AI Content Team

1. Overview

The AI Content Team is a desktop platform for YouTube creators that automates the repetitive steps of content production — from trend research to video publishing — with AI agents.

The app is organised in three sequential phases:

• Research — Analyses the channel, identifies trends and organises the editorial calendar
• Production — Generates script, thumbnail, editing guide, B-roll, rough cut and avatar
• Report — Compiles and sends a weekly digest by email

Dashboard — pipeline overview, channel status and video in production

Available pages

2. Licence and Free Trial

2. Pricing

3. Getting Started

After activating the trial or the licence, the app shows a 3-step onboarding wizard:

Step 1 — Channel

• Channel name — as it appears on YouTube
• Handle — the channel @handle (e.g. @mychannel)
• Channel ID — found in the advanced settings of YouTube Studio

Step 2 — Channel Profile

Defines the editorial identity used by the AI agents to generate content aligned with the channel:

You can edit this information any time in Settings → Channel Profile.

Step 3 — API keys

External integrations require individual API keys. See the Settings section for the full list and where to get them.

Multiple channels

The app supports multiple channels. Use the channel selector in the sidebar to switch between them. Each channel has its own profile, calendar and production history.

4. Common Interface Elements

Video selector

Displayed at the top of Production pages. Selects which video from the calendar is in focus. The selected video determines which script, thumbnail and other assets are shown and edited.

Execution history

The "History" button at the top-right of every page opens a side panel with all previous agent runs — date, duration and result. You can review and restore previous outputs.

Real-time progress

When you run an agent, the output streams line by line in the result panel. You can follow processing without waiting for completion.

Confirmations and notifications

Destructive actions (delete, clear history) show a confirmation dialog. Action results (save, send, generate) show toasts — temporary notifications in the corner of the screen.

Missing-key alerts

If a page requires an API key that is not configured, a banner appears at the top with the key name and a direct link to the relevant section in Settings.

5. Research Phase

The research phase feeds the editorial calendar with real data from the channel and the market.

5.1 Channel

Shows the channel's current metrics (subscribers, total views, published videos, analysed videos) and generates a full strategic analysis based on the latest published videos.

Channel page — metrics and AI-generated strategic report

How to use

1. Click "Analyse channel" to generate a new report
2. The AI analyses the latest videos: performance, positioning, growth opportunities
3. The report is saved and can be reviewed via History

Requires: YOUTUBE_API_KEY

5.2 Trends

Identifies trending topics inside the channel's niche and classifies them by urgency (High, Medium, Low). You can approve, reject or postpone trends from the calendar.

Trends page — topics identified with urgency and action

Sections of the page

• Approved — trends that will enter the calendar
• Pending — awaiting review (Approve / Reject)
• Rejected — dismissed, can be reviewed later
• Outliers — atypical viral videos in the niche, useful as reference

Requires: YOUTUBE_API_KEY

5.3 Calendar

Visual editorial calendar where approved videos are scheduled. Shows every item with date, status and current production phase.

Calendar page — editorial grid with planned and in-progress videos

Available actions

• Create a new item manually
• Change the planned publish date
• Sync with Google Sheets (requires GOOGLE_SHEETS_ID)
• Jump directly to the production page of each video

5.4 Channel Profile

The Channel Profile is the editorial identity that every AI agent reads from to keep scripts, thumbnails, trends and emails aligned with the channel's voice. Initially populated by the 3-step onboarding (see Getting Started), it can be edited any time.

Channel Profile — editorial identity fields driving the AI agents

Where to edit

• Dedicated page: /perfil (Channel Profile).
• Or from Settings → Channel Profile.

What's in it

Same fields as the onboarding Step 2: niche angle, target audience, brand tone, content pillars, brand voice, title formulas, and what to avoid. Changes here flow into every subsequent agent run — re-running Script or Thumbnail after editing the profile will reflect the new tone.

6. Production Phase

The production phase covers everything between the planned video and the file ready for publishing. Use the video selector at the top of each page to make sure you are working on the right video.

6.1 Script

Generates the full video script organised in sections (Hook, Context, Body, CTA…). Each section includes a timecode hint, camera direction and spoken text.

