About the Role

Agentspace.ai runs an in-house generative AI content engine. We produce faceless video channels, AI-driven ad campaigns for clients, and realistic avatar content, at a sustained volume of roughly 14 to 15 finished videos every week.

This role is the person who turns generated raw material into a finished, postable video. Our AI Content Creator writes the script and generates the shots. You take those shots, the voiceover track and the written brief, and you cut the video: pacing, sequencing, sound design, captions, colour, export. You are the last person who touches the file before it goes live, and the quality of the final cut is your accountability alone.

This is not a role where you wait for a shot list and execute it silently. Our briefs arrive as structured documents produced through Claude, and you are expected to read them, work out what they are asking for, and come back with specific questions when something in the brief does not hold together. Reading the documentation is part of the job, not a favour to the person who wrote it.

What You Will Do

- Cut the final video: Assemble generated clips, voiceover and music into a finished short-form video, from first frame to export, working from the written production brief for that video.

- Own retention through pacing: Build the first 3 seconds so the viewer stays, then control cut rhythm, hold lengths and beat changes across the rest of the runtime so the drop-off curve stays flat.

- Sound design and audio finishing: Sync voiceover to picture, place music beds and sound effects, duck and balance levels, clean up artefacts in generated audio, and deliver a consistent output loudness across every video we publish.

- Caption and subtitle craft: Produce word-level animated captions using Submagic or VEED.IO, then correct the timing and the spelling by hand. Auto-generated captions are a starting point, never the deliverable.

- Continuity across generated shots: Our recurring AI personas are generated shot by shot, so faces, wardrobe, lighting and colour drift between clips. You catch that drift and fix it in post through grading, reframing, regeneration requests or cut choices.

- Flag and fix bad generations: Identify glitching hands, warped text, lip-sync failures and frame-rate stutter in generated footage, and either repair it in post or send a precise regeneration request naming the exact clip, the exact timestamp and the exact defect.

- Generate replacement shots yourself: When a single shot is missing or unusable, generate the fix yourself in Higgsfield rather than waiting for someone else to unblock you.

- Run asset discipline: Every generated asset, project file and export is uploaded to the shared Google Drive folder for that video within the same working session it is created. Nothing critical lives only on your local machine at the end of a day.

- Hold the weekly output rhythm: Deliver against a fixed weekly publishing calendar across multiple channels and client campaigns running in parallel, and raise a delay the moment you see it coming, not on the day it is due.

- Document what you build: Write down your export presets, caption styles, colour settings and repeatable sequences so the next editor and the next intern can reproduce your output without asking you.

Tools You Will Use Every Day You will not be trained on all of these from zero. We expect fluency in the conventional editing stack and working competence in at least three of the AI tools named below,



with the ability to learn the rest inside your first month.

Conventional Editing and Post-Production (Required)

- Adobe Premiere Pro for timeline editing, multi-track assembly and export management.

- CapCut (desktop and mobile) for fast short-form cuts, caption styling and platform-native formats.

- Adobe After Effects for motion graphics, text animation, masking and clean-up compositing.

Working knowledge of any of the following is accepted in place of the equivalent Adobe tool, and is a positive either way:

- DaVinci Resolve for colour grading and finishing.
- Adobe Audition for audio repair, noise removal and levelling.
- Adobe Photoshop for thumbnails, still frames and asset clean-up.

AI Editing and Post Automation (Required)

- Submagic for word-level animated captions, B-roll insertion and hook framing on short-form cuts.
- VEED.IO for subtitling, translation, background removal and quick browser-based turnarounds.

AI Generation Stack (Required)

- Higgsfield as the primary generation platform: running image and video jobs, maintaining locked character identities, managing credit spend and diagnosing why a generation came out wrong.

- Image generation models: Nano Banana Pro and Seedream 5 Pro.

- Video generation models: Seedance 2.0, Google Veo 3.1 and Gemini OmniFlash.

Large Language Models (Required) A large part of how we work is written. Briefs, prompt packs and editor handoff documents are produced through Claude, and you will read them, question them and often extend them.

- Claude, ChatGPT or Google Gemini: daily hands-on use of at least one, with the ability to pick up the others quickly.

