Captions Teardown — Gaurav Misra's $60M Pivot from Captions to AI Avatars

TL;DR

Captions started as the simplest possible product — auto-generated subtitles for vertical videos on iOS. Four years later it has raised roughly $60M from Sequoia, a16z, and Index, ships an AI Studio that generates fully synthetic talking-head videos from a script, and clears an estimated $2.5M MRR (~$30M ARR). The interesting story is not the revenue. It is that Gaurav Misra, an ex-Snap engineering lead, took the riskiest unforced pivot in his category: he moved from a working consumer captions tool with millions of installs to a generative AI avatar product, and did it before the market was forcing him to.

If you are trying to clone this, the literal product is already taken. The wedge that is not taken is vertical AI video for one specific creator niche — fitness coaches, therapists, real estate agents, financial advisors, dentists — with avatar libraries and script templates tuned to that profession's content patterns.

Quick Facts

Product	Captions (captions.ai)
Founded	2021
Founder	Gaurav Misra (CEO), Dwight Churchill (CTO)
HQ	New York, NY
Total raised	~$60M (Sequoia lead Series C, a16z, Index, Kleiner Perkins)
Estimated MRR	~$2.5M ($30M ARR)
Pricing	Free / Pro $9.99/mo / Pro Annual $69.99/yr / Max plan for AI Studio
Platform	iOS-first, Mac app, web companion
Wedge	Mobile-native AI video creation for solo creators

The Product

Captions is two products stitched together that share a single iOS app and a single subscription.

Product one is the original Captions: import a vertical video, get word-level animated subtitles styled to match TikTok, Reels, and Shorts conventions. There is a B-roll feature that auto-suggests stock footage, a teleprompter mode that runs on your phone screen while you record, and a noise-removal step that runs locally on device. This is the thing that got the first five million installs.

Product two is AI Studio, launched in late 2023. You type a script. You pick an "AI Creator" — a synthetic talking-head avatar trained from real captured performances. The system generates a video of that avatar reading your script with lip-sync, gesture, eye-contact, and B-roll cuts automatically inserted. There is also an "AI Edit" mode that takes raw footage of a real person and produces a polished short — captions, cuts, zooms, B-roll — in one tap.

What makes the product feel different from HeyGen or Synthesia is not the model quality. The model quality is roughly comparable. It is that the entire pipeline runs as a native iOS experience. HeyGen and Synthesia are web tools that assume you are at a desk. Captions assumed you were a creator on a couch with an iPhone.

A few specific design choices worth noticing:

• One subscription, everything unlocked. There is no per-credit pricing for AI generations on the consumer tier.
• Avatars as named characters, not anonymous models. "Karen," "Brian," "Aisha" — each has a backstory, a voice, a style.
• Vertical-first composition. Captions assumes 9:16.

Gaurav Misra Story

Gaurav Misra spent six years at Snap before founding Captions. He was a senior engineering manager on the camera and creative tools team — the part of Snap that built filters, lenses, and the recording stack.

The first transferable lesson: creator tooling on mobile is a different category from creator tooling on desktop. The second: the next wave of camera-native tools would be generative, not capture-based. The original Captions product was a Trojan horse. Build a useful captions tool, accumulate a creator user base, train models on millions of hours of vertical video, and when the generative video stack was ready, you would be the only consumer app already in the camera roll.

Misra hired video ML researchers in 2022 when the company was still ostensibly a captions tool, which is the kind of forward hire that does not happen by accident.

The Pivot Decision

In mid-2023, Captions had a working, growing, revenue-positive consumer iOS app with several million installs. Nothing was on fire.

A safe CEO ships incremental features against Submagic for two more years. Misra instead announced AI Studio in late 2023 and rebuilt the brand around AI Creators within a year.

The strategic insight: captions was a feature, but AI video creation is a category. Features get absorbed — TikTok now ships native captions and so does CapCut. Categories get companies. Misra saw the feature collapse coming and jumped the gap before it caught him.

The other underrated decision was keeping the brand name. Most founders would have rebranded. He did not, partly because the SEO and App Store search position on "captions" is worth millions, and partly because keeping the brand creates a soft on-ramp.

Business Model

Plan	Price	What you get
Free	$0	Watermarked captions, limited AI generations
Pro	$9.99/mo or $69.99/yr	Unlimited captions, B-roll, teleprompter, Pro AI generations
Max	~$24.99/mo	Full AI Studio, longer videos, premium avatars

A few things are notable:

• iOS App Store rev share. Apple takes 30% on year-one subscriptions, 15% after. Captions has aggressively pushed annual.
• Compute as the dominant cost. A generated AI Studio video is reportedly multiple dollars of inference cost.
• Customer acquisition is mostly organic. App Store search + creator demos on TikTok/X drive the bulk of installs.

The AI Studio Max tier is where margins are actually made and lost.

Captions vs Submagic vs Synthesia vs HeyGen vs Opus Clip

	Captions	Submagic	Synthesia	HeyGen	Opus Clip
Primary platform	iOS native	Web	Web	Web	Web
Core wedge	Mobile AI video for solo creators	Web captions	Enterprise AI video	Prosumer AI video	Long-to-short
AI avatars	Yes (AI Studio)	No	Yes	Yes	No
Vertical-first	Yes	Yes	No	No	Yes (output)
Pricing entry	$9.99/mo	$16/mo	$22/mo	$29/mo	$19/mo
Estimated ARR	~$30M	~$15M	~$70M+	~$50M+	~$25M
Funding	~$60M	Bootstrapped	~$160M	~$75M	~$30M

The crucial read: the only direct head-to-head competitor in the mobile-native AI creator slot is nobody yet. Submagic stayed in captions. Synthesia is moored to enterprise web. HeyGen has been adding mobile but their iOS app reads like a port, not a native product.

Distribution

Captions runs three distribution loops:

1. App Store search and category placement. "Captions" the keyword is a top-tier App Store search term.

2. Creator demo content on TikTok and X. Captions seeds early access to specific power creators, lets those creators produce content using AI Studio.

3. Sequoia/a16z portfolio leverage and PR.

What they do not do that competitors do: heavy paid Meta/Google ads, influencer flat-fee deals, content marketing/SEO at scale, affiliate programs.

For someone trying to clone this in a niche: you cannot win the App Store search position on "captions." But you can absolutely win the App Store search position on "fitness video AI" or "therapy short video" or "real estate listing video."

Why Now

Video generation model quality crossed the acceptability threshold for solo creators. Mobile compute and on-device ML reached the point where the round-trip experience is acceptable. The creator economy has matured into "creator as small business" rather than "creator as performer."

The window on Captions' specific positioning is closing. Within 12-18 months, HeyGen will likely ship a competent native iOS app, TikTok will integrate generative avatars directly, and OpenAI's video stack will be exposed enough that someone builds the consumer wrapper.