Gemini 3.8 Live: talk through an idea, get a structured doc back, and what that means for voice in games
From the OpenDrama Create Wiki
genevieve__h, who works in developer relations for Google AI Studio, built Clever Brainstorm as a demo app for Gemini 3.8 Live: a voice orb next to a document. She talks through how to show the model off on X, and while she talks the document fills itself in: sections, open questions, saved nuggets, an action item. Her post argues voice models are "only as useful as their ability to reason and make tool calls" mid-conversation. For creators, that's the difference between an NPC that chats and one that does things.
Members read on
The rest of this page is for members
- 4 more steps
- 4 more sections
Create a free account, top up once — from $0.99 — and the whole wiki is yours.
- Every breakdown, Skill and Prompt
- New pages every day
- Prompts copy in one tap
Tags:#Voice & dubbing#Gemini#AI NPCs#AI agents#Game design
Written by the OpenDrama editors from a public post. Stills are quoted from the original video for commentary; tools, settings and steps are our own reconstruction. All rights in the original work stay with its creator — ask us and we take a still down.
Skills and prompts in this breakdown
More in Sound
View all →AI sound design for games and short videos
How to brief AI music, SFX and ambience, build seamless game loops and stingers, sync sound to cuts and gameplay events, hit platform loudness targets such as −14 LUFS, and avoid licensing traps.
Gemini 3.8 TTS: cast a voice in one sentence, then direct game dialogue and narration with tags
genevieve__h demos Gemini 3.8 TTS: a one-sentence casting note becomes a persistent voice, and direction lines plus inline tags steer every beat. Here's how to use it for game dialogue and video narration.
Gemini 4 Argon builds a Jupiter/Moog combo synth in one HTML file, in under 30 minutes
maxescu had Gemini 4 Argon build an 8-voice retro synth in under half an hour, all in a single HTML file. We read the live source to show how it's structured and how to get a model to hand you a playable instrument.
An AI voice reads everything AI voices can't do: a 44-second spec ad for Eleven v4
Former ElevenLabs employee Burak Tuyan wrote a list of 'can't dos' and had Eleven v4 perform each one. The script trick, the shots that hide the mouth, and the twist.
Voice cloning from a Mac with no NVIDIA card: IndexTTS 2.5 on a Colab L4, near real time
nicekate drives the open-source IndexTTS 2.5 on a Colab L4 from her Mac: 34.69 seconds of narration in about 35 seconds. Her three-layer setup, reference audio, settings and compute budget.
