Turn any link into
a short-form video.

Paste a GitHub PR, an arXiv paper, a Hacker News story, or any article. Pick a tone, a voice, and a format — solo monologue or a two-character dialogue with sticker bodies reacting per line. Get back a render-ready vertical MP4 with karaoke captions over a looped gameplay clip. Built for builders who don't feel like editing in CapCut.

See pricing

Free. No card. 2 renders to see if the output sounds like you before you commit.

Made by Tobi.

01 · Source

You paste a link, we fetch the content.

GitHub PR, arXiv, HN, articles, YouTube transcripts, PDFs, Google Docs, or X threads. All parsed server-side.

02 · Script

Claude writes a monologue or a dialogue.

claude-sonnet-4-5 streams the script in your tone. Switch the format toggle and you get a 4-8 line back-and-forth between two characters instead of a single voiceover.

03 · Video

ElevenLabs voices, ffmpeg renders the MP4.

Single voice for monologue, alternating voices + sticker character bodies in dialogue mode. B-roll images pop in on cued phrases. Karaoke captions over a 90s gameplay loop. 1080×1920, TikTok / Shorts ready.

How it works

Paste a link, ship a video. Three minutes, not three hours.

Paste a source

GitHub PR, arXiv paper, Hacker News story, or article (free). YouTube, PDF, Google Doc, or X thread (Pro). We auto-detect the type from the URL, no picker. PDF uploads work too.

Pick a tone and format

Tone: Educational, Hot Take, or Hype. Format: solo monologue (one voice, ~30-45s) or dialogue (two characters back and forth, ~50s, sticker bodies reacting per line). Same source, very different videos.

Generate, edit, render

Script streams in about 10 seconds — fully editable line-by-line in dialogue mode. Pick voices (per-speaker in dialogue), pick a caption style, and render. 1080×1920 MP4 with karaoke captions over a gameplay loop, sticker characters in the bottom corners during dialogue, b-roll images popping in on cued phrases.

Features

Everything between source and shipped.

Eight source types, one input

Free: GitHub PR diff, arXiv abstract, Hacker News story + comments, any article (Mozilla Readability). Pro: YouTube transcripts, PDFs (URL or upload), Google Docs, X threads. Auto-detected from the URL — no source-type picker.

Eight ElevenLabs voices, per speaker

Three male narrators (Liam, Drew, Adam), three female (Rachel, Aria, Charlotte), two character voices (gravelly Clyde, witchy Glinda). In dialogue mode you pick one voice for the asker and one for the expert independently, so the back-and-forth has real contrast.

Monologue or two-character dialogue

Solo voiceover for explainers, or a 4-8 line back-and-forth between an asker (naive, hooks the viewer) and an expert (drops the punchline). Per-line reaction poses drive the sticker characters that animate during their lines.

Eight sticker characters, gender-matched

Four asker bodies + four expert bodies (dev-bro, dev-girl, gamer, skater, anime-girl, founder, scientist, professor). The renderer pairs gender-to-voice automatically, or you pick the exact pair in the UI. Six reaction poses each.

Two caption styles + b-roll overlays

Clean (Inter 72, amber karaoke fill) or Bold 3D (Inter 96, chunky black-stroke TikTok feel). B-roll images pop in for ~2s on cued phrases — Unsplash photos when the prompt is concrete, Flux Schnell when it's abstract. SRT export for CapCut.

MP4 over a gameplay background

1080×1920 vertical, captions burned over a looped 90-second Subway Surfers parkour clip — the brainrot template that holds retention on TikTok. Rendered server-side with ffmpeg, drop straight into TikTok or Shorts.

Streamed monologue, per-line editable dialogue

Monologue scripts stream token-by-token so you can read and edit while the model is still typing. Dialogue scripts come back as 4-8 inline-editable cards so you can polish individual exchanges before voicing.

Pricing

Try it free. Upgrade when you're hooked.

Free

$0forever

Enough to see if the output is good enough to post.

2 video renders, lifetime
4 source types: GitHub PR, arXiv, Hacker News, articles
All three tones, all eight voices
Clean and Bold 3D caption styles
SRT caption export
Cross-device history
YouTube, PDFs, Google Docs, X are Pro

Recommended

Pro

$12/month

For when one render isn't enough.

Unlimited MP4 renders
All 8 source types: includes YouTube, PDFs, Google Docs, X threads
Looped gameplay background, vertical 1080×1920
Two caption styles (Clean amber karaoke, Bold 3D)
7-day refund window, no questions
Cancel anytime in one click

Plans are billed monthly via Stripe. Refund policy.

FAQ

Quick answers.

What can I paste in?+

Eight source types in total. Free: GitHub Pull Requests, arXiv papers, Hacker News stories, and any article URL (we use Mozilla Readability to extract the body). Pro: YouTube videos, PDFs (via URL or upload), Google Docs (shared publicly), and X threads. YouTube, PDFs, Google Docs, and X are Pro-tier sources. Everything is auto-detected, no source-type picker.

How long is the output video?+

Monologue scripts are tuned for 80-120 words → 30-45 seconds spoken. Dialogue scripts target ~130 words across 4-8 alternating lines → ~50 seconds. Both windows fit within TikTok's retention sweet spot.

What's the difference between monologue and dialogue?+

Monologue is the original single-voice explainer: one ElevenLabs voice reads a script you can edit in a textarea. Dialogue is a 4-8 line back-and-forth between two characters — an asker who hooks the viewer with the naive question and an expert who drops the punchline. You pick a separate voice for each speaker, edit each line individually, and the renderer overlays sticker character bodies in the bottom corners that swap per line with poses driven by reaction tags.

Can I edit the script before the voiceover?+

Yes. Monologue comes back in a single editable textarea. Dialogue comes back as 4-8 inline-editable line cards — tweak any individual exchange before clicking Generate voiceover. Your edits feed directly to ElevenLabs.

What if the AI gets a fact wrong?+

Edit it. The script is a draft, not the final cut. Every project we tested needed at least one factual correction, that's why the script is editable instead of one-shot.

What does the rendered video look like?+

1080×1920 vertical, 30-50 seconds. Karaoke captions in the middle of the frame over a looped 90-second Subway Surfers parkour clip. Top of the frame is reserved for b-roll image overlays that pop in for ~2 seconds on cued phrases — Unsplash photos when the prompt is concrete, AI-generated images otherwise. In dialogue mode, the bottom-left and bottom-right corners show sticker characters that swap per line with reaction poses (idle, confused, smug, excited, eye-roll, shocked). Two caption styles: clean amber karaoke (Inter 72) or bold 3D (Inter 96 chunky white with a thick black stroke + drop shadow).

Which voice should I pick?+

Eight ElevenLabs voices total. For explainer content, Liam (warm male narrator) or Rachel (warm female) read most naturally. For hot-takes try Drew (punchy male newscaster) or Charlotte (expressive female storyteller). For hype, Adam (energetic male) or Aria (bright female). Clyde (gravelly war-vet) and Glinda (witchy dramatic) are character voices for satire and bits.

Will it post to TikTok or YouTube Shorts automatically?+

No. You download the MP4 and post it yourself. Auto-posting is a roadmap item; right now you stay in control of when and where it lands.

How private is my data?+

Your runs are stored against your Google sign-in. We don't share them, don't train on them, and don't show them publicly. You can wipe your history any time from Settings.

Can I cancel?+

One click from Settings → Manage billing. We also honor a 7-day no-questions refund on your first charge. Full details on the refund policy page.

Turn any link intoa short-form video.