Repository navigation
feat: Gemini 3.8 TTS dialect + corpus enforcement (0.5.0) - #7
Merged
Merged
Conversation
New audio-tag dialect for gemini-3.8-flash-tts / -flash-lite-tts, mirroring the ElevenLabsV3 no-SSML architecture: - [laugh]/[sigh]/[gasp]... -> <laugh>/<sigh>/<gasp> angle-bracket vocal bursts (aliases normalized to documented spellings, e.g. [cheering] -> <cheer>, [whew] -> <phew>, [ahem] -> <throat-clearing>) - breaks -> <short pause>/<long pause> (two steps; sub-beat -> ellipsis) - emphasis (moderate+strong) -> CAPS, Gemini's documented mechanism - backchannels [mhm]/[oh]/[wow]... -> plain text (transcript-as-speech) - sound effects [applause]/[boo] dropped (prompting guide: avoid SFX) - (text)[whisper] -> <whispers> prefix tag - rate/pitch/volume/style dropped inline: turn-level speech_metadata.style is the documented channel (engine concern) - IPA/say-as -> the word itself (native normalization) - gemini_capabilities() documents the dialect and its limits
…back; single whisper prefix Loop-1 reviewer findings: - modifier_caps_text treated emphasis:"reduced" as CAPS — a semantic inversion (reduced is *less* emphasis, CAPS is Gemini's strongest form). Reduced now passes through unchanged. - expressive_mapping's wildcard returned PlainText while two comments claimed 'best-effort vocal burst pass-through'. Comments now state the actual policy (unknown tags degrade to plain text); dead let binding removed. - section_prefix could emit <whispers> twice when both spellings were present; now guarded to at most one prefix.
[chuckle]/[chuckles], [snicker], [snort], [sob], [shriek], [shout], [growl], [grunt], [hiss], [grr], [breath], [exhales], [argh] — all on Google's documented Gemini 3.8 burst list — now parse as Expressive nodes; the gemini dialect maps them to their documented <angle-bracket> forms. Output-neutral on every other platform (Expressive renders verbatim [tag], identical to the previous literal fallthrough). Review loop 4 flagged [chuckle] as unmapped; this fixes it at the parser level so the whole documented burst set is first-class.
- integration runner gains a .gemini.ssml arm (11 fixture sets now byte-enforced against Platform::Gemini) - tests/test-data submodule bumped ee7460e -> 22663dc (gemini fixtures; lands on master when PR #6 in speechmarkdown-test-files merges) - cargo fmt (includes some pre-existing drift)
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.
Platform::Gemini + GeminiFormatter (angle-bracket vocal bursts, two-step pauses, CAPS emphasis, plain-text backchannels, SFX dropped), gemini_capabilities(), corpus enforcement via 11 .gemini.ssml fixture sets (submodule bumped), parser vocabulary extended with the documented Gemini burst set. Version 0.5.0.