6 minMCP & Token Economics

Cutting 9:16 Reels, 4:5 Feeds, and 16:9 YouTube Videos From One Footage Shoot With an AI Video Editing MCP

Quick Summary

Blindly FFmpeg-cropping a rendered 16:9 master into 9:16 or 4:5 slices off typography, split-screen B-roll windows, and platform safe zones. With Mixbash’s Video Editing MCP (https://mixi.mixbash.com/api/agent/mcp), your agent normalizes and registers source clips once via register_media, then opens dedicated 9:16, 4:5, and 16:9 projects via create_project so server workers recompose captions, shot framing, and frame-level visual QA natively for each aspect ratio in under one cent of client tokens.

At 4:10 PM on launch day, someone drops a finished 16:9 YouTube product walkthrough into Slack and asks if an agent can quickly center-crop a 9:16 Reel and a 4:5 LinkedIn cut before 5:00 PM.

One FFmpeg crop=ih*9/16:ih command later, the founder’s left shoulder is gone, half of every product callout sits outside the frame, and Instagram’s bottom UI bar covers the pricing disclaimer.

Aspect ratios are not cookie cutters you punch out of a flattened MP4. On a 16:9 canvas, split-screen B-roll sits comfortably beside a speaker; on 9:16, that same moment needs vertical stacking or a timed full-bleed cutaway with captions lifted clear of Reels and Shorts overlays; on 4:5, mobile LinkedIn and Instagram feed readers expect tighter headline hierarchy without wasting vertical headroom.

Trying to maintain three responsive layouts inside a local coding agent usually triples your headache and triples your frontier token spend.

Through https://mixi.mixbash.com/api/agent/mcp, you pay the upload and transcription cost once and reuse the underlying footage cleanly across formats. Normalize your raw camera clips locally with mixbash-intake-v1 (1080p H.264, CRF 23, AAC), push them to private R2 via prepare_upload and complete_upload, and attach your Whisper word timestamps with register_media.

Your agent then calls create_project for each target format (9:16 for Reels and Shorts, 4:5 for feed posts, or 16:9 for YouTube and VSL embeds) locked to your brand style slug from get_styles.

When plan_visuals and generate_preview run on each project, Mixbash’s server workers compose typography, B-roll windows, and safe-zone margins natively for that exact canvas geometry and verify every scene with automated visual QA.

Even across all three aspect ratios, total client orchestration hovers around twenty short JSON tool calls (~6,000 tokens combined), costing roughly $0.004 on Claude Haiku 5.5 or DeepSeek V4.1 Flash instead of $30+ debugging three separate Remotion compositions on Fable 5.2 or GPT-6.2 Astra.

Be realistic about one workflow detail: each aspect ratio generates its own preview URL via get_project_page and requires its own explicit approve_review before render_final runs. Spending ninety seconds glancing at three browser tabs is slightly slower than a blind shell crop, and it is the exact reason none of your platforms end up with decapitated headlines.

Once all three masters come back from get_artifact, piping them into https://post.mixbash.com/api/mcp lets the same agent queue YouTube, Instagram Reels, and LinkedIn posts in one clean session.

Common questions on this topic

Which aspect ratios does Mixbash’s create_project MCP tool support?

Mixbash’s create_project tool on https://mixi.mixbash.com/api/agent/mcp natively supports 9:16 (Instagram Reels, YouTube Shorts, TikTok), 4:5 (Instagram and LinkedIn feed video), and 16:9 (YouTube long-form, web explainers, and VSLs).

Do I have to re-encode and re-run Whisper three times for three aspect ratios?

No. You normalize source clips once with mixbash-intake-v1 and generate local Whisper word timestamps once on your machine, then reference your clean media metadata across 9:16, 4:5, and 16:9 projects.

Continue reading

TRIAL OFFER

Email us your footage, we will make a demo reel for free.

Whether you want us to run edits for you or want to plug your custom brand style into ChatGPT, Claude, Gemini, or DeepSeek Harness via Mixbash MCP — test it on one real video first.

View Video Editing MCP Setup