ChannelHelm is a video-to-publishing command center. It watches your video — audio, visuals, and meaning — and drafts every asset for every platform. You review, edit, approve, and ship.
Transcription and speaker diarization with word-level timing.
Scene cuts, frame descriptions, and on-screen text (OCR).
Audio and visual aligned into one timestamped scene log.
Topics, hooks, retention: the brief every asset is drafted from.
A single upload needs a YouTube title, description, chapters, tags and thumbnail; short-clip cuts; a blog draft; and a post for every social network — each on-brand, each different. ChannelHelm does the first draft of all of it, locally, in one pass.
Not just a transcript — a four-layer read of what's said, what's shown, and what matters, fused into a timestamped scene log.
One source video becomes a canonical Publishing Package: every derivative asset, each with full provenance (which model, prompt, and inputs made it).
Local-first by design. Your media and transcripts never touch a cloud SaaS; everything runs on your own computer.
Drop a file or paste a link. For a YouTube URL, the brand is auto-detected from the channel — no manual picking. A package is created and the pipeline queues.
Background workers transcribe the audio, analyze the visuals, fuse them into a scene log, and extract the intelligence layer: topics, hooks, and retention windows.
Every asset is drafted and waiting. Read scored options, edit inline, regenerate a section, or generate one on demand — and watch the rest fill in as the pipeline completes.
Approve what's ready and dispatch: YouTube and social through your connected accounts, editorial to your publishing destination. Track each asset through to published.
One ingest produces the whole kit — scored where it counts, editable everywhere.
The per-package review is where you live. Three layouts, one keystroke apart — pick the one that fits the job.
Two-pane daily review: platform rail, video plus live pipeline plus stacked assets, and a confident approval panel.
A file tree of every asset, a focused single-asset editor with side-by-side comparison, and a provenance inspector.
An overview canvas of every platform with completion percentage. Triage what's ready; click in to focus.
Titles and tags carry 0–100 scores against character budgets, so the best pick is obvious at a glance — then editable inline.
Don't like a section? Regenerate just that one. Empty section? Generate it on demand straight from the transcript.
Run many channels. Paste a link and ChannelHelm matches the right brand from the channel, falling back to the website domain.
Choose the AI you trust for each kind of work — cloud or fully on-device — set per task or as a default, with your own image provider for AI thumbnails too.
A four-layer progress indicator shows exactly what's done and what's still generating. Partially-ready is a first-class state.
Every generated asset records the model, provider, prompt version, and inputs that produced it. Nothing is a black box.
Re-mine an old video through the pipeline with your current prompts — defaults to the cheap transcription-only profile, so back catalogues yield fresh kits without re-recording.
A plain-English walkthrough of how ChannelHelm turns one upload into a whole channel's worth of on-brand content — the learning loop, the planning desk, and the business layer.
Run it on your own computer, keep your media local, and turn one upload into a complete, on-brand publishing kit.