improvement(chat-voice): modernize ElevenLabs TTS to Flash v2.5#4943
Merged
Conversation
- Switch default TTS model from eleven_turbo_v2_5 to eleven_flash_v2_5 (ElevenLabs recommends Flash over Turbo in all cases; ~75ms latency) - Drop deprecated optimize_streaming_latency knob plus legacy use_pvc_as_ivc / enable_ssml_parsing flags - Move output_format to the query string and raise it from mp3_22050_32 to mp3_44100_128 for higher audio quality - Switch apply_text_normalization from off to auto for correct number/date pronunciation
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
Contributor
Greptile SummaryThis PR modernizes the deployed-chat TTS path by switching the default model from
Confidence Score: 5/5Safe to merge — changes are additive quality improvements with no logic regressions on the audio proxy path. The diff is small and targeted: a model name swap, a bitrate bump, and removal of deprecated fields that ElevenLabs no longer honours. The contract default and the hook default are kept in sync, output_format placement matches the ElevenLabs streaming endpoint spec, and no auth or data-flow logic is touched. No new error paths are introduced. No files require special attention. Important Files Changed
Sequence DiagramsequenceDiagram
participant Client as Browser (useAudioStreaming)
participant Proxy as /api/proxy/tts/stream
participant EL as ElevenLabs API
Client->>Proxy: "POST {text, voiceId, modelId="eleven_flash_v2_5", chatId}"
Proxy->>Proxy: validateChatAuth(chatId)
Proxy->>EL: "POST /v1/text-to-speech/{voiceId}/stream?output_format=mp3_44100_128"
Note over Proxy,EL: body: {model_id, voice_settings, apply_text_normalization="auto"}
EL-->>Proxy: audio/mpeg stream (128 kbps, 44100 Hz)
Proxy-->>Client: streamed audio/mpeg via TransformStream
Client->>Client: arrayBuffer() → decodeAudioData → play
Reviews (1): Last reviewed commit: "improvement(chat-voice): modernize Eleve..." | Re-trigger Greptile |
Replace the legacy Sarah default (EXAVITQu4vr4xnSDxMaL), which has no high-quality eleven_flash_v2_5 base, with Jessica (cgSgspJ2msm6clMCkdW9) — a current premade conversational voice verified against the live account and optimized for Flash v2.5.
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
eleven_turbo_v2_5toeleven_flash_v2_5— ElevenLabs now recommends Flash over Turbo in all cases (functionally equivalent, ~75ms latency, built for real-time/Agents)EXAVITQu4vr4xnSDxMaL) with Jessica (cgSgspJ2msm6clMCkdW9) — Sarah has no high-qualityeleven_flash_v2_5base; Jessica is a current premade conversational voice optimized for Flash v2.5. Verified against the live account.optimize_streaming_latencyknob and the legacyuse_pvc_as_ivc/enable_ssml_parsingflags from the proxy requestoutput_formatto the query string (where ElevenLabs reads it) and raise it frommp3_22050_32(32 kbps) tomp3_44100_128for noticeably better audio qualityapply_text_normalizationfromofftoautoso numbers/dates are pronounced correctly (level-4 latency mode had silently disabled the normalizer)Scope is limited to the deployed-chat voice path (
/api/proxy/tts/stream, its contract, the audio-streaming hook, and the chat default voice). STT (scribe_v2_realtime, single-use token flow) was already current and is untouched.Type of Change
Testing
Smoke-tested the exact new request (Jessica +
eleven_flash_v2_5+output_format=mp3_44100_128+apply_text_normalization=auto) against the live ElevenLabs API → HTTP 200, valid 128 kbps / 44.1 kHz MP3, numbers/dates normalized correctly.bun run check:api-validationandtsc --noEmitpass clean.Checklist