Make a video with one request

What it makes: a ~30s violet-themed video about the company's latest earnings call — presenter over a stock chart, animated fact cards, and a quote card.

What we do with your ticker: we find the company's newest earnings call transcript, cut it to management's prepared remarks (the analysts' Q&A and the boilerplate are dropped), and our model writes the narration, the title and three on-screen facts from those remarks only. The quote is never written: a real sentence is lifted verbatim from the transcript with the executive's name on it. Leave the ticker out and it covers the newest earnings call on a US listing.

Use it for: earnings-day coverage of any company on your watchlist — {"template": "earnings-call", "ticker": "NKE"} the morning after Nike reports.

What it makes: a ~30s forest-green video about the company's most recent notable SEC filing — headline card, presenter over a chart, animated points, quote card.

What we do with your ticker: we pull the company's recent filings from an events feed that grades each one (a guidance change or a CFO departure is a catalyst; a routine proxy is not) and take the newest one worth covering. The narration, title and points are written from that filing's analysis; the quote is copied verbatim from the press release the company attached to the filing on the SEC's own site, with the executive's name. If every recent filing is routine we refuse with a clear error rather than narrate nothing. Leave the ticker out and it covers the newest confident catalyst in the whole market.

Use it for: reacting to news — an 8-K lands, you submit the ticker, the explainer exists 20 minutes later.

What it makes: a ~10 second alert: what just happened, and why it matters. The fastest format (~5–8 min).

What we do with your ticker: same events feed as recent-event — the ticker's most recent graded catalyst — but compressed to a two-sentence spoken alert instead of a full explainer. Refused at submit if the ticker has no current catalyst.

Use it for: high-frequency posting — a stream of quick hits across a portfolio.

What it makes: the same ~10 second alert, split in half — the top half is an 8-second cinematic establishing shot generated for this event, looping over the alert; the bottom half is the presenter on a soft blue gradient, with a black VARIANT / variantmobile.com channel bar riding the seam between the halves.

Where the shot comes from: the pass that writes the alert also writes a purely-visual b-roll direction for the company and its industry, and a video model renders it in a fixed house style — glossy 3D-animated, cinematic, with no text, logos, brand names or people talking. It generates while the avatar renders, so the clip costs little extra time.

Use it for: the alert with production value — when the quick hit should look like a produced segment, not a template.

What it makes: a ~15s clip of the presenter over the company's live one-year stock chart, with captions.

What we do with your ticker: the chart is drawn from live market data for that symbol, and the narration is a short take on the company written by our model. Pass tweet_url and the post is pinned above the chart as a card — and the narration is written as a reaction to that post instead of a generic take.

No ticker? An X post alone carries the clip: the card takes the chart's place, the identity bar drops its ticker chip, and the presenter reacts to the post. One of the two is required — with neither there is nothing to put on the page.

Use it for: a quick market-commentary clip when there is no specific filing to point at — or a straight reaction to someone's post.

What it makes: the presenter over the ticker's live chart on a slate field — a navy gradient page with a steel-blue glow behind the presenter, the chart card in a matching dark theme, and the identity bar in white. Captions, closing splash, 9:16.

What we do with your ticker: only the chart — it is drawn from live market data for that symbol (span withrange: 1y, 6m or 2y). Every spoken word is yours: send notes (the model writes the script and title from them) or script (spoken verbatim). No feeds are touched.

Use it for: your own commentary on a name — the take is yours, the chart keeps it anchored to the tape.

What it makes: the fight card — a bull-versus-bear debate as generated arcade-style artwork per round, each side's claims spoken over it in its own voice, and a declared winner at the end.

Where the material comes from: the Variant insights feed — the two-and-two bull and bear case for the ticker — so variant.api_key with insights access is required. Each side's argument is written by our model from its side's points. Leave ticker out and the newest insight is covered.

What you supply: two face images (they become the fight artwork), optionally a voice per side, and the winner.

The fight card carries no identity bar by construction — no handle, channel or tier appears on it, only the ticker and company being debated. is_general is accepted and changes nothing here; it exists for the presenter formats.

curl -X POST https://variantvideo.com/api/v1/videos \
  -H "Authorization: Bearer vv_YOUR_KEY" -H "Content-Type: application/json" \
  -d '{
    "template": "showdown-no-lipsync",
    "ticker": "HOG",
    "faces": {
      "bull": "https://.../bull.jpg",
      "bear": "https://.../bear.jpg"
    },
    "voices": { "bull": "am_michael", "bear": "af_heart" },
    "winner": "bull",
    "variant": { "api_key": "ftk_YOUR_VARIANT_KEY", "publish": false }
  }'

Example output: a fight card. The lip-synced three-hander (showdown, example) is studio-only — three avatar generations per render is the farm's heaviest job, so it stays off the programmatic door.

What it makes: the simplest thing that works — your presenter over a flat background, speaking your material, with captions. No ticker needed.

What you supply: notes (the model writes the script and title from them) or script (spoken verbatim), plus a face — photo_url or your Variant avatar via the key.

What it makes: the split-screen frame — an 8-second cinematic establishing shot generated from your material loops in the top half, your presenter speaks in the bottom half, and the channel band rides the seam. No branded opener.

What you supply: notes orscript plus a face, like talking-head — no feeds are touched, and a ticker is optional (it becomes the chip). Pick the page with background: blue (default), green, red, black, purple or orange — every choice is the same pale-into-deep gradient at a different hue. With a verbatim script, one small model pass still writes the shot's visual direction from it.

Use it for: produced-looking commentary on anything — your words carry the clip, the generated footage sets the scene.

What it makes: your presenter announces a move on location — the face you supply is re-staged into the chosen scene before it is animated, so the whole frame is the shot. Full-bleed 9:16 with captions and the closing splash. No ticker needed. Built for short clips ("just bought some ETH, come ride the wave"): the render time tracks the length of what is said.

The scenes (pass one as scene; blank picks the beach): beach — a lounge chair on a sunny tropical beach, laptop on their lap, waves rolling in behind them. limo — the back of a sleek black limousine at night, city lights streaking past the window. penthouse — a rooftop terrace over a glittering skyline at dusk.

What you supply (one of three, in priority order): script is spoken verbatim; trade is a facts object the farm's own model turns into a punchy first-person trade-alert line (direction-guarded, ~24 words); notes is free text the model narrates. Plus a face — photo_url or your Variant avatar via the key — and optionally the scene.

The trade object: {"platform", "asset_in", "asset_out", "side", "chain", "market"} — asset_in is what was bought, asset_out what was sold, side is buy/sell/long/short. A swap ("I swapped USDC for Ethereum") needs both assets; a plain position needs asset_in or market. When you send trade facts, the $TICKER chip and the TRADED ON badge fill themselves from them (explicit ticker/platform still win), and the video's title is derived from the facts ("USDC → ETH", "Long MON").

The top bar: the same identity strip as every format — handle, channel, tier and the $TICKER chip (pass ticker to show one, e.g. ETH) — plus this format's own TRADED ON badge: pass platform (say, "Metamask") and it renders beside the chip. Publishing with as_channel keeps the clip general and drops the identity bar, same as everywhere else.

Good to know: the face in the finished video is the face you sent, preserved through the scene change. A clear, well-lit source photo makes a better scene.