Feature request: silent mode / no-TTS option (bring-your-own voiceover)

[iavankda]

Summary

Add a "silent mode" option to POST /api/short-video that produces the video without TTS narration, returning visuals + captions + optional background music only. Users who want spoken audio record their own voiceover externally and mux it in post.

Proposed API shape (either works):

{
  "config": { "voice": "none" }
}

or

{
  "config": { "tts": false }
}

When set, the render skips the Kokoro TTS step entirely; timing comes from durationInSeconds per scene (or the sum of paddingBack values).

Motivation

Non-English content is blocked today

Kokoro only ships English voices (28 in af_*, am_*, bf_*, bm_*). Submitting non-English text either produces phonetically-mangled pronunciation or crashes the render process — I hit a pod restart today submitting Brazilian Portuguese text, losing all in-memory video state for that session. The current workarounds are unpleasant:

1. Use English TTS and live with robotic/mismatched narration for non-EN audiences (not viable for brand work).
2. Render with placeholder English text, then strip audio with ffmpeg, then mux in a human voiceover — three manual steps per video.

A silent mode would collapse that to one render + one mux (or zero mux if music-only is fine).

Beyond non-English users

Silent mode is also valuable for creators who want:

• Human voice for trust — founder narration, customer testimonial audio, recorded interview clips. TTS can't substitute for a known voice.
• Multi-language distribution — render visuals once, overlay different VO tracks per locale.
• Higher production quality on a budget — self-recorded VO on a decent mic beats Kokoro for brand content, at zero API cost.
• Integration with other TTS providers — users who already pay for ElevenLabs/Azure/Google TTS in another pipeline can feed output into the mux step.

Proposed behavior

• If config.voice === "none" (or config.tts === false) is set, skip TTS entirely.
• Captions are still rendered on-screen if config.captionPosition or config.captionStyle is set — they're visual elements, not audio-derived.
• Scene timing:
  • If durationInSeconds is set (global or per-scene), honor it.
  • Otherwise, use the sum of paddingBack values, or a sensible default (e.g., 3s per scene).
• Output: MP4 with video track + optional music track + no voice track.
• GET /api/voices should include "none" in the list (or null) for clients enumerating options.

Related

• Feature Request – Add API for Voice Only (TTS MP3) and Captions for Uploaded Video #42 — voice-only (TTS MP3) is the opposite shape of this request; the two are complementary.
• Option to disable music or to set it to none in the video creation #24 (closed) — disabling music was a similar "turn off one stream" ask; the pattern is already in the codebase.
• Gemini API TTS request #46 — alternative TTS providers (Gemini) would fit naturally under the same ttsProvider switch once the "none" case exists.

Context

Using the service via https://remotion.abckx.com.br in production for a small Brazilian-Portuguese content holding. Happy to test a PR or provide real-world payload examples if that helps. Thanks for the work on the project — the REST API is really clean once you get past the language limit.

[Anil-matcha]

@iavankda Looking for a free open-source alternative with no watermarks and no per-clip credits? Check out AI-Youtube-Shorts-Generator — turns any long-form video into viral 9:16 shorts using LLM highlight detection, Whisper transcription, and auto vertical cropping.