Own build, same tools and same structure. No copied code, images, logos, or brand assets from his tool. Tool names match so the workflow feels familiar.
Dashboard
Welcome back. This is your private studio — no approval queue, no credits, no per-video gate. It runs on your own keys and your own GPU time.
Member
Private studio owner Member since: this studio (no account system)
Access
Full access — all 15 tools No plan, no upgrade screen
Credits
No credits Costs are your own API keys and Modal GPU time
Usage
Loading…
Devices: not tracked in this studio
History — saved projects
Every job you ran, newest first. Click a row to open its tool panel. Projects are never auto-deleted here.
Cloud import
Drive, MEGA, Dropbox, and OneDrive accounts are not connected to this studio, so there is no one-click cloud sign-in. What works today: upload a file from your device, or paste a direct link (up to 500 MB) in any tool that accepts files. The studio downloads the link itself.
Niche Hunter
Find growing faceless channels. RPM is private YouTube data — the report never guesses it. Channel creation date, subscribers, and country need a YouTube Data API key; without one they show as unknown.
Faceless cannot be detected from public data — this flag is your own label on the report, not a measurement.
Saved
Title Hunter
Our 5 template methods (not his prompt — we do not have it). Competitor mode learns formats from real titles; Master mode uses the templates; Live mode adds fresh news items with date and source.
Saved setup
Script Hunter
Script writing here stores your scripts and counts words today. The AI style-clone and viral-rewrite passes need an LLM and are a later upgrade — the panel saves your inputs and says so plainly instead of faking a rewrite.
Jobs
VO Hunter
Voice model comes from Settings (ElevenLabs, MiniMax, Fish Audio, or free Piper). The browsable list below shows staged free voices with previews, plus your configured provider voice. Only staged voices render without spending — other languages fail plainly naming the missing voice file.
Choose a voice
StampHunter
Exact timestamps per line, measured with faster-whisper — never estimated. Lines break between words, never mid-word, because stamps come from speech segments.
Custom time-range rules — one row per part. Default: 0–1 min every 3s, then every 6s.
Dub Hunter
Dub a whole video into another language at the same length. New voices use one consistent Piper voice for the whole video — Original voices (each speaker keeps their own voice) need diarization + cloning and are not built yet. Translation needs a LibreTranslate URL or DeepL key in Settings, or paste translated lines below.
Checked mixes the original audio quietly under the dub — music survives, but original speech bleeds quietly too. Unchecked replaces the audio entirely.
Videos can be up to 3 GB in his tool. Here, direct-link imports cap at 500 MB — for bigger files, upload the file or split it first.
Prompt Mapper
One image prompt per timestamp. Prompts today are template-based (skill style + script line) — an LLM pass is a later upgrade. His tool has no skill named stickman; our Stickman skill is ours, not his.
Make a skill
Up to 5 competitor frames, up to 10 own skills. The skill saves your style notes + frames; frame analysis is manual for now — no automatic style extraction is claimed.
Image Batch
Every picture in one run, via our Modal FLUX.1-schnell backend (licence-clean). Character and style consistency here means the same anchor words in every prompt — a prompt-level trick, not a trained character model.
External option (not primary): a Gemini bulk-generation link using your own Google account. This studio never sends your prompts to Gemini itself — that path is yours, in your browser.
Video Batch
Every shot in one run. Image + zoom shots are real (AI picture + zoom + voice line) — NOT AI video generation. True AI shots need a video model endpoint in Settings.
External option (not primary): Google Flow using your own Google/Flow credits. This studio never spends your Flow credits itself.
Clip Hunter
Stock clips from Pexels and Pixabay (free keys in Settings), YouTube reference links for research, or per-scene research rows. Stock counts page through the provider's API — square has no API orientation value, so it searches unfiltered.
Frame Hunter
Assemble the finished video from pictures and clips + voiceover. A file with only a start time ends when the next file starts (auto-end). Videos are preferred over same-name images. This panel renders independently of the Phase 0 chain — or reuse a Phase 0 job's images and voice below.
Media
Cloud: Drive, MEGA, Dropbox, OneDrive are not connected — paste a direct link or upload from your device.
Voiceover
Settings
Sizes are estimates from his demo (a 21-minute 2K video ran ~1.8 GB at 30fps). Your actual file size shows after the render.