- Structured problem framing: You can take a vague creative problem, break it into parts, put the right question to the model, judge whether the answer is any good, and iterate. We are testing your thinking, not your typing.

- Working from written systems: Our production instructions are documented. You are expected to read a document end to end before starting work on it, and to ask a specific question when something is unclear rather than guessing and producing the wrong cut.

What We Are Looking For

- Demonstrated editing craft: A portfolio of short-form video you personally cut, showing control over pacing, framing, sound and caption timing. Weight of real work matters more to us than years on a resume.

- Hands-on generative AI experience: You have personally generated video or image assets with AI tools and cut them into finished pieces. You know what generated footage does wrong and how to work around it.

- Short-form retention instinct: You can explain, on a specific video of your own, why you cut where you cut and what it did to watch time.

- Independent problem solving: When a render fails, a model refuses a prompt or a file will not conform, your first move is to diagnose it, not to report it and stop.

- Written communication: You can write a clear message describing what is blocking you, what you have already tried and what you need from whom.

- Volume tolerance: You are comfortable finishing multiple videos a week on a repeating calendar without quality drifting down by Friday.





- Quick responsiveness: You reply on our Slack channel quickly, not hours later, and you are easy to reach by phone call when something needs a fast answer. This role sits in the middle of a fast-moving pipeline, and a slow reply from you slows down everyone waiting on the cut.

Non-Negotiables We work to a fixed publishing calendar, so the following are conditions of the role rather than preferences:

1. Responsiveness: You reply to messages on the team channel within working hours on the same working day. If you go quiet with work in your hands, that is the single fastest way to lose this role.

2. Deadline honesty: A delay raised 24 hours early is a scheduling problem and is fine. A delay raised on the due date is a trust problem and is not.

3. Uninterrupted work setup: You work from our Mumbai office Monday to Friday. For the Saturday work from home day, you need a machine and an internet connection that survive a power cut, whether that is a laptop battery, an inverter or a backup connection. Power outages at home are not accepted as a reason for a missed Saturday deliverable.

4. Machine specification: For the work from home day you need a machine that can run Premiere Pro on 1080p and 4K timelines without stalling.

Valuable to Have None of the following is required. Each of them makes an application stronger, and we are actively interested in AI tools you use that are not on our list.

- Additional AI tools in your kit: Opus Clip, Descript, ElevenLabs, Topaz Video AI, Runway, Kling, HeyGen or Midjourney.

- Hindi and English: Ability to cut and caption content in both languages, including Hinglish, and to catch a mistranscribed Hindi word that an automatic captioning tool got wrong.

- Channel ownership: You have run your own YouTube, Instagram or TikTok channel to real, verifiable numbers.

- Thumbnail and packaging craft: You have designed thumbnails or title cards that you can show performed better than an alternative.

- Colour grading depth: You can match the look of clips generated in separate runs so a sequence reads as one continuous scene.

Culture and Logistics

- The workplace: A small, fast-moving team where you work directly with the founders, ship every week and see your work published immediately. There is no layer between you and the decision.

- Location and schedule: Office based in Mumbai (Jacob Circle / Mahalaxmi), Monday to Friday in person. Saturday is Work From Home.

- Growth: As output scales, this role grows into ownership of the post-production function, including the systems and the people who join it.

- Compensation: Fixed monthly salary in INR, discussed at the first interview and set against demonstrated skill.

How to Apply Send all four of the following. Applications missing the portfolio link will not be reviewed. Here is the Google Form link for the same: https://docs.google.com/forms/d/e/1FAIpQLSdgNw4dSZ5xLzUtDoeMJ3iRg_uw4uBa3Zy_u4NdN206qW0ufA/viewform?usp=dialog

1. Your resume.
2. A link to your portfolio (showreel, published channel, client work or personal projects). Link to work you personally edited, and say what your role was on each piece.

3. One video you cut end to end, with three lines explaining one specific cut decision you made in it and why you made it.

4. Two sentences naming the AI tool you currently use most in your editing workflow and the exact task you use it for.

Shortlisted candidates will be given a short paid trial edit before any offer is made.

📌 Video Editor (Mumbai)
🏢 Agent Space
📍 Mumbai