Localize a 30-second holiday ad into 8 languages for about $2.65

Localizing a 30-second holiday ad into eight languages costs about $2.65 in Sume usage: $0.33 per language for voice, assembly and captions, plus one $0.01 transcription of the source. The translation step itself is yours (a person or a language model), so it is not in that number.

The pipeline is four calls: transcribe the original, translate the script, speak the translation, and re-render the video with the new voice and captions.

Per-language cost line by line

A 30-second ad has roughly 450 characters of script. Sume TTS bills $0.0475 per 1,000 characters, so 450 characters are $0.0214, billed as $0.03 after rounding up. Timeline renders at $0.10 per output minute, rounded up, so a 30-second video is $0.10. Standalone captions are $0.20 for videos up to 60 seconds. Transcribing the source at $0.01 per audio minute happens once.

Where this can go wrong

Voice and language must match. TTS checks the requested language against the voice's primary language and returns 409 tts_voice_language_mismatch when they differ, so choose a voice per language, not one voice for all eight. Length also drifts: German and French often run longer than English, so a 450-character English script can become 550 characters and overrun 30 seconds. Check the audio length before you build the Timeline.

For captions, the language hint only tells speech-to-text what to expect and never selects the style or font. A Korean script with a Latin-only style such as slam is refused with caption_hangul_text_latin_style, since those faces render Hangul as empty boxes. Pick a Hangul style such as korean-ad for Korean.

Cheaper variants

For the dubbing pipeline end to end, including audio joins, the guide linked below covers each call.

• Skip burned-in captions for markets where the platform auto-captions: the per-language cost falls to $0.13, and the total to $1.05.
• Caption once on the silent source and reuse authored cues only where the timing does not change.
• Do the two highest-revenue languages first and ship them; add the other six once the ad is proven.