Captions
Karaoke in this panel falls back to Bold — the file timeline has no word timings. Karaoke works in the Phase 0 chain and in Timeline, where word timings are saved.
Caption preview — drag me
Timeline
Each file shows start, end, duration, and zoom %. Files with only a start time end at the next file (auto-end note).
Hunter Timeline
A 4-track editor: Captions, Overlay, Main, Audio. Shortcuts work when this panel is open: Space play/pause preview, S split, Q cut left, W cut right, D duplicate, Delete remove, Ctrl+Z undo, Ctrl+Y redo. Magnetic snapping pulls cut ends to neighbours.
Cloud: not connected — upload from your device or paste a direct link in the voiceover field's job flow.
Stock visuals need a Pexels key in Settings. AI visuals use the Modal stills backend.
Generates with the export below; Download .srt appears in the result.
Preview
Tracks
Captions—
Overlay—
Main—
Audio—
Cuts
One row per keep-range, seconds. Empty = whole voiceover. Magnetic is on: ends snap to the next start when within 0.5s.
TT Hunter
Long video to viral shorts, ranked by a template viral score (not a learned ranker). Blurred intro cards, sound effects, background music, and silence cutting are NOT built — the result lists anything you asked for that was not applied.
Genre is stored on the result — the template score does not vary by genre yet.
Auto edit
HunterChat
Ask in plain words, any language. Rule-based v1: it picks the tool, runs light tools (titles, jobs, settings checks) right here, and guides you to the right panel for heavy tools. An LLM agent is a later upgrade — this chat never pretends to be one.
Cloud import: not connected — upload from your device, or paste a direct link into the matching tool panel.
Try:
FB Hunter
His FB Hunter is still SOON in his tool — there is nothing of his to copy yet. What you see here is OUR basic Facebook Reels version on the TT Hunter engine: same clips, Facebook format. It is not presented as his FB Hunter.
Make Facebook Reels in the TT Hunter panel — pick format "Facebook Reels" and run. The result carries the same plain label.
Voice + image settings
Keys are stored on the studio storage volume only, never in this repo and never shown back. Empty means not set yet.
What is real, what is not yet
Voice: API voices when keys are set; free Piper voice otherwise (real speech, clearly labelled). Speed and audio name are real; word-check runs at the Stamp stage.
Stamps: measured with faster-whisper word timings (tiny = Fast, base = Best), not estimated. Gap rules include Sentence, 3s, 5s, 8s, and custom time ranges.
Images: Modal FLUX.1-schnell (licence-clean). Never FLUX.1-dev, never XTTS v2. Aspect sizes are real; consistency toggles are prompt-level only.
Frame: real FFmpeg render with size (1080p/2K/4K/Shorts), FPS, Fit/Fill, zoom range or none, caption styles + colours + draggable position, quality presets, and export length. Karaoke in the file-timeline panel falls back to Bold (no word timings there).
Niche: real public channel data (API key, or public page without one). RPM is never guessed. Creation date, subscribers, and country need the API key. Faceless is your label, not a measurement.
Titles: real competitor formats + our 5 template methods. News items carry date + source. LLM titles are a later upgrade.
Clips: real scene research; stock search is real only with a free Pexels/Pixabay key, else it fails plainly. Stock counts page through the API.
Dub: real Whisper + translation + Piper + length stretch + mux. One voice for the whole video — Original voices need cloning, not built. Background-music mixing is real (quiet original under the dub).
TT Shorts: real Whisper windows + template viral score + reframe + captions. Facebook Reels is our basic version, not his (unreleased). Blurred cards, sound effects, background music, and silence cutting are not built and are listed as not applied.
Timeline: real captions, cuts, zoom, speed, 4 tracks, shortcuts, magnetic snap, extract audio, text overlays, and Auto Video (AI/Stock/both). Karaoke uses saved word timings.
Video Batch: image-zoom shots are real. AI video shots need a model endpoint — not deployed yet. Flow is an external option only, never primary.
HunterChat: rule-based v1 — picks tools, runs light ones, guides heavy ones, names missing keys exactly. LLM agent is a later upgrade.
All 15 tutorial functions are built. Facebook uses our basic version (his FB Hunter is still unreleased).