Script page — script broken down by section with timecodes and directions

How to use

1. Pick the video in the selector at the top
2. Click "Generate script"
3. Wait for the generation (streamed in real time)
4. Expand and edit the sections as needed
5. Click "Copy" to copy the entire script

Teleprompter

Click "Teleprompter" to open the presentation view: large text, controllable auto-scroll, adjustable speed, font size and visible lines.

Teleprompter mode — text with auto-scroll for recording

Requires: ANTHROPIC_API_KEY

6.2 Avatar new

Generates a presenter video with a digital avatar lip-synced to the script using HeyGen. The result can be used as main footage in the Rough Cut.

Avatar page — configuration and generation of the digital-avatar video

How to use

1. Configure HEYGEN_API_KEY, HEYGEN_AVATAR_ID and HEYGEN_VOICE_ID in Settings
2. Open the Avatar page with the video selected
3. The script text is extracted automatically
4. Click "Generate Avatar"
5. Track progress — generation takes a few minutes
6. Download the video when it's done via the download button

Requires: HEYGEN_API_KEY, HEYGEN_AVATAR_ID, HEYGEN_VOICE_ID

6.3 Thumbnail

Generates thumbnail visual concepts with AI (Gemini). Presents multiple options to pick from and lets you set the official thumbnail of the video.

Thumbnail page — AI-generated concepts to choose from

How to use

1. Click "Generate thumbnails"
2. The AI generates 3+ concepts based on the script and channel profile
3. Preview each concept and pick your favourite
4. Click "Set as official" to record the choice

Requires: GEMINI_API_KEY

6.4 Editing

Generates a detailed editing guide with shot list, B-roll suggestions, captions and soundtrack, based on the script. The video editor uses this guide as a reference when assembling the video.

Editing page — full guide with shot list and assembly instructions

What's in the editing guide

Requires: ANTHROPIC_API_KEY

6.5 B-Roll

Searches and manages supporting footage (B-roll) from multiple sources: Pexels stock library, AI generation via Runway and manual uploads.

B-Roll page — clip manifest with download status and cost

B-Roll sources

• Pexels — free stock footage, keyword search
• Runway ML — AI video generation (paid per generated second)
• Manual upload — add your own clips

Available actions

• Download clips individually or as a ZIP
• See the estimated cost per source
• Delete unneeded clips

Requires: PEXELS_API_KEY and/or RUNWAY_API_KEY

6.6 Rough Cut

Automatically generates an editing timeline with timecodes mapped from the script to the main footage. Exports an XML file compatible with Premiere Pro and DaVinci Resolve.

Rough Cut page — timeline with timecodes and XML export

How to use

1. Make sure the script and B-roll are ready
2. Click "Generate Rough Cut"
3. The system maps the script timecodes to the footage
4. Review the timeline and tweak manually if needed
5. Export the XML for Premiere Pro or DaVinci Resolve

6.7 Studio new

The Studio is the in-app HyperFrames editor: an LLM-driven chat on the left and a live composition preview on the right. Use it to build the final video without leaving the app — as an alternative to exporting XML to Premiere Pro or DaVinci Resolve.

Studio — chat-driven HyperFrames editor with live preview and scene markers

Layout

• Left panel — chat with the LLM. Ask it to add a scene, change a transition, swap a B-roll, tweak captions, etc. The panel is resizable; collapse it to give the preview more room.
• Center — the HyperFrames player rendering the current composition. A toolbar above gives quick access to History, Render and Advanced Editor.
• Scenes bar — markers below the player let you jump between scenes by timecode.

How to use

1. Pick the video in the selector at the top — the Studio loads the latest composition for it (or starts a blank one)
2. Describe changes in the chat: "add a 3-second title card at the start", "swap the B-roll on scene 2 for the lighthouse clip"
3. The LLM rewrites the composition HTML; the preview updates in place
4. Use the scenes bar to navigate, the History button to revert, and Render when you're satisfied

Advanced Editor

Click "Advanced Editor" to open the full HyperFrames Studio in a separate window — Inspector, Capture and Renders panels included. It opens against a local sidecar that serves the composition you're working on, so any edit there shows up back in the chat preview.

Requires: ANTHROPIC_API_KEY

6.8 Edit Video new

Edit Video auto-trims your raw footage based on its transcript: silences, false starts, retakes and filler words are removed so you get a clean take to feed the Rough Cut. Use it before assembling the timeline.

