Gemini 3.8 Live migration errors and Sume's request-response audio
Moving to gemini-3.8-live: thinking_level unsupported, async function calling by default, proactive audio always on. Where Sume's audio jobs fit instead.
Google's Gemini 3.8 Live page lists what breaks when you move from gemini-3.1-flash-live-preview: the model string changes to gemini-3.8-live, thinking_level is not supported, async function calling (behavior: NON_BLOCKING) is the default, and proactive_audio: false returns an error. Those are live-session concerns. Sume's STT and TTS are request-response jobs with a bounded wait of up to 30 seconds, then polling.
Sources: Google's Gemini 3.8 Live page and Sume's API reference, read 2026-10-01.
What do I change when I migrate?
| Item | Behavior on gemini-3.8-live |
|---|---|
| Model string | Use gemini-3.8-live |
thinking_level | Not supported; omit it and thinking_config |
| Function calling | NON_BLOCKING by default; BLOCKING for the old mode |
| Proactive audio | Permanently enabled; false returns an error |
| Affective dialogue | Removed; delete enable_affective_dialog |
| Response modality | Audio; enable output transcription for text |
What does a Sume audio job look like instead?
A Sume STT or TTS call is not a session. You submit a request and get a job: with mode: async the response returns status_url, result_url, events_url and cancel_url; sync and subscribe wait up to wait_timeout_seconds, at most 30. If the job is still running, poll status_url rather than resubmit. The 30-second ceiling bounds the HTTP wait, not the job.
Which one should I use for what?
Use a live model when a person talks back and forth with the system. Use Sume's jobs when you have a finished script to voice or a recording to transcribe: TTS takes a transcript of up to 20000 characters and a required Router model, and STT returns words[] timings under the public id sume/stt-1.0.
Does Sume offer real-time conversation?
This post makes no such claim; the cited sources describe request-response jobs only. For the avatar side of the question read real-time or rendered avatars.
Sources
Related posts
More in Developers
- Gemini 3.8 TTS returns WAV by default; Sume TTS returns MP3
Gemini 3.8 TTS unary requests return WAV with a RIFF header. Sume TTS 1.0 defaults to mp3 at 44100 Hz, so ask for wav when your pipeline needs it.
- Gemini TTS reads stage directions aloud: Sume emotion field
Gemini 3.8 TTS speaks its text field verbatim, so style goes in speech_metadata. On Sume, keep the transcript clean and put delivery in generation_config.
- Gemini API file size limit: 2GB per file, 20GB storage
Gemini's rate-limits page lists a 2GB input file limit and 20GB file storage under Batch API limits. Sume takes public HTTPS URLs instead of uploaded files.
- Gemini static vs dynamic webhooks: Sume sets one URL per job
Gemini splits project-level static webhooks from per-request dynamic ones. Sume has only the per-job kind: pass webhook_url on each submit call.
Written by Sume