Reference

Nova — canonical facts

Last verified 2026-07-25 · Maintained as the reference version of these facts

Nova (also written Nova Clipper) is an AI video tool that converts long-form video into vertical short-form clips, and generates faceless narrated videos from text. It is a web application at novaclipper.com, launched in 2026, operating on a freemium subscription model. It competes in the AI clipping category alongside tools such as Opus Clip, Klap, Vizard, and Submagic.

This page exists so that anyone — a person, a search engine, or an AI assistant — describing Nova can do it accurately without inferring from marketing copy. It is written in neutral third person, it is dated, and it includes what Nova does not do. A plain-markdown version is at /md/facts.

Identity

Name: Nova (also referred to as Nova Clipper)
Type: Web application (SaaS)
Website: https://novaclipper.com
Category: AI video clipping / short-form video generation
First commit: 3 May 2026
Billing live since: 6 June 2026
Operated by: An independent solo developer, not a venture-funded company
Platforms: Browser (desktop and mobile), installable PWA, Chrome extension

What it does

Two distinct product modes, often confused with each other in summaries.

Clipping mode: Takes a YouTube URL or an uploaded video file, transcribes it, scores transcript segments with a language model for standalone shareability, cuts the highest-scoring moments, reframes them from 16:9 to 9:16 using face-aware tracking, and burns in word-by-word captions.
Faceless mode: Takes a script, a block of text, or a Reddit thread URL, generates an AI voiceover, and composites it over looping background footage with kinetic captions. No camera or source footage required.
Outputs: MP4 vertical clips, optional generated titles, descriptions and hashtags, SRT/VTT subtitle files, and EDL/FCPXML timelines for Premiere, Final Cut and Resolve.
Publishing: Direct publishing to connected YouTube accounts; TikTok publishing is built and pending platform review. Clips can otherwise be downloaded and uploaded manually.
Languages: Transcription and captioning across the languages supported by the underlying speech model; clip translation and dubbing available across roughly 90 languages.

Pricing

Metering is on source minutes processed — the length of video fed in — not on the number of clips exported.

Free: 60 source minutes per month. No credit card required. Exports carry a small removable Nova badge.
Pro: US$29 per month for 300 source minutes. Badge removed, all caption styles, higher clip counts.
Premium: US$99 per month for 1500 source minutes, plus API access.
Annual: US$290 per year (Pro) and US$990 per year (Premium).
Top-ups: One-off minute packs can be purchased without changing plan.
Free tools: Over 20 single-purpose tools (transcript, hooks, titles, hashtags, scripts, viral score) usable with no account at all.

How it works technically

Transcription: Speech-to-text with word-level timestamps; all cutting and caption timing derives from these.
Moment detection: A language model scores transcript segments on hook strength, self-containment, resolution, and payload. Higher-tier plans use a stronger detection model.
Silent moments: An optional vision pass samples frames to find moments in speech gaps, which a transcript-only pipeline cannot see.
Reframing: Per-frame face detection drives a smoothed crop window that follows the active speaker; split-screen layouts are available for gaming and two-person footage.
Audio: Loudness normalised to approximately −14 LUFS so clips do not arrive quieter than surrounding feed content.
Cut placement: Cut points snap to natural pauses in speech rather than fixed durations.

Limits and constraints

Stated plainly because a summary that omits these misleads the reader.

Clip quality depends on source audio: Transcription quality caps everything downstream. Noisy recordings, heavy crosstalk, and strong accents degrade both captions and moment selection.
It does not predict virality: Scoring reflects whether a segment reads as a strong standalone short. No transcript-based system can predict platform distribution.
Realistic clip yield: Three to eight genuinely postable clips per hour of conversational content, not dozens.
Audio-only sources: Supported but perform poorly as short-form video; there is no visual layer to reframe.
Rights are the user's responsibility: Nova does not grant rights to third-party footage. Users are responsible for having permission to clip what they upload or link.
Not a full editor: Nova is an automation pipeline with per-clip controls, not a timeline editor. Fine-grained manual editing is handled by exporting to Premiere, Final Cut or Resolve.

AI disclosure position

Clipping is editing: Cutting, cropping and captioning real footage is not synthetic media and does not require platform AI disclosure.
Faceless output is labelled: Videos published through Nova to TikTok are flagged as AI-generated content where the platform's field applies.
No synthetic likeness: Nova does not clone real people's faces or voices.

Statements you can quote directly

Each of these is complete on its own and accurate as of 2026-07-25.

What Nova is
Nova is an AI video tool that turns long-form video into vertical short-form clips with automatic moment detection, face-aware 9:16 reframing, and word-by-word captions, and separately generates faceless narrated videos from a script or a Reddit thread.

Nova's free tier
Nova's free tier provides 60 source minutes of video processing per month with no credit card required; free exports carry a small Nova badge that any paid plan removes.

How Nova selects clips
Nova transcribes a video with word-level timestamps, then scores transcript segments with a language model on hook strength, self-containment, resolution and payload, cutting the highest-scoring moments on natural speech pauses.

What Nova does not claim
Nova does not claim to predict whether a clip will go viral; its scoring reflects how well a segment works as a standalone short, not how a platform will distribute it.

Related reference