GoCreator turns local slides plus narration text into narrated videos from the command line.

• Reads slide assets from data/slides
• Infers narration from matching .txt and audio sidecar files in data/slides
• Supports PNG, JPG, JPEG, PDF, MP4, MOV, AVI, MKV, and WEBM inputs
• Expands PDFs into one page per slide before rendering
• Translates narration into multiple output languages
• Generates text-to-speech audio or uses prerecorded narration
• Renders per-slide video segments and combines them into final outputs
• Applies optional post-processing such as voice overrides, subtitles, audio mixing, intro/outro clips, exports, metadata, chapters, and thumbnails
• Caches translations, audio, video segments, and PDF preprocessing artifacts

GoCreator is now local-only and CLI-first:

• gocreator init creates a starter project layout
• gocreator create runs the full generation pipeline
• Slides are discovered only from the top level of data/slides
• Slide ordering uses natural filename order (slide2 comes before slide10)

The current media contract is:

• One narration entry is required for every final slide after PDF expansion
• Image slides and PDF pages use narration duration
• Video slides use clip duration by default
• Video slides can instead align to narration duration with timing.media_alignment: slide
• Video slides mix embedded clip audio with narration when the clip has audio

• Go 1.24+ if building from source
• ffmpeg and ffprobe in PATH
• OPENAI_API_KEY set in the environment when TTS or translation is needed
• For PDF input: pdfinfo, pdfseparate, and pdftocairo in PATH

PDF support currently relies on those PDF utilities during preprocessing so multi-page PDFs can be split and rendered into slide assets.

go install github.com/Napolitain/gocreator/cmd/gocreator@latest

go build -o gocreator.exe ./cmd/gocreator

Initialize a project:

gocreator init

This creates:

• gocreator.yaml
• data/slides/
• data/out/
• data/cache/

Add your assets:

• Put images, PDFs, and/or video clips in data/slides/
• Put matching narration files in the same folder using the same basename

Examples:

data/slides/01-cover.png
data/slides/01-cover.txt

data/slides/02-demo.mp4
data/slides/02-demo.wav

data/slides/03-summary.png
data/slides/03-summary.fr.txt

Create videos:

gocreator create --lang en --langs-out en,fr,es

GoCreator now infers narration from files placed next to each slide:

• basename.txt: source-language text for TTS or translation
• basename.<lang>.txt: language-specific text override
• basename.mp3 / basename.wav / other supported audio formats: source-language prerecorded audio
• basename.<lang>.mp3 / basename.<lang>.wav: language-specific prerecorded audio

Inference rules:

• If matching audio exists for the requested language, GoCreator uses it directly
• Otherwise, if matching text exists for the requested language, GoCreator uses TTS on that text
• Otherwise, if source text exists, GoCreator translates it and uses TTS
• If both text and audio exist for the same slide/language, audio wins
• Sidecars can be interleaved in any order; media ordering is driven only by slide filenames

For PDF pages, use the expanded page basename:

• 02-handout-page-0001.txt
• 02-handout-page-0002.fr.txt
• 02-handout-p003.wav

timing.media_alignment still controls video-slide timing:

• video: keep the clip duration
• slide: trim or loop the clip to narration duration

• PDFs are discovered alongside images and videos
• A multi-page PDF is expanded into one final slide per page
• A single-page PDF still goes through the same preprocessing path
• Invalid or encrypted PDFs fail the run
• Expanded PDF artifacts are cached under data/cache/pdf/

If data/slides contains:

01-cover.png
02-handout.pdf   # 3 pages
03-demo.mp4

then your narration sidecars could look like:

1. 01-cover.txt
2. 02-handout-page-0001.txt
3. 02-handout-page-0002.txt
4. 02-handout-page-0003.txt
5. 03-demo.wav

Creates a starter config and project layout in the current directory.

Runs the generation pipeline.

Common flags:

• --lang, -l: input language
• --langs-out, -o: comma-separated output languages
• --config, -c: config file path
• --no-progress: disable the progress UI

1. Load slides from data/slides
2. Expand and cache PDF pages when needed
3. Match per-slide text and audio sidecars from data/slides
4. Translate only the slide texts that do not already have a target-language sidecar
5. Generate TTS only for slides that do not already have a matching audio sidecar
6. Render one video segment per final slide
7. Concatenate segments into a master language render
8. Optionally post-process with subtitles, music, intro/outro, exports, metadata, chapters, and thumbnails

The main create flow actively uses:

• input.lang
• output.languages
• output.directory
• output.format
• output.quality
• output.formats
• voice
• cache
• encoding
• effects
• audio
• subtitles
• intro
• outro
• metadata
• chapters
• transition
• timing.media_alignment
• multi_view

Supported effects in the core pipeline are:

• ken-burns for still-image slides
• text-overlay
• blur-background
• color-grade
• vignette
• film-grain
• stabilize for video slides

Effects are optional. When effects is absent, the normal rendering path stays on the lightweight fast path.

Optional post-processing features now supported by create include:

• per-language TTS voice overrides
• background music, ducking, and timed sound effects
• generated .srt / .vtt subtitles plus optional burn-in
• intro/outro clips or generated template cards
• multi-format export (mp4, webm, gif) with encoding presets
• metadata, chapter markers, and thumbnail generation

When those post-processing sections are absent, create keeps the direct fast path and writes the primary video without extra FFmpeg passes.

The config schema is still larger than the runtime surface: pip remains declarative only.

• examples/minimal-sidecar-tts/ - single image plus .txt sidecar (requires API key)
• examples/video-prerecorded/ - single video plus prerecorded .wav sidecar (no API key)
• examples/video-align-to-audio/ - video plus longer prerecorded audio using timing.media_alignment: slide
• examples/language-overrides/ - mixed per-language .txt and prerecorded audio sidecars
• examples/getting-started/ - five-slide starter project with matching .txt sidecars (requires API key)
• examples/demo/ - two-slide multi-language CLI example with inferred sidecars (requires API key)
• examples/demo-multiview/ - multi-view example with per-slide .txt sidecars (requires API key)
• examples/*.yaml - config schema examples

Common commands:

go mod tidy
go fmt ./...
go vet ./...
go test ./...
go build -o gocreator.exe ./cmd/gocreator
go build -o perftest.exe ./cmd/perftest
go build -o cache-perf-test.exe ./cmd/cache-perf-test

The main runtime path is:

cmd/gocreator -> internal/cli/create.go -> internal/services/creator.go

Core services handle:

• text loading
• translation
• audio generation
• slide discovery and PDF preprocessing
• video assembly
• transitions and multi-view layout composition

GPL-3.0