You said you want AI to do the manual labour without leaving DaVinci. Everything in this stack respects that. Half of what the internet tells you to buy, you don't need. The other half, you need the right version of.
"I pumped out the entire timeline as an SRT file. It's all text but the text has the time code. I'll just go on the timeline and look at 16 seconds to 24 seconds."
"Obviously, that process can all be cut down instead of being manual. I don't know how to do it yet."
That's what The Bridge fixes — it takes whatever Claude hands you (SRT, markdown, bracketed timecodes) and emits a DaVinci timeline with all your beats already named and placed.
The picks are opinionated. Where Codex and Gemini both told me the same answer, I listened. Where they disagreed, I went with the one that fits a DaVinci-first, paid-client workflow.
One Python file. No dependencies, no API key, no internet. You point it at the file Claude gave you and it writes an FCPXML you drag into Resolve.
When you open the timeline, every beat is already placed at the right timecode, named (`HERO INTRODUCTION — Architect walks…`, `VILLAIN PRESSURE — Developer's tower…`, …), with a role-tagged marker (`[HERO]`, `[VILLAIN]`, `[CRESCENDO]`) at each beat's start. Resolve doesn't render the marker colours the page mocks up — only the labels — but you scan them at a glance from the markers row.
It creates named gap placeholders + markers, not media clips. Your job is still to drop the actual shots in. The Bridge buys you the timeline scaffolding.
The manual timecode hunting ends here.
You said you already treat Claude like a second brain — make that official. In Claude.ai (Pro), create a Project with custom instructions: "output every outline as ## HH:MM:SS — BEAT NAME headings, beat names from EXPOSITION / HERO / VILLAIN / VALLEY / CRESCENDO / RELEASE." Now every conversation gives you Bridge-ready output.
Paste Claude's output into a text file. Save it. Or skip this step entirely — The Bridge also reads SRT and bracketed-timecode formats.
python3 bridge.py outline.md --fps 24 — that's it. Writes outline.fcpxml next to your input.
File → Import → Timeline. Pick the .fcpxml. Your timeline is now populated with named gap placeholders + role-tagged markers sitting at exact timecodes.
Stylised mockup of the timeline you get from the sample outline (6 beats, 24 fps, 2:17 long). The markers row carries storytelling tags. V1 carries named placeholder clips at exact timecodes. You stop hunting, you start filling.
Drop the right shots under each named clip. Your creative judgement still runs the show — the manual navigation just stopped costing you hours.
Beyond the Bridge: five more CLIs, five practical playbooks, a Claude.ai Project starter pack, and a Claude Code onboarding guide. All offline, all stdlib-Python, no subscription anywhere. Download what you want.
filmkit new "Anya Wedding" creates the full client folder tree (00_ADMIN → 08_ARCHIVE), drops a Claude brief template, writes a START_HERE.md telling you the next 6 things to do. Every other tool plugs in.
Dual-destination copy with MD5 manifest. Verifies every file on every drive before it lets you format the card. "Never format until ingest exits 0." One bad card = relationship over.
Scans a folder, flags clips with dead air, audio clipping, black frames, tiny file sizes. Outputs sortable markdown + CSV. Catches bad takes before they slow your edit.
Reads Claude's outline (markdown or SRT), emits FCPXML 1.9. Resolve imports your storytelling beats as named placeholder clips with role-tagged markers, sitting at exact timecodes.
ffmpeg wrapper. Given a master + cut sheet + aspect ratio (9:16 / 1:1 / 16:9), renders normalised cuts with -14 LUFS audio loudness. One command replaces 60 minutes of click-cut-export.
Three sample outlines (markdown headings, bracketed timecodes, SRT) + three generated FCPXMLs at 24 / 25 / 23.976 fps. Open them in Resolve before you run on your own footage so you know what to expect.
The single most-important runbook. Card labelling, dual-drive ingest, what-to-check-before-format. Read this once before every wedding. Lose one card and the relationship is over.
Top-to-bottom checklist for the most common Resolve emergency. Drive renamed? Codec issue? Cache mismatch? Six diagnostic steps + the nuclear options when nothing works.
The 7 phases in the right order: radio edit → outline → audio-only assembly → visual coverage → picture lock → music/grade/sound → deliverables. Most editors burn weeks doing this in the wrong sequence.
The 12-beat scaffold. Anticipation → VO hook → prep → travel → first-look → ceremony → vows → kiss → speeches → first dance → party → goodnight. Drop into Claude as a template; cut to 7-9 min.
