Skip to content
Story AI ✦ Features Pricing Blog Try Demo — free →

For the Editor in Premiere: AI Metadata via Native XMEML — No Translation Report

Tomislav Brdjanović
Published date:

Premiere Pro has had an XML import bridge since the FCP7 era. It works by translating foreign XML into Premiere’s internal format — and when the source XML contains things Premiere’s bridge doesn’t recognise, like Final Cut Pro’s native .moti title generators, you get a Translation Report. Items get dropped. The sequence you built doesn’t match what lands in Premiere.

AVScript doesn’t go through that bridge.

When you export from AVScript for Premiere, you get an xmeml file — the same FCP7 format Premiere has read natively since version CS5. File → Import. No translator, no Translation Report, no dropped items. The sequence opens exactly as you built it: clips in order, gaps preserved, text captions on V2, source descriptions in the project bin.

The cut stays what you made it.


Your Timeline Stays Yours

AVScript never modifies your Premiere project. It doesn’t open your .prproj, doesn’t reorganise your bins, and doesn’t propose a cut you didn’t ask for.

What it does: it takes the footage you’re about to edit, runs it through AI analysis — transcription, emotion scoring, visual tagging — and hands the result back as native XMEML you import on your own terms.

Your domain — Premiere Pro
Sequence structure and cut decisions
Clip selection and timing
Color, audio mix, delivery
Final approval of everything in the timeline
AVScript — outside your NLE
AI analysis of source footage
Searchable transcripts and metadata
Team review and sequence proposals
XMEML export — sequence and metadata land in Premiere as a native import

The two worlds only connect when you choose to import the XMEML. Until then, your Premiere project is exactly as you left it.


Keep Your Originals. Upload Proxies.

Your camera originals are large. Uploading REDCODE RAW or ARRI MXF to a cloud platform to run AI analysis makes no sense if you already have them on a fast local drive.

There’s a better path: generate proxies, upload the proxies to AVScript, run all the AI analysis on those, then export XMEML and relink to your originals in Premiere.

What you upload to AVScript
🎞
interview_proxy.mp4
H.264 · 1080p · 480 MB
🎞
broll_proxy.mp4
H.264 · 1080p · 210 MB
same timecodes
same duration
What stays on your drive
🎬
interview_original.R3D
RAW · 8K · 84 GB
🎬
broll_original.braw
RAW · 6K · 38 GB
1
Generate proxies for your camera originals
In Premiere: right-click your clips → Proxy → Create Proxies. Or use Ingest Settings to auto-create on import. Proxies share exact timecodes and duration with the originals — that’s the only thing that matters for this workflow.
2
Upload proxies to AVScript
Drag the proxy MP4s into your AVScript library. Upload is fast — these are compressed files. The AI analysis runs on the proxy, but every timecode maps 1:1 to your originals because the duration is identical.
3
Run Story AI analysis
Transcription, Emotions AI, Visual AI — all running on the proxies. One job, done once. The data stays on those files permanently — use them in a new project months later at no extra cost.
4
Export Premiere XML from AVScript
··· → Export Premiere XML opens the modal. Choose your options — captions on V2, descriptions in the bin, FPS if auto-detection misreads it. Export produces the xmeml file.
5
File → Import into Premiere — relink to originals
Premiere opens the XMEML natively. If it can’t find the source files (because you uploaded proxies, not originals), the Link Media dialog appears. Point it at your originals. Premiere relinks by duration match — timecodes transfer cleanly.

Why the relink works

Premiere’s Link Media matches on duration and frame rate, not on filename. As long as your proxy and original were generated from the same camera card and share the same duration — which they always do when created with Premiere’s built-in proxy workflow, or any proper transcoding tool — the metadata transfers cleanly. Your originals never had to leave your drive.


Your Team Works Outside Your Edit

Here’s the part that matters most if you work on documentary, long-form, or any project with stakeholders who have opinions about the cut.

They will always have opinions. The question is where they voice them.

In a traditional workflow that answer is often “over your shoulder at the monitor” or “in a voice note with a timecode they read off a screenshot.” You translate their notes into edits. It works. It’s also slow.

AVScript gives them a better lane — without giving them access to your Premiere project.

