Captions?

NOT REALLYA prototype of the interface is possible, but the durable product depends on difficult systems beyond the UI. Build a safer adjacent workflow instead.

BUILD FIRST

A non-sensitive or read-only version of the core experience
A manual review step wherever correctness or trust matters
A narrow companion tool that removes a real piece of work

LEAVE FOR LATER

Regulated decisions, payments, or irreversible actions
Claims of correctness, compliance, or uptime that the MVP cannot support
Sensitive integrations until security, support, and auditability are ready

RESEARCH NOTES

What this tool actually does.

Checked 2026-09-06

Captions (captions.ai) is an AI-powered video creation and editing application available on iOS, web, and desktop. Its positioning is 'AI that edits like a professional editor,' covering automatic editing, trimming, captioning, and generating footage from prompts. For voice AI specifically, Captions offers AI voiceover generation from text using multiple external models (ElevenLabs, Cartesia, PlayHT, OpenAI), AI voice cloning via a one-minute script read, and AI dubbing into other languages while optionally using the user's cloned voice. Voiceover tracks can be edited, volume-adjusted, split, and used to drive lip-synced avatar videos in its 'Prompt to Video' flow. Other AI tools include auto-zoom, censor, denoise, sound effects, music, background cutout, and simulated eye contact. Official docs note platform-specific availability: AI Voiceover, Denoise, Sound Effects, Music, Cutout, and

SOURCES

RESEARCHED BUILD PROMPT

Give this to your coding agent.

Build a focused replacement for Captions' voice-AI workflow, not its full video editor. Core user journey: a creator imports a video clip, writes or pastes a script, selects an AI voice or their own cloned voice, generates a voiceover track, inserts it on the timeline, then optionally dubs the video into a target language using that same voice, with lip-sync roughly aligned. User-facing requirements: (1) text-to-voiceover generation with multiple selectable voice models and preview; (2) voice cloning from a 60-second sample; (3) multilingual dubbing that preserves the chosen voice's identity; (4) simple audio-track editing (trim, volume, split) over the original clip; (5) automatic caption generation for the final video. Technical requirements: use modern TypeScript/React frontend and Node or Python backend; integrate at least one commercial TTS/voice-clone API (e.g., ElevenLabs, Cartesia, PlayHT) and one dubbing/voice-conversion API; store user clips, scripts, voiceover audio, and cloned-voice profiles in a database with per-user auth; expose an API for generating voiceovers and dubs asynchronously with status polling; implement basic lip-sync by time-stretching generated audio to clip pacing rather than true model-driven lip sync. Sensible approaches: Next.js or Vite frontend, Express or FastAPI backend, PostgreSQL for user data, S3-compatible storage for media, and WebSocket/SSE for generation progress. Explicitly do not copy the Captions brand, logo, marketing copy, or visual design. Out of scope: the full video editor timeline (no multi-track, filters, overlays), AI zoom/eye contact/censor, prompt-to-video avatar generation, and mobile native apps. Write a README covering setup, required API keys, how the async generation and polling works, and important limitations (voice clone quality depends on sample clarity, dubbing cannot guarantee perfect lip-sync, latency and cost of external APIs).

Visit the official product page