Enter your own API keys to use the generator. Keys are stored on the server and never shared.
Estimated using published/list rates. Cross-check against provider dashboards for exact billing.
Nothing is deleted automatically — past work stays until you remove it. When the disk gets full, delete what you no longer need here.
A topic becomes narration: question-first hook, claim → named source → counter-case → honest verdict. Plain spoken text, ready for the voiceover and the video pipeline.
A profile is that channel's standing instructions — persona, format rules, structure, guardrails. Paste them from your Claude project once and reuse them here. The profile wins wherever it disagrees with the house style.
Drop in YouTube Studio screenshots — audience age and gender, geography, retention, traffic sources, comments — and any notes. Claude reads the real numbers and writes the profile around who is actually watching. With a profile selected it revises that one instead of starting over.
This profile researches the web while writing — prices, recalls, ratings and owner reports. Expect a slower run and a list of sources with the result.
Every script is fact-checked after it is written: claims are searched, wrong figures corrected, unsupported ones removed. Findings appear below the script, never inside it. Numbers are written as words so the voice reads them correctly.
Any text to one audio file. Long text is split and rejoined automatically.
Cloned through ai33, falling back to Minimax if ai33's cloning is down. Saved to your voices when it finishes.
ai33's whole catalogue — seventeen thousand ElevenLabs voices alone. Previews are ai33's own and cost nothing. Saved voices appear under My voices.
The next video to ship — topic, titles and thumbnail concept — read from a channel's own numbers, on an 80/20 proven-vs-experiment split.
Pulls top videos, recent uploads and breakout outliers from NexLev (3 quota). Needs a NexLev key in Settings.
Click one of yours to load it above. Tick competitors to read their breakout videos too (1 quota each, up to 4).
Will Google treat this IP as a normal user, or as automation? Checks each IP or proxy against a dozen sources, asks Google and YouTube directly through the proxy, and looks at the surrounding /24 for proxy fleets. Vendor fraud scores are capped, and the report says where the sources disagree and which side to believe.
0 targets · up to 100 per check, duplicates and #comments are skipped. A bare IP gets a reputation lookup; a proxy is also tested live. Passwords are never saved.
Ping and speed are measured from the machine running this app. On the hosted version that is the server, not your computer. Optional extra sources (AbuseIPDB, IPQualityScore, ipinfo, GreyNoise, proxycheck.io) switch on when their keys are in Settings.
Who can use what, and what they spend on your API keys. A member past their monthly limit stops getting your keys; their own keys (Settings) keep working. Changes apply within about a minute. Admin status itself is changed only in the Clerk dashboard.
Clerk emails them a sign-up link. The access you set here is applied to their account the moment they sign up.
Everything you have generated, newest first. Only you can see your own files.
Tip: download from YouTube with yt-dlp on your computer, then upload here
Minimax: cheaper, broader languages, native cloning (needs MINIMAX_API_KEY). Ai33.pro: aggregator of ElevenLabs/Minimax/Edge/Kokoro with one key (needs AI33_API_KEY).
If unchecked, the voice is deleted from the TTS provider after the translation finishes (one-time use; Minimax doesn't support API deletion so the voice stays on your account).
The speaker's voice will be cloned from the video and saved to My Voices
Listen to the voiceover. If it sounds good, click Continue to render the video(s).
Upload a custom thumbnail to translate its text to your selected languages.
Each edit applies to the previous result. Click any thumbnail in the history below to revert.
Describe your video or paste a script to generate 3 thumbnail options.
When both are set: format from 1st, colors/typography/treatment from 2nd.
Tip: Use a video from the channel you're designing for as the Style reference.
Searches real YouTube thumbnails by look — describe one, or paste a YouTube video URL to find thumbnails like its own. Needs a NexLev API key in Settings. Pick a result to use it as the format or style reference.
Upload, paste from clipboard, or use a YouTube URL to translate text into multiple languages using AI.
Upload, paste, or use a YouTube URL — then describe changes using natural language.
Provide a second image (e.g. a face, a logo, a specific object) and reference it in your edit prompt — like "swap the face in the main image with the face from the reference" or "add the logo from the reference to the top-right corner." The reference is auto-cleared after each edit so subsequent edits operate on the latest result only.
After the first edit, subsequent edits are applied to the previous result. Click any thumbnail in the history to revert.