🎬
Your Premiere edit
stays untouched
📤
Share player link
any clip, any moment
👥
Team reviews
producer, director, client
New sequence proposed
in AVScript, not Premiere
You decide
import, ignore, or adapt

A producer can open AVScript, look at the analysed footage, build a sequence from suggested clips, and send it to you as a project. You open it in AVScript, see what they had in mind, export XMEML if anything is worth keeping — and the rest never touches your .prproj.

Nobody is in your edit without your permission. But they’re not sending you WhatsApp voice notes either.


The Storyteller’s Workspace

On most productions there’s a person whose job is to shape the story before it reaches the editor. A producer, a director, a journalist, a story editor — call them the storyteller. Their job is to figure out what gets told, in what order, from which takes. Your job is to make that vision real inside Premiere.

The traditional version of this handoff is a timecode log on paper, a shared Google Doc, or a stream of voice notes. Useful, imprecise, and entirely dependent on your ability to translate someone else’s intent into an edit.

AVScript gives the storyteller a dedicated workspace for exactly this.

🪑
In the edit suite with you
Sitting behind you while you cut
Searches all footage by keyword, emotion, or transcript — instantly, without interrupting your flow
Plays back any moment directly in AVScript — no “can you pull that clip up”
Suggests timecodes with actual context: “the line at 14:17, where she says she didn’t sleep”
Never touches your Premiere project — you decide what to act on
💻
At home, on their own laptop
Working independently, async
Browses the full analysed library — same footage, same AI metadata, from anywhere
Builds a sequence proposal at their own pace — pulls dialogue lines into order, structures the story
Exports a PDF script and XMEML when ready — you open it in Premiere in the morning and know exactly what they had in mind
No calls, no screenshots, no “I’ll send you my notes later”

The workspace is built around the dialogue track — interviews, narration, presenter pieces, anything where words carry the structure. The storyteller lays out the verbal spine; you dress it with visuals once it lands in Premiere. When the structure is ready, they export a PDF script — speaker names in caps, frame-accurate timecodes, verbatim dialogue, full asset list — and the same sequence as XMEML, ready to open natively in Premiere.

You use the PDF the way production used to use a paper cut off a Betacam session: as a roadmap. Import the XMEML, use what works, throw out what doesn’t. You’re starting from something precise rather than from a feeling.


What Lands in Your Premiere Project

When you import the XMEML, three things appear in Premiere at once: the sequence, the source clips in the project bin, and the metadata.

📁 Morning News Package — Premiere Pro
NameDurationDescription
📂 Sequence: Chapter 3 — The Decision
🎞interview_studio.mp4
01:04:32:08
Key testimony — breaking down the decision and what it cost the team
🎞interview_day2.mp4
00:48:11:00
Follow-up interview, more reflective tone, admits doubts she didn’t voice on day one
🎞broll_cityscape.mp4
00:12:08:12
Establishing atmosphere — dusk light, city quiet, use under act 2 opening
🎞broll_office_evening.mp4
00:09:44:00
Empty office, after hours — contrast shot for the “nobody left” moment

The Description column in the Premiere project bin shows the notes from AVScript’s clip overlay field. These come from whatever the storyteller or editor wrote during the assembly — tone notes, usage suggestions, line summaries — mapped to <logginginfo><description> in the XMEML. In Premiere you can sort and filter the project bin by this column, or use it as a reference while searching for the right moment.

The sequence itself opens with V1 carrying your clips and V2 carrying the caption overlay — text generators with the clip’s description visible directly on the timeline. Scrub through the sequence and you see the notes over the corresponding footage, without opening any separate panel.

Chapter 3 — The Decision · Sequence
V2
”Key testimony…"
"Follow-up, more reflective…"
"Contrast shot…”
V1
interview_studio
■ slug
interview_day2
broll_office
A1
interview_studio
■ silence
interview_day2
broll_office

Black gaps between clips land as proper slugs on V1 with matching silence generators on A1 and A2 — Premiere imports them cleanly with no audio track length mismatch. Coloured slugs (solid-colour cards) come through as colour matte generators with the exact RGB values from AVScript.

Every source file appears only once in the Premiere project bin, regardless of how many times it’s used in the sequence. That’s the <masterclipid> at work — each source file gets a stable identifier so Premiere groups repeated uses under the same master clip instead of creating a duplicate entry per edit.


