UnmarkAI pricing, free plan and limits

Last updated: 2026-10-27

UnmarkAI is a browser-based AI video cleanup and localization platform that uses AI inpainting to erase hardcoded captions, on-screen text and scene objects from video you own or have permission to change, rebuilding the background. It then translates and dubs the cleaned video into 69 languages, with a REST API and a hosted MCP server for automation.

About UnmarkAI

UnmarkAI is a browser-based video cleanup and localization platform run by a small independent studio based in Hangzhou, China. It targets one recurring problem for creators and localization teams: footage they own or license that carries hardcoded captions, old campaign copy, timestamps, or on-screen objects that need to come out cleanly before the clip can be re-subtitled, dubbed, or republished. The vendor describes itself as rights-aware: users confirm ownership or permission before upload, and its policies rule out stripping attribution, ownership marks, required notices or platform identification.

The core mechanism is AI inpainting: the video text remover and object remover analyze each frame, erase the flagged region (a caption, a lower third, a passerby, a prop) and rebuild the background behind it instead of blurring or cropping over it. Text cleanup works on one selected box, up to three boxes, or the full frame; object cleanup uses either a fixed mask or automatic tracking. The cleaned footage then feeds a translation workbench that produces subtitles and AI dubbing into dozens of target languages from 116 source languages, with downloads of the dubbed MP4, original and translated SRT files, and the voiceover MP3. Direct YouTube link import is not offered; you upload a local file.

Its clearest use cases are short-form video repurposing for creators clearing burned-in captions before a re-edit, a job OpusClip approaches from the auto-captioning side rather than AI inpainting, overseas short-drama localization, where whole series go through in batches, and e-commerce teams turning one product demo into several dubbed market versions. Developers get two routes into the same engine: a self-serve REST API with bearer-token keys, asynchronous jobs, signed webhooks and required idempotency keys, and a hosted MCP server that exposes nine tools to agents such as Claude Code, Cursor and Codex. A Shopify app can also append a finished video to a merchant's product listing.

UnmarkAI is web-only, with no desktop or mobile app; teams that need an installed editor should look at CapCut or Filmora 15 instead. Access is priced on points, with a free starter allotment, monthly and yearly subscriptions, one-time point packs, shared Team plans and custom Enterprise terms; the API bills from a separate prepaid USD wallet. Points are charged by processing time and task type rather than a flat per-video fee, and the plan also sets the per-video size and length cap.

Its terms of service and acceptable use policy require that uploaded content be owned, licensed, or explicitly authorized for editing, and prohibit removing watermarks, logos, notices or attribution in a way that breaches copyright, trademark, or platform rules, with a zero-tolerance, mandatory-reporting policy for child-exploitation material. The vendor has not published an independent third-party security certification, worth flagging for teams with a formal vendor-risk review process.

UnmarkAI is not the right fit for every video job, though. Teams that only need voice dubbing and have no burned-in text to remove first get more from a voice-first tool like ElevenLabs, Murf AI or Fliki, and teams that want lip-synced avatar dubbing should compare HeyGen or Rask AI. Anyone comparing the wider field should start with Hokai's AI video generator directory, the dubbing and video translation roundup, the translation tool buyer's guide, or the video editing and content creation guide; the OpusClip vs CapCut EditPilot comparison and the best AI video generators roundup cover the captioning and general-purpose alternatives in more depth. E-commerce teams weighing the point system against a flat-fee generator should also compare it with Creatify or a template-driven tool like Lumen5. Teams that need to generate new footage rather than clean up existing footage want a text-to-video model like Runway; Hokai's guide to AI image and video creation covers that category split in more detail.

Pricing

UnmarkAI runs on points, with standard prices in USD before tax on its pricing page (table dated 29 August 2026). Free gives 10 points on signup, processes only the first 30 seconds, and free downloads carry a watermark. One-time Point Packs run from Quick Fix ($4.90, 100 points) to the 5,600 Credits Pack ($99).

