Any language in, your language out
Sources in English, Spanish, Arabic… transcribed in their own language; the script is written in the language you choose.
Documentary
Video → Documentary
Several videos on one topic → your own original, fact-based documentary.
Video → Documentary is Film To Story for facts. You give it several source videos on one subject — in any language — and it merges them into one shot map where every shot knows its source. Your AI writes an original script in your language: a hook fact, idea blocks, a counter-intuitive turn, a takeaway, references — pulling shots from every source and adding full-screen B-roll pictures where footage cannot show an idea. Charts are drawn from the script's numbers, pictures are generated in a chosen visual style, presenters and channel logos never make it in.
- Input2–8 videos (PC, links, YouTube pages, magnets), any language
- OutputOriginal documentary — YouTube 16:9 and/or Reel 9:16
- FactsTranscripts + OCR text of every source, references in the script
- B-rollCharts drawn locally, pictures by AI in a chosen style
- SafetyNo presenters, no logos, no original audio, never mirrored
Who it is for
- Documentary, science, history and explainer channels — especially faceless ones.
- Creators who translate knowledge across languages: sources in English, documentary in Arabic or French.
- Educators who want a structured 8–15 minute piece with references.
- Teams that need consistent visual style across many episodes.
What you get
- An original scriptNever a translation of a source: hook fact first, 3–5 idea blocks, a counter-intuitive turn, a takeaway, in the tone you pick — with a references list.
- Shots from every sourceA merged shot map with a source letter per shot; the prompt caps any one source at 50 % and the importer warns when a source is unused or dominates.
- B-roll pictures & chartsFull-screen pictures the script defines (BR01, BR02…) with their own duration, motion and visual style (photoreal, satellite, blueprint, watercolor, isometric…); charts drawn by the app from the script's real numbers.
- Voiceover, captions, musicSame engines and controls as Film To Story; rhythm marks (pause, fade to black, echo, slow motion) for chapter changes.
- Presenter- and logo-free footageShots of hosts talking to camera are hidden per source; static channel logos are detected and zoomed out, blurred or marked by you.
- Thumbnail, publish texts, cloud syncEverything Film To Story offers on the Publish page, including cross-PC restore.
How it works, step by step
Same eight pages as Film To Story, with a multi-source Import, an OCR-first Analyze and a B-roll card on the Story page.
- 1
Sources
Create the project with 2–8 source videos: PC files, direct links, YouTube / video page links or magnets. Each gets a letter (A, B, C…); a YouTube source takes the video's title as its name. Add or remove sources later; Import downloads only the missing ones. Each source has a "Has a presenter" switch (on by default).
- 2
Analyze
Every source is analyzed on its own — shots, keyframes, dialogue in its own language, descriptions — then merged into one shot map where each shot carries its source letter. A re-run only analyzes new sources.
- OCR on by default: lower-thirds, names, dates, figures and signs (Latin + Arabic + the spoken language's script) become a TEXT column the AI may use as facts; text cards and "subscribe" screens are excluded precisely.
- Presenter filter: shots of a host / interviewee talking to camera are hidden per source; "View all shots → hidden" restores any the filter got wrong.
- Channel logos: a static logo / watermark is detected per source ("top-right") with a choice — Auto (zoom out when it fits, else blur), Zoom out, Blur, Ignore — or draw the rectangle yourself.
- 3
Documentary settings
Subject, optional angle, narration tone (Explainer, Investigative, Nature / calm, History, "Did you know" list), whether the script may name the source channels, plus language and duration.
- 4
Master Prompt
The AI is told to write an original script with true, verifiable facts; to pull shots from every source (≤ 50 % from one); never to use a presenter; to list its references; and, where footage cannot show an idea, to add a B-roll defined once in a
brolllist and placed in the beat's shot list with its own duration, motion and visual style — never any text on it. The prompt lists the project files an AI IDE may open: the shot map, every source's timestamped transcript and the keyframes. - 5
Script + B-roll pictures
The import checks shot IDs, word budget and schema, warns about unused or dominating sources, shows the references and the shots-per-source split. The B-roll pictures card lists every picture the script asked for: charts are drawn by the app; maps, diagrams and illustrations get an AI prompt in their style — generate one by one or all at once with Gemini / ChatGPT, import your own, change style, aspect, prompt, motion or duration. A B-roll without a picture is skipped at render, never blocks.
- 6
Review
Voice engine, model and voice; per-beat generation and regeneration; the shot strip shows B-roll with a teal frame and lets you move or remove it like a shot; rhythm marks per beat; music with ducking.
