MODULE 05
Tools Implementation: Choose What Works For You or Build Your Own
Two things happen tonight. We put the selection criteria for the Berlin workshop on the table, and we find out what each of you needs from a tutor before 7 September. Everything else on this page — the screening, the tool landscape, the reel built without a timeline — exists to make those two conversations concrete.
Where this sits
18:03 · 7 min
Creator's workflow
- 1Idea and the topic
- 2Research to collect and verify materials
- 3Decide how to present the document
- 4Write a screenplay and record the voiceover
- 5Assemble the visuals with tools that fit you — you are here
- 6Polish: subtitles, visual effects, sound design
- 7Pre-publishing screening and evaluation
Schedule
| 14.08 | 01 Introduction and Finding the Focus |
| 18.08 | 02 Research Phase — AI-powered or AI-distorted |
| 21.08 | 03 Ethics and Techniques of Intervening in Archival Footage |
| 25.08 | 04 Editing as a Way of Thinking Through a Fragmented Reality |
| 27.08 | 05 Tools Implementation: Choose What Works or Build Your Own — today |
| 28.08–01.09 | Consultations, 30 min per group |
| 02.09 | 06 Production: Visual Effects and Sound Design |
| 07.09 | 07 Final Screening and Wrap Up |
| 21–27.09 | Berlin workshop · 16 places |
You have a text and a read. From here on, timing is not invented — it is derived from the audio. That is true in every editor named further down this page, and the last section makes it literal.
Tonight
| 18:03 | Where we sit in the workflow |
| 18:10 | Criteria of selection |
| 18:22 | Questions on the criteria |
| 18:30 | What you need — three lines, then booking |
| 18:40 | Screening — four groups, five minutes each |
| 19:00 | Reflection across the four |
| 19:08 | The tool landscape |
| 19:20 | The 89 reel, built without a timeline |
| 19:30 | Homework for Module 06 |
| 19:33 | Feedback form, close |
| 19:35 | Open floor — booking and questions, tutors stay |
Criteria of selection
18:10 · 12 min · Berlin, 21–27 September, 16 places
The criteria describe readiness to make a film in a room in six days. They do not describe talent, and they do not rank you against each other on taste.
Two criteria
| № | Criterion | Read as |
| 1 | Feasibility of implementation in person | the level of detail in the concept · the presence of a storyline · the amount of material gathered |
| 2 | Relevance of the chosen focus to the workshop's theme | your focus answers “How do wars end?” and can be defended as answering it |
What each sub-point means
- Detail in the concept — a named focus object, a stated question, a stated answer. The Module 01 move: topic → document → focus.
- Storyline — the order in which the viewer receives things. Result → Question → Story → Lesson, if you use our frame. Any deliberate order counts. No order does not.
- Material gathered — files you have, verified, with sources. Not a list of what exists somewhere.
A strong idea with no images and no line cannot be made in Berlin in a week. A modest idea with forty verified images and a clear line can. That asymmetry is the whole of criterion 1.
Not criteria
- —Technical polish
- —Which editor you used, or how many effects are in it
- —Fluency in English
- —Whether the rights to your material are cleared yet
Procedure
| When | immediately after the final module, 7 September, on what has been submitted by then |
| Who assesses | the tutors, against these two criteria |
| Who decides | Museum Berlin-Karlshorst |
| Unit | individual places — 16 of them |
| How you hear | individually |
| Precondition | completing the online intensive: one finished short video submitted |
Questions on the criteria and on the date are taken now, for eight minutes. Everything else goes to the channel or to your consultation.
What we need from you
18:30 · 10 min · shared doc
One question, in writing: what stands between you and meeting those criteria by 7 September?
Three lines, in your group's section
- 1What we are stuck on right now
- 2What we want a tutor to do with us in thirty minutes
- 3Who we think that is — a name, or “don't know”
Four minutes of silent writing, then we read the four sections aloud and the named tutor confirms or redirects in one sentence. No discussion here — the discussion is the consultation itself.
The consultation slot being filled tonight is the one between this module and the next: 28 August to 1 September, thirty minutes. It is in the curriculum, it is already funded, and it is the last one before the production module. Book it before you close the laptop.
