Today / Studio command

Build one clear, reviewable content run.

Sign in for AI + saved work

Studio Signal consultant

Plan by text or start a reviewed live voice session through the protected AI Gateway route.

Checking gateway
JarvisWhat are we making? Give me the channel, audience, output format, and goal.
FahimPlan three original kid-friendly 3D sing-along shorts.
JarvisI would use one repeatable character kit, three learning themes, and a 20–30 second format. I can prepare the first brief for review.

Live voice consultant

Model: openai/gpt-realtime-2 · browser auto-stops after two minutes

Ready for review. Microphone is off.

One connected run

Every role feeds a single reviewable output.

No automation
01
Research signalTopics, source links, audience need
Manual
02
Script + scene planOriginal hook, beats, visual direction
Draft
03
Creative render jobVisual, voice, and music cost review
Gated
04
Approval + scheduleHuman review before outbound action
Off

Studio truth

Useful progress, without fake status lights.

2planned jobs
0linked channels
0approved renders

Unified production queue

Jarvis, research, creative work, and approvals all land in one place.

Creation queue

Adding a request creates a local draft only. No generation, upload, message, or payment is triggered.

Marketing operating map

The useful core of the “AI marketing team” concept—split into accountable, reviewable roles.

Research lab

Collect trusted sources, then turn the question into a brief. This interface does not search the web or open links by itself.

Source review

Research agent, honestly

Build target: a private research agent that returns source links, summaries, and a confidence note. Until an approved search provider is linked, this page is a source-review board—not “live web access.”

Research-to-brief

Use reviewed sources to make the next job precise.

Channel board

Formats and production goals are ready. Real YouTube numbers appear only after a read-only connection.

Learning Loop
Kids 3D sing-along · sample channel
Planning
0
connected YouTube views
Character kit needed0 videos
Daily Signal
AI + founder Shorts · sample channel
Research
0
connected YouTube views
Topic map needed0 videos
Story Spark
Animated micro-stories · sample channel
Draft
0
connected YouTube views
Scene recipe needed0 videos

What gets tracked

A channel is not “connected” until Fahim approves Google’s access request.

01 / OVERVIEW

Channel pulse

Videos, views, subscribers, and publishing cadence.

02 / CONTENT

Video scorecard

Hook, retention markers, format, and next experiment.

03 / STRATEGY

Gap analysis

What the channel needs next—not made-up trend data.

04 / PUBLISH

Schedule gate

Draft queue before any upload or public post.

Generate Script

Turn one brief into an original hook, script, scene list, voice direction, and rights check.

AI Gateway

Script output

Model: openai/gpt-6-luna

Waiting for a brief.


        
Content factory
Make a production brief that can travel from research to script to scenes to approval.
No render
Factory rule
One job has a visible path: research → script → scenes → assets → review. Every later connected service must return its output and cost back to that same job.
Output checklist
What a completed brief contains.
A
Hook + promiseThe first seconds and viewer value
B
Original scriptBeats, dialogue, title, and ending
C
Shot listReusable character and scene directions
D
Render planProviders, credits, consent, review owner

        
Factory queue
Move work forward only as a draft or review artifact.

      


      

        
Generate Image
Create one original visual after reviewing the model and per-image charge.
Cost review
Image output
Model: recraft/recraft-v4.1-flash
No image generated.

        
Generate Video
Render one 4-second original clip through the protected AI Gateway route after a cost review.
Paid render
Video output
Model: google/veo-3.1-fast-generate-001
No video generated.

        
Creative studio
Four lanes for the visual workflows in the reels—kept behind provider, cost, and rights checks.

        
OpenArt lane
Visual direction lab
Prompt packs, art direction, model choice, and image/video render jobs attached to one brief.
Requires Fahim’s OpenArt account approval and a visible cost check.
Provider check
Cinematic scene lab
Environment changes, motion briefs, and scene continuity—without claiming a provider is already connected.
Use only a provider with a verified supported workflow and Fahim’s credits.
Fictional only
Character lab
Create a consistent original fictional avatar or animated character kit for a channel, with clear disclosure and review.
No impersonation, real-person likeness, or sexualized automation.
Audio lane
Voice + music desk
Voice direction, original narration, and licensed/original music planning attached to each scene list.
Requires authorized voice and music services; no account is connected.

