immediate use in Script and Idea generation.

Generate Ideas

Steps:

Context & Source Assembly: The system dynamically compiles all active input sources enabled via checkboxes (Channel Information, Competitor Data, Trending Topics/Keywords, and Scanned Sources/Articles), alongside any attached file, document, or image and the multi-turn conversational chat history.
Model Routing: The packaged context, prompt instructions, and media attachments are securely routed to the currently active LLM model (e.g. Gemini 3.1 Pro Preview).
AI Prompt Injection: The system injects the following base system instruction into the model's configuration (you can edit, reset, or set your custom prompt as official default here):

System Instructions:
Conversational Generation: The active model processes your prompt alongside the assembled context and conversational memory, producing high-impact video concepts, angles, and content recommendations.
Delivery & Persistence: The generated response is rendered directly into the chat stream and automatically saved to the channel's project history for future brainstorming or export.

Generate Titles

Steps:

Input Context Assembly: The system collects all enabled research inputs selected via checkboxes:
- Channel: Uses the analyzed Channel Information document (target demographic, niche, content pillars).
- Competitors: References competitor video performance data, concepts, and benchmarks.
- Trends: Injects real-time trending topics and keywords from Google Trends.
- Sources: Reads all downloaded articles and documents from the channel's Sources folder.
- Ideas: Gathers notes and brainstorming ideas from the saved Ideas Info document.
Model Routing: Packages the compiled context sources and routes them directly to the active LLM (e.g. Gemini 3.1 Pro Preview).
AI Prompt Injection: The system injects the following base system instruction and conditional formatting rules into the model configuration (you can edit, reset, or set your custom prompt as official default here):

System Instructions:

If Listicle Format is selected:

If Question Format is selected:

If News Headline Format is selected:
Format Application & Processing: The model applies the chosen format mode (Default, Listicle, Question, or News Headline format), evaluates the combined context, and generates exactly 10 high-CTR video title ideas ranked in order.
Delivery & Selection: Titles are rendered as selectable radio options in the UI. Selecting a title immediately syncs it as the designated project title across Script, Thumbnail, and Data generation workflows.

i
Generate Data

Steps:

Context Assembly: The system gathers the active script text from the editor. If populated, it will also attach the active Channel Information document and the Target Video Title into the prompt context window.
Instruction Formatting: The AI is heavily instructed to act as a "YouTube optimization expert" and is given a very strict formatting template passed via the system prompt (you can edit and save your custom prompt here):

System Instructions:
API Execution: The system passes the context and instruction to the active model. The generated text is parsed by the UI and injected into the appropriate text areas on the Data page.