Doza Assist for Documentary Editors.
Twenty hours of interviews, one story hiding inside them. Doza Assist transcribes everything on your Mac, reads the whole corpus, surfaces the beats, and hands your selects straight to Final Cut Pro, Premiere Pro, or DaVinci Resolve.
Doza Assist is a local-first AI editing assistant for documentary work. It transcribes interviews in about 100 languages on your Mac, identifies speakers, analyzes story structure across your whole corpus, and exports selects and stringouts to Final Cut Pro, Premiere Pro, and DaVinci Resolve — for a one-time $199 license. Footage never uploads to any server.
01How does a documentary paper edit work in Doza Assist?
A paper edit — reading the transcripts, marking the lines that carry the film, and assembling a radio cut on paper before you ever touch a timeline — is still the fastest route from a mountain of verité and sit-downs to a first cut. Doza Assist runs that pass with you instead of leaving you alone with a highlighter and a binder:
-
01 · Wrap
The story is in there. Somewhere.
Take an example feature: thirty-two sit-downs, a hundred and twenty hours. Watched down at the usual two hours per hour of footage, that’s six working weeks before anyone makes a cut, and the line you need is always in hour four.
-
02 · Import
Bring the whole shoot in.
Drop in camera files, or the FCPXML you exported from Final Cut Pro or DaVinci Resolve. Multicam and sync clips come in as they are, and nothing is copied: your media stays where it lives. Two or more files become one collection, transcribed and analyzed one after another.
From here on, the window follows one real interview: a 27-minute sit-down with papermaking artist May Babcock.
- MOV · MXF · MP4 · WAV
- FCPXML 1.8–1.14
- Multicam & sync clips
-
03 · Transcribe
Every word, typed on your Mac.
On the M2 Max Mac Studio it ran on, her 27 minutes came back in 17 seconds: 292 segments and a timestamp on every word. On camera files that carry source timecode, one click shows it. Nothing uploads, so there’s no queue and no per-minute meter.
- Word-level timestamps
- Source timecode
- Auto-detects about 100 languages
-
04 · Speakers
Every voice, named.
Speaker ID ran on the same Mac, told the three voices apart and labeled them from what was said: “Mae Babcock,” spelled the way the transcript heard her, a “Reporter” who was really the interviewer, and a “Panelist.” Double-click a word to fix it, rename a voice once in the Speakers panel, and the transcript, clips and exports follow. The timing never moves.
- Built in, no account
- Your renames always win
-
05 · Story
Read the whole interview, end to end.
AI Analysis read all 27 minutes and pulled her strongest soundbites, each with a line on why it works: “Reveals the emotional core/appeal of her art.” Across a whole collection, Story Brief does that for the film: the arcs, the tensions, who echoes whom, and where coverage is thin.
- Gemma 4, running locally
- Checked against the transcript
-
06 · Selects
Mark it up like paper.
Jump to any timecode and paint the lines you want with the five-color highlighter, and name the colors for your film (say chestnuts, process, identity, climate, cold open). Trims snap to word boundaries, so no select starts mid-word.
- 5 renamable colors
- Trim to the word
-
07 · Paper edit
The radio cut, already sketched.
AI Analysis also marked seven story beats, from her craft to the chestnut trees to the land’s memory. Back to back they run ten minutes and fifty-five seconds: a first assembly to react to, not a finished cut. Want a tighter angle? Tell Story Builder what you’re making and it builds a sequence with a role for every clip.
- 7 story beats
- Story Builder
- My Style
-
08 · Timeline
Send it to your timeline.
Tick what to include, choose a pre-cut timeline, cuts and markers, or markers only, and send it. Final Cut Pro gets FCPXML, Premiere Pro an XML to import, and DaVinci Resolve Studio pulls it in for you. On a Final Cut Pro timeline, every clip carries its words in its Notes.
- FCPXML · Premiere XML
- Words in each clip’s Notes (Final Cut Pro)
02Why local matters when the footage is the film
Documentary interviews are frequently the most sensitive material in the building: unreleased access, embargoed subjects, survivors telling a story once. Cloud transcription suites ask you to upload all of it before you can search a word of it. Doza Assist runs the entire pipeline — transcription, speaker ID, story analysis, chat — on your Mac. It works with the network cable unplugged.
That has a practical edge beyond privacy: there is no upload window for a 2TB doc project, no per-minute transcription bill that scales with your shooting ratio, and no monthly seat fee that outlives the production. A 100-hour project costs the same $199 as a 10-hour one.
Collections group a series into one project, and Chat, Story Brief, Story Builder and the Clip Library work across every episode's interviews at once — the character arc that spans episodes 2 through 6, the moment two subjects describe the same night differently. True crime timelines, sports-doc season arcs, and festival sizzle pulls all work the same way: the corpus is one searchable body.
03How is this different from cloud AI editing tools?
Most AI assistants for documentary editors are cloud services: footage uploads, processing is metered by credits or minutes, and access lasts as long as the subscription does. Doza Assist takes the opposite architecture.
| Doza Assist | Typical cloud suite | |
|---|---|---|
| Where footage is processed | On your Mac, offline | Uploaded to vendor servers |
| Pricing | $199 one-time, yours forever | Monthly subscription or credit meter |
| Transcription cost per hour | $0 — unlimited, local | Per-minute or credit-based |
| NLE handoff | FCP timeline, Premiere XML, Resolve | Varies; often export-only |
| Multicam / sync-clip aware | Reads multicam and sync clips from FCPXML | Usually flattened media only |
| Learns your editorial voice | My Style profiles | Generic output |
| Works with no internet | Fully offline | No |
If your deliverable is captioned vertical clips for social, a cloud clipper may fit better — Doza Assist is built for editors whose deliverable is a film, a cut, or a stringout inside a real NLE.
04Frequently asked by documentary editors
Can Doza Assist handle a 100-hour interview corpus?
Yes. Drop in a stack of interviews and Create Collection transcribes and analyzes every one, one after another, unattended. Since processing is local there's no upload time and no per-minute cost — the practical limit is disk space and how fast your Mac chews through audio. 16GB of unified memory is the tested minimum; 32GB or more is recommended for large multicam libraries.
Does it work with my Final Cut Pro projects?
Through its XML. Export an FCPXML from Final Cut Pro (File › Export XML) and drop that in: multicam and sync clips come in as they are, with no re-transcode and no flattening, and finished sequences go straight back to an FCP timeline. Premiere Pro gets an XML to import, and DaVinci Resolve imports the FCPXML.
My interviews are in Spanish and French. Does that work?
Yes — transcription understands about 100 languages and detects them automatically, all on your Mac (the first non-English file downloads the language model once). The AI writing (summaries, soundbites, Story Brief) can come back in any of 30 output languages, set per project — cut a Spanish-language master and deliver notes in English, or the reverse. See every language.
Will the AI actually make my selects for me?
It makes a strong first pass — the strongest soundbites, story beats and social clips in every interview — and with My Style on, Story Builder and Chat work the way you cut. You stay the editor: the 5-color highlighter, selects-aware chat, and word-level trim are built for refining its pass into your cut, not accepting it blind.
Is there really no subscription?
Really. $199 one-time includes the first year of updates; after that, updates are an optional $49/yr and the app keeps working forever either way. There's a 30-day money-back guarantee if it doesn't fit your workflow.
Doza Assist — transcription, Speaker ID, Story Brief, Collections, My Style, and NLE export, all on your Mac.
Free trial: every feature on your first 2 minutes — no account, no card · 30-day money-back guarantee · Requires a Mac with Apple Silicon (M1 or later)