        
Creative review rules
What prevents “generate and hope” from becoming the process.
01 / RIGHTS
Original or authorized
Confirm character, voice, music, and reference rights.
02 / COST
Show cost first
Display provider, credits, and expected output before render.
03 / REVIEW
Human checks export
Review consistency, claims, audience fit, and policy.
04 / RECORD
Keep the receipt
Save source, prompt version, asset status, and next action.

      


      

        
Agent team
These are the real departments behind the “zero employees” reel. Today they are configured roles, not hidden autonomous workers.

        
Autonomy boundary
Agents can prepare work. Publishing, inbox replies, direct messages, paid generations, and account changes remain human-approved.
What “works while you sleep” really means
A future worker can process already-approved research or drafting jobs on a schedule. It must be visible in the queue, limited by cost, logged, retry-safe, and unable to publish or message anyone without an explicit final approval.

      


      

        
Approval desk
Every draft needs a clear human decision before it can become a paid render, scheduled upload, or public message.
Outbound off
Approval is a feature
Marking a job ready here does not publish it. It only records that it is prepared for Fahim’s review. Uploading and scheduling stay separately disabled until exact account permissions are approved.
Before a render
A small checklist to prevent surprise cost or rights issues.
1
Provider selectedOnly an account Fahim has approved
2
Cost visibleCredits and expected outputs are clear
3
Rights checkedVoice, music, character, and references

      


      

        
Connection control room
Every integration starts with the smallest safe permission. None are connected from this dashboard.
0 connected
QW
Private AI consultantConnect a protected server-side route only after fresh health and authorization checks.
YT
YouTube Data + AnalyticsRead-only channel data first; upload/publish needs a separate explicit approval.
WR
Web researchApproved search and citation workflow for research briefs.
OA
OpenArtCreative provider via Fahim’s account approval, with a cost review before generation.
AU
Voice + musicAuthorized narrator and music service, selected per production job.
PB
Publishing serviceDraft/scheduling only after account and final outbound approval are clear.
Why every card says “plan”
Connection buttons are intentionally not OAuth buttons. A real connection needs a server-side credential vault, provider approval, a tested callback, scope review, error handling, and a live health check.
Integration route
The flow that turns a planned tile into a trusted connection.
1. Pick scopeRead-only first, clear purpose, no broad access.Fahim approves
2. Connect safelyServer-side secrets, callback, test account, audit log.Developer work
3. Prove itHealth, sample output, cost/permission display.Evidence required

      


      

        
Jarvis Memory
Save explicit facts, preferences, and project decisions. Chat is not memorized automatically.
Private records
Memory boundary
Only facts Fahim saves here become memory records. The consultant does not silently ingest private conversations. Records are tied to the signed-in studio owner and can be edited or deleted.

      


      

        
Automations
Save repeatable text-drafting recipes. Run one on demand after reviewing the model and usage.
Manual run
What is live now
Recipes persist and can generate a reviewed text draft through AI Gateway. Daily/weekly labels are plans, not background jobs. This deployment has no connected scheduler; it will not run while the dashboard is closed. No media, publishing, or account actions run from a recipe.

      


      

        
Asset Library
Private uploads and generated media live here with their origin and model recorded.
Private storage
Review before publishing
Uploaded or generated assets are private workspace files. Saving an asset does not upload it to YouTube or grant rights to a character, voice, image, or soundtrack.

      


      

        
Content Calendar
Plan production and review dates across channels. This does not publish to YouTube.
Planning only
Scheduling boundary
These are internal production dates only. YouTube and other platform publishing remain disconnected and require a separate account connection and final approval.

      


      
    

  

  
Connection plan
Plan a connection
This dashboard has not requested access. A live integration needs Fahim’s account approval and a server-side implementation.

  
Media cost review
Generate one image?
Model: recraft/recraft-v4.1-flash
Current listed rate: $0.007 USD per image. Rates can change; check live pricing. One click generates one image and charges the Vercel team.

  
Media cost review
Generate one 4-second video?
Model: google/veo-3.1-fast-generate-001
Listed base rate: $0.10 USD/second, about $0.40 for 4 seconds. Audio or provider options may cost more. Check current price and terms before continuing. This starts a paid render; it can take several minutes.

  
Text run review
Run this recipe once?
Model: openai/gpt-6-luna. This creates a paid text draft, saves the result, and does not publish or generate media. Check current pricing.

  
Live voice cost review
Start a timed voice session?
Model: openai/gpt-realtime-2. Vercel bills by usage; the final charge depends on how much you speak and hear. This browser auto-stops the session after two minutes or when the tab is hidden. It is not a server-side spending cap. Check current model pricing.