[Bugfix] Accept 'speaker' as alias for 'voice' in TTS speech API (#2424)

Signed-off-by: marksverdhei <marksverdhei@hotmail.com>
Signed-off-by: marksverdhai <249650165+marksverdhai@users.noreply.github.com>
Co-authored-by: marksverdhai <249650165+marksverdhai@users.noreply.github.com>
This commit is contained in:
Markus / Markandmarksverdhai authored and GitHub committed 2026-04-07 15:07:01 +00:00
1 parent 408365fa0d
commit feefdaee9f
4 files changed
+255 -16

No files matched your search

@@ -2,7 +2,7 @@ import math
from typing import Literal
import numpy as np
from pydantic import BaseModel, Field, field_validator, model_validator
from pydantic import AliasChoices, BaseModel, Field, field_validator, model_validator
_MAX_EMBEDDING_DIM = 8192
@@ -10,8 +10,12 @@ _MAX_EMBEDDING_DIM = 8192
class OpenAICreateSpeechRequest(BaseModel):
input: str
model: str | None = None
# Accept both "voice" (OpenAI convention) and "speaker" (model/internal
# convention) as input keys. Intentionally global — all TTS backends
# (Qwen3-TTS, Voxtral, Fish Speech) use this field for the speaker name.
voice: str | None = Field(
default=None,
validation_alias=AliasChoices("voice", "speaker"),
description="Speaker/voice to use. For Qwen3-TTS: vivian, ryan, aiden, etc.",
)
instructions: str | None = Field(