Gemini 2.5 TTS Model Not Working at Low Temperatures (original) (raw)

Is anyone having this issue with Gemini 2.5-Flash-TTS?

Model will not produce any output when temperature is set to a low value (<0.5 approx.). Model will simply process and time out with no response, or return a short snippet with the rest cut off (about 2 mins of silence). I can say with some degree of certainty that this is a problem with the model, given that responses are working with higher temperatures, with the only variable changing between my experiments is the temp. This is using the ‘GoogleGenAI’ library in a TS project.

My prompt:
const response = await ai.models.generateContent({
model: “gemini-2.5-flash-preview-tts”,
contents: [
{
parts: [
{
text: `Read aloud in a narration style: ${transcript}`,
},
],
},
],
config: {
responseModalities: [“AUDIO”],
temperature: 0.4,
speechConfig: {
voiceConfig: {
prebuiltVoiceConfig: { voiceName: “Kore” },
},
},
},
});

Successful response (Temp = 0.5):
GenerateContentResponse {
sdkHttpResponse: {
headers: {
‘alt-svc’: ‘h3=“:443”; ma=2592000,h3-29=“:443”; ma=2592000’,
‘content-encoding’: ‘gzip’,
‘content-type’: ‘application/json; charset=UTF-8’,
date: ‘Thu, 02 Apr 2026 07:40:02 GMT’,
server: ‘scaffolding on HTTPServer2’,
‘server-timing’: ‘gfet4t7; dur=21244’,
‘transfer-encoding’: ‘chunked’,
vary: ‘Origin, X-Origin, Referer’,
‘x-content-type-options’: ‘nosniff’,
‘x-frame-options’: ‘SAMEORIGIN’,
‘x-gemini-service-tier’: ‘standard’,
‘x-xss-protection’: ‘0’
}
},
candidates: [ { content: [Object], finishReason: ‘STOP’, index: 0 } ],
modelVersion: ‘gemini-2.5-flash-preview-tts’,
responseId: ‘UR3OadmXLdvJjuMPg_ra0A8’,
usageMetadata: {
promptTokenCount: 149,
candidatesTokenCount: 896,
totalTokenCount: 1045,
promptTokensDetails: [ [Object] ],
candidatesTokensDetails: [ [Object] ]
}

NON-Successful Response (Temp = 0.1):
GenerateContentResponse {
sdkHttpResponse: {
headers: {
‘alt-svc’: ‘h3=“:443”; ma=2592000,h3-29=“:443”; ma=2592000’,
‘content-encoding’: ‘gzip’,
‘content-type’: ‘application/json; charset=UTF-8’,
date: ‘Thu, 02 Apr 2026 07:38:56 GMT’,
server: ‘scaffolding on HTTPServer2’,
‘server-timing’: ‘gfet4t7; dur=148571’,
‘transfer-encoding’: ‘chunked’,
vary: ‘Origin, X-Origin, Referer’,
‘x-content-type-options’: ‘nosniff’,
‘x-frame-options’: ‘SAMEORIGIN’,
‘x-gemini-service-tier’: ‘standard’,
‘x-xss-protection’: ‘0’
}
},
candidates: [ { finishReason: ‘OTHER’, index: 0 } ],
modelVersion: ‘gemini-2.5-flash-preview-tts’,
responseId: ‘EB3OabBVtrKO4w_t-v6gDw’,
usageMetadata: {
promptTokenCount: 149,
totalTokenCount: 149,
promptTokensDetails: [ [Object] ]
}