Individual subscriptions start at Starter ($9.90 a month, 300 points, up to 500 MB and 5 minutes per video, 1 job at a time), then Creator ($19.90, 1,000 points, 1 GB and 30 minutes per video; the first month is $13.90), Pro 3,000 ($49.90, 2 GB and 60 minutes per video, 4 concurrent jobs), Pro 10,000 ($159), Pro 30,000 ($499) and Pro 50,000 ($759). Yearly billing saves up to 30%, from $99 a year for Starter to $6,369 for Pro 50,000. Team plans share points across 3 operation seats with unlimited viewer seats: Team Starter $69 a month (3,000 points), Team Growth $179 (10,000) and Team Scale $519 (30,000), with extra seats at $12 a month.

Enterprise is custom-quoted. Points spend by task: 5 or 10 points per 30 seconds for selected-area versus full-frame text cleanup, 5 or 10 per 15 seconds for masked versus tracked object cleanup, 10 per 30 seconds for base translation (15 for advanced quality) plus 8 per extra language, and 5 to 30 points per 5 seconds for upscaling from 720p to 2K. The API is billed separately from a prepaid wallet, at $0.225 a minute for selected-area text removal and $0.45 for full frame, with top-up discounts of up to 40%; developers comparing that against a dubbing API such as Rask AI or ElevenLabs should check the yearly subscription math too.

Key Features

• Video Text Remover: AI inpainting clears hardcoded captions, timestamps, lower thirds and owned on-screen copy from a single region, several regions or the whole frame, and fills in what was behind them.
• Video Object Remover: Removes ordinary scene distractions such as a passerby, a prop or a cable, using either a fixed mask or automatic tracking for things that move.
• AI Video Upscaler: Enhances cleaned or source footage to 720p, 1080p, source size or 2K as a separate step, priced by the target resolution.
• Video Translation Workbench: Translates and dubs uploads up to 2 GB, returning the dubbed MP4, both SRT files and the voiceover MP3; the short-drama flow takes 2 to 200 episodes per upload.
• Self-Serve REST API: Bearer-token uma_live_* keys, asynchronous jobs, signed webhooks and a required Idempotency-Key on job creation, with 4 concurrent jobs by default and capacity packs for more.
• Hosted MCP Server: The same engine runs as an MCP server at api.unmarkai.net/mcp, giving agents such as Claude Code, Cursor and Codex nine tools with one API key and no SDK.

Pros

• Ships a working self-serve API (bearer-token auth, async jobs, signed webhooks, required idempotency keys, a default concurrency of 4 parallel jobs) plus a hosted MCP server, so the automate-it claim is checkable against real endpoint documentation rather than being sales-only.
• Publishes a full standard-price and point-usage table, including Team plans and API wallet rates, so the cost of a job can be worked out before signing up.
• States a specific, named zero-tolerance policy with mandatory NCMEC reporting for child-exploitation material in its Acceptable Use Policy, a more concrete governance disclosure than most small video-cleanup competitors publish.

Cons

• Free plan processes only the first 30 seconds and stamps a watermark on the download, so judging full-length results requires a paid plan.
• Lower plans cap each video: Starter accepts files up to 500 MB and 5 minutes, and only Pro 3,000 and above reach 2 GB and 60 minutes.
• No published SOC 2, ISO 27001, GDPR, or HIPAA attestation, and the privacy policy does not state whether uploaded video is used to train models, worth checking before enterprise use.
• Web-only with no desktop or mobile app, no direct YouTube link import, and the vendor's own notices say results vary with heavy motion, low source resolution, or complex backgrounds.
• Operated by UnMarkAI Team, an independent developer based in Hangzhou, Zhejiang, China, per its own Terms of Service, rather than an incorporated, independently audited company, relevant for procurement teams doing vendor-risk review.

Visit UnmarkAI Official Website