Edit Video — automatic trimming of raw footage from the transcript

How to use

1. Click "Pick file" and select the raw footage on disk (or drag-and-drop into the page)
2. Click "Edit video" — the agent transcribes the footage and decides where to cut
3. Watch progress streamed line by line
4. Preview the trimmed result in-page and confirm the path
5. The edited file becomes the main footage input for the Rough Cut

Requires: ANTHROPIC_API_KEY (decision-making) and ELEVENLABS_API_KEY (transcription)

6.9 Publish

Prepares the video metadata and uploads directly to YouTube, with optional scheduling. A Library sub-page lists every video you've already published from the app.

Publish page — metadata and scheduled YouTube upload

How to use

1. Select the video file (rendered or external)
2. Authorise YouTube via OAuth (first time only)
3. Review title, description, tags and chapters — pre-filled from the script
4. Tick the AI disclosure boxes where applicable
5. Pick: publish now or schedule (minimum 15 minutes in the future)
6. Click "Publish" or "Schedule"

Library

Open Publish → Library (route /publicar/biblioteca) to browse every video you have uploaded through the app: title, publish date, YouTube link, scheduling status and quick actions to re-open the original metadata.

Requires: GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET, YOUTUBE_API_KEY

7. Report Phase

7.1 Email

Generates and sends a weekly HTML digest summarising production — published videos, metrics, trends and next steps.

Email page — weekly digest template and Gmail send

How to use

1. Click "Generate email" to build the weekly digest
2. Review and edit the HTML template if needed
3. Authorise Gmail via OAuth (first time only)
4. Click "Send"

Requires: GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET, EMAIL_RECIPIENT

8. Settings

Centralises all API keys, visual preferences, execution history, licence status and app updates.

Settings page — API keys and management tabs

8.1 API keys

Every external integration requires individual keys. A progress indicator shows how many keys are configured.

8.2 Appearance

• Theme: Dark / Light / System
• Font size: Small / Medium / Large / Extra Large

8.3 Language

Switch the interface language between English, Português (Brasil) and Español. The choice is saved per profile and takes effect immediately.

8.4 Channel Profile

Shortcut to the editorial profile of the active channel (also accessible via 5.4 Channel Profile). Edit niche angle, audience, brand tone, pillars, voice, title formulas and exclusions. Changes flow into every subsequent agent run.

8.5 History

Lists every agent execution. You can delete individual runs or clear the entire history. The retention limit is configurable.

8.7 Updates

Checks manually for new app versions. When an update is available, it shows the changelog and an install button. The app can also update itself automatically in the background.

9. Support new

AI Content Team includes built-in live-chat support accessible directly inside the app.

How to open support

1. Click "Contact support" at the bottom of the left sidebar
2. In the window that opens, fill in your name and email (saved automatically for next sessions)
3. Live chat opens automatically

Data sent automatically

To speed up support, the following is sent automatically:

• Installed app version
• Operating system (Windows / macOS / Linux)
• Licence status (active, trial, inactive)

10. Full Workflow

Typical weekly production cycle using AI Content Team:

Monday — Research

1. Run Channel — refresh metrics and strategic analysis
2. Run Trends — identify trending topics
3. Approve the relevant trends
4. Run Calendar — spread approved videos across dates

Tuesday to Thursday — Production

1. Pick the video of the week in the video selector
2. Run Script — generate and review the script
3. Run Avatar — generate the presenter video (optional)
4. Run Thumbnail — generate and pick the visual concept
5. Run Editing — generate the editing guide
6. Run B-Roll — search and download supporting footage
7. Record your raw footage, then run Edit Video — auto-trim silences and retakes from the recording
8. Run Rough Cut — generate the timeline and export the XML
9. Choose one path to finish the video:
   • Studio — assemble and render directly in the app (chat-driven HyperFrames editor)
   • External editor — open the XML in Premiere Pro / DaVinci and use the editing guide

Friday — Publish and Report

1. Run Publish — upload and schedule on YouTube
2. Run Email — generate and send the weekly digest

Pipeline execution (Dashboard)

The Dashboard offers chained execution modes that run several agents in sequence:

11. Quick Reference by Page

Glossary

Version History