Screening — the advertising reel
18:40 · 20 min · four groups, five minutes each
The assignment, as set
| Task | a 40–60 second sequence advertising an exhibition about the capture of Berlin at the Karlshorst museum |
| Material | pre-cut shots from two films only |
| Film A | Berlin, 1945, dir. Yuli Raizman — best documentary, Cannes 1946 |
| Film B | The True Glory, 1945, dir. Garson Kanin and Carol Reed — Academy Award |
| Condition | it must work as advertising and at the same time not lie about the material |
| Brought | the sequence, and an explanation of the choice of shots and of that structure |
One Soviet film, one British-American. Advertising by definition wants to stitch things together and to convince; that makes it constructive by definition. Holding both requirements at once was the task, and how was left to each group.
Five minutes per group
- 1 minthe sequence plays — no introduction
- 1 minthe group explains the shots and the structure
- 3 minresponse — one tutor leads
The question asked of every group, in the same words: where in your sequence does the advertising win over the material — and did you decide that, or did the edit decide it for you?
What the response stays on
- Which cut asserts a causal link the footage does not carry.
- Where a shot from one film is made to stand for the other film's claim.
- Where the soundtrack does the persuading instead of the images.
- Where the group chose to leave a seam visible, and what that cost.
Running order is drawn at the start of the block, so nobody prepares to go last. At five minutes the next group plays, finished or not — what does not fit goes into your consultation.
Reflection
19:12 · 8 min
Not a fifth review. What was common to the four.
Three positions to hold against each other
| Bazin | the shots are given the logical structure of oratory, and the speech acquires the credibility of a photograph. An illusion of proof arises where there are only ambiguous facts held together by the commentator's words. The objection is to the method, not to the content |
| Soft montage | two images placed side by side that do not resolve into one conclusion and hold the tension instead. It makes familiar footage strange again |
| The chain | event → recording → selection → preservation → archive → digitisation → editing → viewer. At every stage decisions are made |
Tonight's homework added one more link to that chain, and every group felt it: advertising.
Four groups fought the same fight in four different editors. The fight was about order and about claim. The editor was almost irrelevant to it. That is the argument of the rest of this page.
The tool landscape
19:20 · 12 min
The reference is Wikipedia's Comparison of video editing software. Around twenty rows. We are not reading it — we are learning which four columns decide anything for you.
The four columns that decide
| Column | Why it decides |
| Licence and price | free tier, subscription, or one-off. Decides whether your whole group can use the same thing |
| Operating system | Final Cut is macOS only; most open-source editors are cross-platform. Decides whether a project can be handed over at all |
| Subtitles and export | imports SRT, burns it in, exports 1080×1920 H.264 at 30 fps. Our deliverable is vertical |
| Scripting or project format | whether something other than your hands can write the timeline. This column leads to the next section |
Shortlist
| Tool | Licence | Use it when |
| DaVinci Resolve | free tier · paid Studio | you want edit, grade and sound in one place |
| Kdenlive | free, open source | you want a full editor with no account and no cloud |
| Shotcut | free, open source | you want the lightest full editor that opens on old hardware |
| OpenShot | free, open source | this is your first timeline |
| Blender VSE | free, open source | you are already in Blender for something else |
| LosslessCut | free, open source | you only need to trim and join, without re-encoding |
| Premiere · Final Cut · Avid | subscription or paid | you already have it through work or school |
| CapCut and mobile editors | free with account | speed — and see the note below |
| ffmpeg and a script | free, open source | the edit is a rule rather than a series of gestures |
One constraint, because it changes what you do this week
CapCut is the fastest way to a finished reel and the one most of you already know. It is still not a way for us. Four reasons, heaviest first:
Why not CapCut
| Built-in assets | sounds, music, fonts and stickers are licensed for use inside the app. A reel published under the museum's name and the funder's cannot be built on them |
| What you upload | the terms grant the platform a broad licence over uploaded content. Putting unpublished Karlshorst material there is the museum's decision to make, not yours |
| Not reproducible | the project cannot be handed to anyone else, rebuilt, or described by a specification. This course rests on disclosed intervention and traceable provenance |
| Practical | free-tier limits, auto-subtitles among them, and an export path that periodically runs into a paywall |
If you are already working in it, keep the built-in effects out of the final mix and take sound from a CC0 source — freesound.org filtered to CC0, or the Internet Archive. Note where every sound came from; the credits will need it either way.