How to set revision-round expectations in the brief, how to handle "one small change" creep, how to picture-lock formally. The protocol that turns runaway projects into profitable ones.
3-minute setup. Upload these files to a Claude.ai Project once. Every conversation in that Project now operates on your editing rules, your story frameworks, your Bridge format. Stops you re-prompting from scratch.
The custom instructions you paste into the Claude Project settings field. Locks Claude into Bridge-ready output by default. Knows your beats, your FPS rules, your default formats.
Six-beat structure with timing percentages, anti-patterns, and variations per film type (wedding / founder / craft / investigative). Claude pulls from this when asked to structure a film.
The exact input format the Bridge CLI parses. Claude follows this so its output is one-shot Bridge-ready. No reformatting, no manual fixing, no parsing errors.
The template Claude uses when you ask for a music brief. Output is Artlist-search-ready: vibe, BPM range, energy curve, two reference tracks, what to avoid. Skip the 800-result rabbit hole.
You already use Claude.ai. Claude Code is the same brain running in your terminal with access to your tools. Install in 5 min, then 10 copy-paste tasks that teach you to improve your kit without writing code.
A one-time payment. It already replaces five separate AI subscriptions other "creator stacks" tell you to buy. If you're still on free Resolve, upgrade before you touch anything else on this page.
Resolve 20 Studio's Neural Engine ships these natively. No round-trips. No subscriptions.
You probably won't. Most editors run with three to five of these. The table is here so you can see the worst-case number before anyone surprises you with it.
| Role | Pick | Cost | Fluff |
|---|---|---|---|
| Everything (worst case) | Full subscription stack | ~$225/mo + $295 one-time | — |
| Recommended starter | Resolve Studio + Bridge + Gling + Artlist | $55/mo + $295 one-time | — |
High-end weddings and Dubai documentaries are reputation-sensitive. Blind DM automation kills the relationship before it starts. Everything here is research + drafting automation, not send automation. You approve every outbound. Always.
Wedding photographers shooting luxury venues are repeat nodes, not one-off couples. Offer "same-day social teaser for your couple + tagged collab reel". One photographer = 8–20 weddings/year.
Scrape London's premium venues (Aynhoe, Brunel, Tanner Warehouse, etc.). Pull their preferred-supplier lists. Planners on those lists shoot 30+ weddings/year — they're the gatekeepers, not the bride.
Not a generic portfolio link. Each warm planner gets a 1-page Cloudflare deploy: 3 films in their tone, 1 line about their clientele, 1 suggested reel concept. I can build the generator — same shape as this page you're reading.
After a wedding appears online, send venue/planner: "Saw the X event. If you're building a preferred-filmmaker list, here are 2 films in that tone." Better than chasing couples directly.
LinkedIn / Crunchbase / event sites for new funding rounds, new offices, major launches, government partnerships, founder awards in UAE/GCC. Every trigger is a documentary brief. Daily digest, you pick which to pursue.
(1) Founder journey, (2) City/heritage story, (3) Capital project/transformation. Each is a 90-second-structure + visual references + budget tier. When a prospect surfaces, you swap the company name and you're 80% of the way to a pitch.
Scrape upcoming Dubai conferences (business, real estate, culture, sustainability). Offer "documentary recap" not "event coverage" — turns a one-day shoot into a brand film. Different price tier entirely.
Don't fully translate — but every outbound has an Arabic subject line + English body. Signals you understand the market. Cheap signal, real ROI in GCC.
Mirror of my job-bot — scrapes the targets above, drafts personalised emails referencing the prospect's specific recent work, queues them in a Telegram bot. You tap ✅ to send, ❌ to discard, ✏️ to edit. Never blind-fires. One-time build, free to run.
Reads replies, tags warm/cold/won/lost, daily digest of who needs a follow-up. Stops leads dying in the gap between send and reply.
ManyChat-style autoresponders kill premium-market relationships. The cost of one badly-aimed DM to a Hampstead bride is the cost of every wedding she ever talks about. Manual approval, always.
Metricool ($25/mo) for IG scheduling + post-time analytics across @justafilmmaker + @darb.agency. Only worth it once you're posting 3+ reels/week and need the rhythm data.
Everything on this page is free or one-time. Nothing is recurring unless you choose Resolve Studio ($295 once) or a subscription tool. Start at day 1 — the kit unfolds from there.
If anything in the kit breaks or doesn't fit your workflow, tell me. Iterating on this is faster than reading 10 YouTubers.