The Practical Workflow, End to End

Here’s how a real project runs with AVScript alongside Premiere Pro:

1
Ingest and proxy in Premiere as normal
Import your camera originals. Use Ingest Settings or right-click → Proxy → Create Proxies to generate working copies. Work as you always have — AVScript doesn’t change your ingest process.
2
Upload proxies to AVScript, run analysis
Drag the proxy files into your AVScript library. Configure the pipeline — Transcribe is enough to start. Run it once. The data stays on those files permanently; you never pay twice for the same footage.
3
Your team works in AVScript
Share player links for review. Producers use the transcript search and Story AI to find the moments worth cutting. They propose sequences inside AVScript — not inside your project file.
4
Export Premiere XML when ready
··· → Export Premiere XML. The modal lets you choose whether to include V2 caption overlay, bin descriptions, and FPS if auto-detection is off. The export produces a native xmeml file.
5
File → Import into Premiere
No XML bridge, no Translation Report. The sequence opens with all clips in position and captions on V2. If Premiere shows the Link Media dialog, point it at your originals — it relinks by duration, not filename.
6
Cut
You have a structured starting point, descriptions in the project bin, and captions visible on V2 for reference. Use what works. Throw out what doesn’t. Now it’s your edit.

When the Client Wants to Participate

Sometimes the storyteller is the client. And clients don’t just want to approve — they want to choose. Normally that means a screen share, a feedback thread with vague descriptions, or a call where they try to describe timecodes verbally.

AVScript has a better answer.

🎛
Multicam edit
If the project was shot on multiple cameras, the client opens the multicam view and chooses the cuts themselves — all angles simultaneously, switching between takes. The result becomes a sequence. They feel ownership. You receive a decision instead of an opinion.
✍️
Text video editor
Every transcribed clip can be edited by editing its text. The client reads the transcript, deletes the lines they don’t want, and the video trims automatically. No timeline skills required. They mark up words; the cuts happen.
🔍
Search
Every spoken word across every file is searchable. The client finds the exact moment — by word, by speaker, by emotion tag — and plays it directly. No scrubbing, no “I think it was somewhere in the second interview.”
📝
Transcribe
The full AI transcript is always there — word-level timecodes, speaker labels. Everything the client interacts with maps back to exact frames in the source files.

All of this happens in AVScript, in a browser, without a single login to Premiere. The client participates productively. You receive structured output — a sequence, a marked-up transcript, a timecode list — instead of a feeling.


What AVScript Is (and Isn’t)

The easiest way to understand AVScript is to think about three tools you already know.

Windows Explorer / Finder manages your files. VLC lets you play and inspect them. Resolve’s Cut page lets you arrange them into a rough assembly.

AVScript lives somewhere in between all three — but with a layer none of them have: AI that reads your footage and turns everything it finds into searchable, exportable structure. You manage your media library. You preview clips. You can build rough sequences. But you’re never in a full NLE. The cut — the real cut — stays in Premiere, where it belongs.

It is a metadata engine and pre-edit workspace that sits upstream of your NLE — logging every spoken word, scoring every emotional peak, tagging every scene — so that by the time you sit down to cut, you already know where everything is. It doesn’t propose a structure. It doesn’t decide what matters. It makes the raw material fully navigable so that you can.

What XMEML doesn’t carry: colour correction values, Essential Graphics text edits (V2 captions use FCP7 legacy text, visible but not editable as Motion Graphics Templates), and multicam angle metadata. These limitations are in the xmeml format itself — they apply to all XMEML files in all NLEs. For delivery, remove the V2 track or replace legacy text with your own Essential Graphics titles.

Working with FCP or DaVinci on the same project? AVScript produces FCPXML for Final Cut Pro and DaVinci Resolve as well. The storyteller’s work — the sequence structure, the transcript, the PDF script — travels to any NLE in the chain. Nobody is locked to a single tool.

The story has an essence. A moment where it lands. AI’s job is to clear the path to that moment. Your job is to recognise it and build toward it. That’s the only division of labour that makes sense — and it’s the one AVScript is built around.

The cut is yours. Always has been.

→ Try AVScript with your next project · → Premiere XML in the editor manual · → How Story AI works

Previous
Story AI Pipeline — Live Pricing Calculator
Next
Multicam Editing in AVScript: All Angles, One Export