Help

AI features in Presentation Builder

Presentation Builder has 3 built-in AI features that work directly with slide objects. They all run on GPT-4o-mini (except VoiceGen, which uses OpenAI TTS).

For AI features in Smart Presentation (AI generation, Restyle, AI on-ramp) see AI features in Smart Presentation. For a standalone translation of entire presentations see Translator.

1. Translate narration text (per slide)

Translates all narration_text (per-slide narration scripts) into a chosen language. Used for:

  • Multi-language MP4 export — rendering MP4 in EN, DE, FR, …
  • Multi-language subtitles in Present mode (toggleable via captions)

Path: Builder toolbar → Translate subtitles (the button only appears if at least one slide has narration_text).

Steps:

  1. Click Translate subtitles → opens a dialog
  2. Choose the target language (EN/DE/FR/ES/IT/PL/SK/UK)
  3. Translate → AI sends all texts at once, returns the translated array, stores it as slide.translations[lang]
  4. After completion, the language picker appears in ExportPublishDialog and Present mode

Cost: ~20 credits per presentation (depending on token count).

Note: this is NOT the same as the Translator tool, which translates the entire presentation (texts on slides, not just narration) and works for both Smart and Builder. This is only about per-slide narration in Builder.

2. VoiceGen — AI voice for narration

Creates an AI spoken narration from slide.narration_text. The generated MP3 is attached as slide.narration_url and plays in Present mode + MP4 export.

Path: Builder → select a slide → Properties tab (slide-level) → Generate voice (admin or opt-in feature flag only).

Steps:

  1. Write the text in Script / subtitles in slide properties
  2. Click Generate voice
  3. Dialog: choose the voice (alloy / echo / fable / onyx / nova / shimmer — OpenAI voices), speed (0.5×–2×)
  4. Generate → backend calls OpenAI TTS, saves the MP3 to R2, attaches the link to the slide
  5. You can preview directly in the dialog

Cost: ~$0.015 / 1000 characters (1 minute of narration ≈ 800 characters ≈ ~$0.012). Converted to credits via lib/openai/logUsage.

Tip: For multi-language presentations, translate first (#1), then generate voice from the translated text — pick the language in the voice dialog.

Multi-language voiceover (dub the translated captions)

Captions (narration_text) can be translated (#1) and each language dubbed separately — the original (Czech) stays intact. The Generate voice dialog has a Voiceover language selector at the top: Original (CZ) or any language the captions are translated into. The selected language pre-fills the text with the translation; the generated audio is saved to a separate slot (Czech is not overwritten). Languages already dubbed show a "✓"; you can play/remove each language version right in the slide properties (Voice language versions).

Switching language during playback or export then plays the matching audio: Present mode and HTML/MP4 export pick the voice by the chosen caption language. Until a language is dubbed, its captions play over silence (Czech audio is deliberately not played under foreign captions).

Full EN flow: Translate captions → EN → on the slide Generate voice → Language: English → in Present/export switch the caption language to EN.

Via the MCP connector your own AI can supply the translated captions for free; dubbing (ElevenLabs) is always a credit-based step in the editor. See MCP connector.

3. AI translation of video subtitles (per cue)

If a slide has a video with subtitles (KonvaVideoNode + cues), you can translate individual cues into another language directly in the subtitle editor.

See Video subtitles — described there.

Cost: calculated based on cue count + text length.

Per-user gating

AI features require credits in the current plan. The free plan with no credits cannot run Translate, VoiceGen or subtitle translation.

You can see your credit balance in the header (avatar → Billing → Credits).

Admin has AI features always available regardless of plan.

Tips

  • Translate only once — if you ever edit the source text, you have to translate again (AI translations don't update automatically)
  • VoiceGen short sentences — TTS sounds better with shorter sentences (clear intonation). Long complex sentences may sound monotonous
  • Translator tool for the whole presentation — if you want to translate not just the narration but all texts on slides, use /tools/translator (lighter workflow for multi-language presentations)