Scenema AI

Turn your ideas into cinema-quality films. AI video, voice acting, & sound on one platform. Visit:

[Scenema AI @scenemaai]

Vox-style animation using Gemini Omni Flash on Scenema. A six-minute AI explainer on FIFA's 2026 World Cup finances.

rt + comment and we'll send you the full prompt.

[Scenema AI @scenemaai]

The narration is one consistent voice from start to finish. The visual language, paper collage on aged newsprint with hot red accents, holds through every single shot.
Try it at https://t.co/glyemSfmyY

[Scenema AI @scenemaai]

Vox-style animation using Gemini Omni Flash on Scenema. A six-minute AI explainer on FIFA's 2026 World Cup finances.

rt + comment and we'll send you the full prompt.

[Scenema AI @scenemaai]

The same recurring worker cutout appears in roughly 60 shots and stays the same character every time he shows up. The stadium, the Aramco logo, and the category 1 ticket also keep their identity across every appearance.

[Scenema AI @scenemaai]

Scenema Audio is now on ComfyUI. You can install through ComfyUI Manager, load the pre-wired workflow, and generate performance quality speech.

https://t.co/ppOanmdhpu

[Beza @BezaBerhan]

1. Starting to experiment with storyboard-based video generation at @scenemaia and I have to share what’s actually going on under the hood.

[Scenema AI @scenemaai]

That’s was really the most fundamental understanding we had to arrive at at Scenema. Text to video is DOA because unless you’re generating one-off, 15s clips, they’re not usable. But for anything that requires visual and vocal consistency, model separation is key.

[Scenema AI @scenemaai]

Experimenting with storyboarding in Scenema. The keyframes were generated using Nano Banana 2 and the shots using Seedance.

[Scenema AI @scenemaai]

Keep your character consistent across every output with https://t.co/nRKttcpdoX.

Choose from 10+ built-in styles, from photorealistic to anime, or create custom styles from reference images.

[Scenema AI @scenemaai]

Thank you for reviewing Scenema Audio, Fahd. Truly appreciate it.

[Fahd Mirza @fahdmirza]

🎙️ Scenema Audio is UNLIKE any TTS you have seen before

🎭 and it actually PERFORMS, not just pronounces

🔹 Built on LTX 2.3's 22B audiovisual model trained on real film audio
🔹 Emotional arcs that shift mid-generation using action tags
🔹 Zero-shot voice cloning, any emotion even if never recorded that way
🔹 10 languages demoed: Arabic, French, Spanish, German, Hindi, Thai, Portuguese, Polish, Swedish, Chinese
🔹 Runs locally on 16GB+ VRAM, full precision on 48GB
🔹 Free, open source, Docker deploy in one command

🔥 Watch the full video below 👇

https://t.co/hTvZx60wmO

[Scenema AI @scenemaai]

Thank you for the shoutout!

[ToZx @Tozxart]

DreamX + MoCam + Scenema — The video AI trio

3 quiet but powerful drops:

DreamX — open-world video generation from a single image
MoCam — motion-aware camera AI for better video capture
Scenema — AI that generates full scene audio from visuals

Video AI is shipping faster than anyone can keep up.

[Sheldon @Sheldon36355913]

Scenema Audio，把 TTS 往“表演型声音模型”推了一步

Scenema Audio 是一个基于 LTX 2.3 音频扩散 Transformer 的 expressive speech / voice cloning 模型。它想解决的不是普通 TTS 的“读得清楚”，而是“能不能像演员一样说”：愤怒、低语、哭、笑、疲惫、恐惧，甚至在同一段生成里做情绪转折。

用法也更像写剧本：prompt 里可以用 <speak> 描述声音、性别、语言、场景，再用 <action> 写舞台指令，比如“声音压低”“忍住情绪”“突然爆发”。模型会把这些表演指令转成语速、停顿、呼吸、情绪弧线和环境音。

关键点：zero-shot voice cloning、13 种语言、48kHz stereo WAV、speech + 环境音；INT8 配置下可在 16GB VRAM 跑，更舒服是 24GB/48GB。模型权重约 29GB，Gemma 3 12B 做 text encoder，需要 HF token 和 Google terms access。

传统 TTS 是“把文本念出来”，Scenema Audio 更像“把一段戏演出来”。适合短剧、游戏 NPC、交互叙事、AI 角色播客

地址：

https://t.co/Q7wLFuddcw

https://t.co/2Bl8PThmaI

[Scenema AI @scenemaai]

@Iancu_ai Would love for you to check us out. Scenema Audio (Zero-Shot Expressive Voice Cloning and Speech Generation) is open source. Whereas you could use both to prompt and direct on our platform https://t.co/nRKttcpdoX
Let us know what you think!

[Scenema AI @scenemaai]

@jennykrakovsky We would love to support you on this exciting journey. Would love for you to check out https://t.co/hrEpdQf6Or. Please DM us or join our discord for further assistance.

[Scenema AI @scenemaai]

@The_Kremlinn @hckinz @ltx_model @resembleai Hey, we plan on releasing a native ComfyUI workflow next week. Follow for updates!

[ModelScope @ModelScope2022]

A new audio model is live on ModelScope! Meet Scenema Audio, a 13B expressive speech generation model. Zero-shot voice cloning, emotional acting, scene-aware audio. 🤖 https://t.co/xZmg8ssVfm

Built on an audio diffusion transformer extracted from LTX 2.3's audiovisual model. It learned how people actually sound in real scenes: angry, laughing, whispering, crying, exhausted, terrified.

🎭 Action tags direct emotional shifts mid-generation. Voice can go from calm to breaking down within a single output.
🧒 Native child voices: six-year-olds, toddlers, teenagers — not pitch-shifted adults.
🌧️ Scene-aware audio: describe the environment and get rain, thunder, crowds alongside the voice.
🎙️ Zero-shot voice cloning: 10-20s reference audio, no fine-tuning needed. Any voice, any emotion.
🌍 13 languages. Long-form narration with automatic splitting and voice continuity.

[Scenema AI @scenemaai]

Scenema Audio, our open-source speech generation model built on the LTX 2.3 audio diffusion transformer.
https://t.co/rCXYn6mVRk

[Geek Lite @QingQ77]

零样本情感语音克隆和语音生成，用文本提示控制声音情绪、语调和节奏，16GB 显存即可运行。

https://t.co/QU2IUD28o6

Scenema Audio 用扩散 Transformer 做语音克隆和生成，底层来自 LTX 2.3 的视听模型。扔 10 秒参考音频进去就能克隆声音，XML 提示词控制情绪起伏、语速、呼吸和表演风格。多语言、环境音效同步生成、长文本自动分块都支持。最低 16GB 显存，Docker 一键启动，带 API 和 Web UI。

[Wildminder @wildmindai]

LTX 2.3 audio as standalone speech model.
Emotional TTS with Scenema Audio.
- Zero-shot expressive voice cloning, speech gen
- 8-step distilled with Gemma 3 12B text encoding
- stage directions via <action> tags
- runs at 1.5x real-time on RTX 4090
- fits in 16GB VRAM
- 13 languages, 48kHz stereo output
it also gens matching environment sounds

https://t.co/7OggjVcQfb

[AI Search @aisearchio]

New open-source speech generator, Scenema Audio

&gt; Incredible emotion control
&gt; Zero shot voice cloning
&gt; Action tags for stage direction, multilingual, long-form narration
&gt; Free &amp; local

Give it a listen!
https://t.co/wLjBvuGUKK