Arsaze Review: A Video Editor Agents Can Drive

The Timeline, and Why the Transcript Matters

The editor is a conventional non-linear one. Clips sit on a real multi-track timeline with trim modes, keyframes and non-destructive colour, and it behaves like one whether or not an agent is connected.

The transcript is what makes the editing fast. A word-level transcript is produced for every clip, and the timeline is bound to that text. Working from the words rather than the waveform is what makes it quick: pick a sentence in the transcript to remove it, lift the filler words out, close up the pauses and pull the retakes, without scrubbing for markers. Taking dead air out of a talking-head recording is one pass rather than a manual hunt.

Colour is handled per shot and non-destructively. There is a keyer, shot matching, the usual light and colour wheels, curve adjustment, HSL qualifiers and a set of film-stock looks, all applied at export rather than baked in. Your own .cube files can be imported if the built-in looks are not the ones you want.

Driving Arsaze From Claude, ChatGPT or Grok

Arsaze is added as a custom connector over MCP, and the developers name Claude, ChatGPT and Grok as supported today. You pick a project, and from then on its calls are applied to the open timeline rather than happening somewhere else and arriving as a finished file you have to take or leave.

Because the tool calls act on the timeline, the useful requests are the ordinary editing ones. Cutting the silences, placing b-roll where you mention something, grading a scene and re-timing a section are all things the agent can carry out against the same project you would work in by hand.

There is a second renderer as well, which builds motion graphics, title cards and kinetic type as code and renders them onto the timeline. That is a separate path from footage editing and it runs alongside it in the same project.

Generation and Client Review

Generation of video, stills and audio all draws on one shared credit balance, included on every paid tier. Output lands on the timeline as a clip, so generated b-roll does not have to be downloaded and re-imported, and prompts and models can be chained into reusable graphs.

Review is built into the editor rather than bolted on. Sharing a link gives clients somewhere to leave notes: each comment attaches to a particular frame of a particular version, replies nest underneath, and individual scenes can be given a score. The agent can read those notes and make the fix, which closes the loop between a client note and the edit.

Exporting: Cloud, Local or FCPXML

By default the editor is a browser tab and exports render on its own servers, drawing on the monthly cloud rendering hours in your plan.

If you would rather use your own machine, the Windows app exports locally, free and with no watermark. Third, there is FCPXML output, which carries cuts, transitions and captions into DaVinci Resolve or Premiere Pro intact. That is the sensible route when a project has to finish in an NLE you already use.

Pricing: Free for Editing, Paid for AI

The free plan is a real editor rather than a demo. You get 50 one-time credits, and those cover audio analysis of your videos; AI generation and the assistant unlock on paid plans. It also includes an hour of cloud rendering a month, 2 GB of storage, three projects and unlimited local Windows exports with no watermark.

Basic at ₹490 a month adds 150 monthly AI credits, all AI features, 2.5 hours of cloud rendering and 10 GB. Pro at ₹2,450 raises that to 800 credits, 12 hours and 50 GB, and is the plan the developers mark as the director's pick. Studio at ₹9,800 gives 2,900 credits, 60 hours and 250 GB with unlimited projects and capsets.

Two details are worth checking before committing. Plans are billed monthly in INR, and no USD pricing is published. Credits can also be topped up separately, in packs from 100 credits that do not expire, so a heavy month does not force a plan upgrade.

Pros and Cons

The strongest argument for it is that the agent and the editor are the same thing. Most AI video tools generate a clip; Arsaze puts the agent inside an NLE that a professional can also use, which means the output is not a one-shot result you have to accept.

The honest constraints are the free tier and the product's age. Audio analysis is all the credits get you before you pay, so evaluating the AI features properly means paying for a month. Local export is Windows-only, cloud rendering is metered in hours, and the product has been in beta since August 2026, which is early enough that the rough edges are real.

Pricing & Plans

Free plan. Basic ₹490/mo. Pro ₹2,450/mo. Studio ₹9,800/mo. Billed monthly in INR.

Best For

Recommended use cases and scenarios where Arsaze shines.

Frequently Asked Questions

Common questions about Arsaze, answered.