The table has about twenty rows and not one of them will make your film good. What the table decides is whether you can finish it, on your machine, with your group, in the time that is left.
Without a GUI — the 89 reel
19:20 · 10 min · the far end of the same spectrum
Shown in Module 03 · 2:44 · no timeline was opened at any point in making it
Two ideas the whole thing rests on
- 1The audio is the master clock. The narration is recorded first and never touched; every shot's duration is derived from where its sentence falls in the read. Nothing is dragged, because there is nothing to drag
- 2The script is known text. We are not transcribing — we align text we already have to audio we already have. No proper noun, rank or date can be misspelt in the subtitles, because none of them is ever recognised
The only file a human writes
shots.csv — twenty rows, one per shot. It holds the single decision a machine should not make: which image stands with which sentence. Four of its rows:
| Cue | Media | Motion | Note |
| 0:00 | Karlshorst-Foyer_square.png | square_in | the foyer wall carries both dates — opens square, pushes to full-bleed |
| 0:42 | Reims-Signing_Jodl.png | zoom_in | “a staff officer with no real power” — Jodl from behind. The wide has only 5 % of zoom headroom |
| 1:01 | MBK_KAPFILM_FINAL.mp4 | fit | Tempelhof. Banded: full width, no enlargement — a 9:16 crop would need 2.4× |
| 2:10 | map_baltics.mp4 | zoom_in | Latvia, Estonia and Lithuania wash in one at a time, in the order the line names them |
Two commands
# align: reads script.md and narration.wav, writes timings python tools/generate_shots.py --project projects/89 --lang en # render: writes 1080×1920 H.264, with .ass and .srt beside it python tools/render_reel.py --project projects/89
Between them there is a report to read, and it is the part worth showing. Every shot carries a confidence figure — how well the read and the script agree there — and warnings about the images:
shot 02 cue 0:06 6.06– 15.12s ( 9.06s, 272 fr) conf 0.83
Karlshorst-Prepared.jpg pan_rl max_zoom 0.96x [CLAMPED]
CLAMPED means this image does not have the pixels for the move asked of it, so the move was reduced. The image was not quietly faked bigger. That decision is the difference between a tool you can trust with an archive and one you cannot.
- Narration
- 164.17 s, recorded human voice, passed through untouched
- Output
- 1080×1920, 30 fps, 4925 frames, 20 shots
- Alignment
stable-ts— timings computed from known text, never transcribed- Render
memoacts_core+ ffmpeg with libass- Human input
- one CSV of twenty rows
- Re-recording
- replace the wav, run the same two commands. Every boundary and every caption recomputes
That last line is the point. Re-reading the narration costs an evening in a timeline and costs nothing here, because no timing is stored anywhere that survives a new read.
What it cost
It took longer to build than the reel took to make. It pays from the second reel, and only if the reels are alike. And it does nothing that matters: finding and verifying the images, and deciding which image goes with which sentence, stayed human work from beginning to end.
Do not install it tonight. It is not the homework. It is shown so that the middle of the range looks like a choice rather than a default.
Discussion
- Everything derivable was derived, and one file was left for a person. Is
shots.csvthe right cut between the two? What would you have kept, and what would you hand over? - A timeline shows you the whole film at a glance. A CSV does not. What do you stop noticing when you stop seeing it?
- The tool refuses to enlarge an image beyond its pixels. Should a tool be allowed to refuse you?
The point
19:30 · 2 min
There is a wide range of tools. You can choose one, or build your own. Neither decision makes the film. What makes it is the story you are going to tell and your picture of how it reaches the viewer. The tool is chosen after that picture exists, and it is chosen to fit it.
Homework
Due before 18:00 CEST, 2 September
1 — Settle your editor
If you have no reason to prefer anything else, keep Shotcut or try DaVinci Resolve. Confirm three things before Module 06: it opens your files, it exports 1080×1920 H.264 at 30 fps, and it imports SRT.
2 — Write the script for your own reel
The text as it will be read aloud, in the order the viewer receives it, with the images you intend for each part named beside the lines.
The script is what criterion 1 reads as detail in the concept and presence of a storyline, and it is what your consultation can work on. A cut made this week would be built on a text that is not settled yet.
3 — Upload tonight's group video
Into the shared folder, with a short document beside it listing the names of everyone in your group.