Add Edge-TTS as a third voiceover provider (free, no API key, incl. Vietnamese) - #46
Open
bin-bard wants to merge 8 commits into
Open
Add Edge-TTS as a third voiceover provider (free, no API key, incl. Vietnamese)#46bin-bard wants to merge 8 commits into
bin-bard wants to merge 8 commits into
Conversation
…ices) Standalone generate_audio()/list_voices() following the same result-dict shape as qwen3_tts.py, so voiceover.py can adopt it as a third provider. No API key or cloud GPU needed. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
argparse only runs %-substitution on epilog text when it contains %(prog); this epilog does not, so the doubled %% rendered literally in --help. The docstring already showed the correct single %. help= strings keep %% (those ARE %-substituted). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds --provider edge-tts alongside elevenlabs/qwen3: argparse choices, --rate/--pitch/--volume flags, a generate_single_audio_edge_tts() wrapper, and a three-way branch through the scene-dir loop, single-file generation, dry-run/JSON output, and brand voice-config resolution. --speaker is reused as the Edge-TTS voice name (default vi-VN-HoaiMyNeural). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Sets the default brand's Edge-TTS voice to vi-VN-HoaiMyNeural, matching the toolkit-wide fallback added in voiceover.py. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
… hints The guided voiceover workflow only offered ElevenLabs and Qwen3-TTS — neither of which supports Vietnamese — so the Edge-TTS provider added in this branch was unreachable from the primary UX. Also adds Edge-TTS as the easiest free fallback in both ElevenLabs failure messages (no API key, no account, no GPU). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A per-scene run reported "success" and exited 0 even when scenes produced
no audio, and the serial loop printed failed scenes as if they had worked
("01.mp3 (?s)"). A silently-dropped scene ships a video with a silent gap.
This was pre-existing in the shared scene loop (it hit ElevenLabs and the
Qwen3 serial path too), but Edge-TTS makes it dangerous: its backend
intermittently returns no audio, so partial failures are routine — and a
failed save leaves a ZERO-BYTE .mp3 behind, which ffprobe reads as no
duration and Remotion renders as silence rather than erroring.
voiceover.py (shared, all providers):
- serial loop prints [FAILED: <error>] instead of a bare filename
- per_scene success is now "every scene produced audio"; adds
failed_scenes; exits 1 when any scene failed
edgetts.py:
- retry transient backend failures (3 attempts, 1s/2s backoff)
- treat a zero-byte write as a failure, and remove the file when the
attempts are exhausted
- fail fast on ValueError (unknown voice, malformed rate/pitch/volume) —
edge-tts validates those client-side, so retrying cannot help
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
brands/default/voice.json and the "Ryan" speaker fallback both defaulted new Edge-TTS scenes to vi-VN-HoaiMyNeural. Fine for this fork's use case, but wrong as the shared default for a general-purpose toolkit: anyone running the documented `--provider edge-tts` example verbatim on English script text got Vietnamese-voice audio. Default is now en-US-AriaNeural; vi-VN-HoaiMyNeural/vi-VN-NamMinhNeural remain available and documented as explicit opt-ins. Also closes registry drift: generate-voiceover's registry entry still described ElevenLabs/Qwen3-TTS only despite the command doc already covering Edge-TTS, and edgetts.py had no standalone tools.* entry. Found via code-reviewer subagent pass ahead of the upstream PR.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds Microsoft Edge TTS as a third
voiceover.pybackend alongside ElevenLabs andQwen3-TTS — free, no API key or account, ~300 voices including Vietnamese
(vi-VN-HoaiMyNeural, vi-VN-NamMinhNeural). Defaults to en-US-AriaNeural.
Changes
backend failures, distinguishes client-side param errors (fail fast) from
transient ones (retry with backoff), never leaves a zero-byte file on failure
dispatch,
--rate/--pitch/--volumeflags, honest per-scene failurereporting (exit 1 +
[FAILED: ...]on any scene that didn't produce audio)— docs and registry entries
Testing
explicit Vietnamese voice, malformed --rate (fails fast, no retry burned),
valid-but-nonexistent voice (retries 3x then cleans up), per-scene batch run
with a forced failure (exits 1, reports which scene failed, no stray files)