- 7
Render
Each shot is cut from its own source video; each B-roll is animated with a slow zoom or pan; documentaries are never mirrored (numbers and signs stay readable) and mixed sources get a light unified look. Logo treatment and presenter hiding are applied at render. Captions, thumbnail text and the rhythm switches as in Film To Story.
- 8
Publish
Per-format cards with preview, the unified row (Render again · Download · Copy link · cloud state · Publish) and the thumbnail / cover composer. The publish modal prefills YouTube, TikTok and Meta fields.
Features in depth
Built so the result is yours — in facts, in footage and in look.
Facts with references
Transcripts and OCR text are the fact base; the script lists its references and the app shows them on import.
B-roll in a consistent style
Photoreal, cinematic 3D, flat illustration, watercolor, vintage archive, blueprint, satellite, isometric, ink sketch, neon dark — per picture or unified.
Charts from real numbers
When the script cites figures, the chart is drawn by the app — no hallucinated bars.
Presenter & logo removal
Hosts hidden per source; logos zoomed out, blurred or marked by hand; applied to every shot of that source.
Source balance
Warnings when one source dominates or is unused; the shots-per-source split is shown at import.
Rhythm marks
Pause after, fade to black, echo→black on the last word, slow motion — written by the script, editable per beat, switchable at render.
Script check
Advisory notes per beat (unrelated shots, single shot for a long narration, B-roll-only beats, missing pictures) with a small "fix this beat" prompt.
Cross-PC sync
Restore the project on another PC; sources re-download on demand.
Every setting you control
Nothing is hidden behind a paywall or an "advanced" plan — these are the knobs in the app.
| Setting | Options | Notes |
|---|---|---|
| Sources | 2–8 · letters A–H · presenter switch · logo treatment (Auto / Zoom out / Blur / Ignore / Mark) | Add / remove any time. |
| OCR | On (default) / Off · languages: Latin + Arabic + spoken script | "Read text now" adds it without re-analyzing. |
| Subject · angle · tone | Explainer · Investigative · Nature/calm · History · "Did you know" list | Tone shapes vocabulary and pacing. |
| Cite sources | On / Off | Whether the script may name the channels. |
| Duration · language | Buckets / custom · script language · voice language | |
| B-roll | Style · aspect · prompt · motion (auto / zoom in / zoom out / pan) · duration | Per picture; charts drawn locally. |
| Voice · music · captions · thumbnail | As in Film To Story | |
| Shield | Zoom · grade · vignette (never mirror) | Unified look across sources. |
| Rhythm | Transitions · Audio effects · Slow motion · fade 0.2–2 s |
What you need
- 2–8 source videos or links on one topic
- A Gemini / ChatGPT account (hidden browser) for the script and the B-roll pictures — or an AI IDE for the script
- An ElevenLabs or Gradium key for the voiceover
- Linked accounts to publish
- A GPU is strongly recommended (several sources are analyzed)
Tips & best practices
- Pick sources that disagree or complement each other — the counter-intuitive turn writes itself.
- Leave OCR on: dates and figures on screen are the best facts you have.
- Choose one visual style for all B-roll of an episode; the card has a unify option.
- Use "Read text now" on already-analyzed sources instead of re-analyzing.
- Check the shots-per-source split after import; ask the AI to rebalance if one source dominates.
- Mark logos by hand when detection is unsure — it takes ten seconds and applies to every shot.
Honest limits
What the tool does not do (yet), so you are not surprised.
- Analysis time scales with the number and length of sources; CPU-only machines will wait.
- Facts come from the sources and the AI's knowledge; the references list helps you verify — verify.
- B-roll pictures depend on your AI provider's image generation; charts never do.
- Arabic voiceover requires ElevenLabs.
- No mirroring for documentaries by design — readable numbers beat a stronger shield.
Frequently asked questions
Is the documentary a translation of one of the sources?
No. The prompt forbids it and the importer warns when one source dominates. The script must be original with its own structure, in your language, citing references.
What if a fact is wrong?
Read the references the script lists, edit the narration on the Review page and regenerate only that beat. The Script check also flags beats whose shots do not match what is said.
How are presenters removed?
On Analyze, shots of a host or interviewee talking to camera are hidden per source; you can restore any shot from the hidden list.
Can I use my own pictures for B-roll?
Yes — import a picture for any B-roll, or generate one with Gemini / ChatGPT in the style the script chose.
How many sources should I use?
Three to five is the sweet spot: enough variety for the 50 % cap, manageable analysis time.
Reel format for a documentary?
Yes, the same script renders as a 9:16 Reel with hook title; B-roll aspect can be set per picture.
Build your first documentary
Free download for Windows. Three sources tonight, an original documentary tomorrow.
Free · No watermark · No clip limits · Runs on your PC