v0.1.4

vLLM Playground v0.1.4 - vLLM-Omni Multimodal

Release Date: February 2026

This release introduces vLLM-Omni integration for multimodal generation - create images, edit photos, generate speech, and produce music/sound effects, all from text prompts. The new unified Studio UI provides a polished experience with adaptive themes and a gallery for managing your creations.

Generate images, edit photos, create speech, and produce music - all with vLLM-Omni.

---

Highlights

• vLLM-Omni Integration - Image generation, image editing, TTS, and audio generation
• Studio UI - Adaptive themes, unified gallery, built-in media players
• CLI Enhancement - vllm-playground pull --omni for vLLM-Omni container
• Pre-commit Hooks - Ruff formatting and code quality checks

---

vLLM-Omni Integration

vLLM-Omni extends vLLM to support multimodal generation. Access it from the vLLM-Omni sidebar menu.

Model Types

Select from Image, TTS, or Audio generation modes.

Image Generation (Text-to-Image)

Generate images from text prompts using DiT (Diffusion Transformer) models:

Model	VRAM	Description
Tongyi-MAI/Z-Image-Turbo	16GB	Fast generation
Qwen/Qwen-Image	>48GB	High quality
Qwen/Qwen-Image-2512	>48GB	Higher resolution (2512px)
Qwen/Qwen-Image-Layered	>48GB	Layered image generation
stabilityai/stable-diffusion-3-medium	16GB	Stable Diffusion 3

Features:

• Prompt templates (Landscape, Portrait, Art/Abstract)
• Configurable image size, inference steps, guidance scale
• Seed control for reproducibility

Image Editing (Image-to-Image)

Upload an existing image and transform it with text prompts:

Model	VRAM	Description
Qwen/Qwen-Image-Edit	>48GB	Image editing
Qwen/Qwen-Image-Edit-2509	>48GB	Advanced editing
Qwen/Qwen-Image-Edit-2511	>48GB	Latest editing
meituan-longcat/LongCat-Image-Edit	>48GB	Alternative editing model

Features:

• Drag-and-drop image upload
• Transform photos with natural language descriptions
• Edit-capable models automatically show upload zone

TTS (Text-to-Speech)

Generate natural-sounding speech from text:

Model	VRAM	Description
Qwen/Qwen3-TTS-12Hz-0.6B-Base	16GB	Lightweight TTS
Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice	24GB	Custom voice cloning
Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign	24GB	Voice design capabilities

Features:

• Prompt templates for common TTS scenarios
• Built-in audio player for playback
• Download generated audio files

Audio Generation (Music/SFX)

Generate music and sound effects using Stable Audio:

Model	VRAM	Description
stabilityai/stable-audio-open-1.0	40GB+	Up to 47s stereo at 44.1kHz

Features:

• Duration control (1-47 seconds)
• Prompt templates for music genres and sound effects
• In-process mode for Stable Audio (bypasses serving limitations)

Preview: Video & Omni Chat (Coming Soon)

The UI includes placeholders for upcoming features:

• Video Generation - Text-to-video with Wan2.2 models
• Omni Chat - Multimodal chat with text and audio I/O (Qwen3-Omni)

These features are disabled in v0.1.4 and will be enabled in a future release.

---

Studio UI

Adaptive Themes

Each generation mode has its own color theme for visual distinction:

• Image - Default studio theme
• Image-Edit - Edit mode indicator
• TTS - Speech synthesis theme
• Audio - Music/SFX generation theme

Unified Gallery

All generated content appears in a unified gallery:

• Display - Lightbox viewer for images
• Download - Save any generated file
• Delete - Clear individual items or entire gallery

Built-in Media Players

Native playback for generated audio and video:

• Audio player with playback controls
• Video player (for future video generation)

---

Run Modes

Subprocess Mode

Install vLLM-Omni locally in a virtual environment:

# Create venv with Python 3.12
uv venv --python 3.12 --seed ~/.venv-vllm-omni
source ~/.venv-vllm-omni/bin/activate

# Install vLLM base (CUDA)
uv pip install vllm==0.14.0 --torch-backend=auto

# Clone and install vLLM-Omni
git clone https://github.com/vllm-project/vllm-omni.git
cd vllm-omni && uv pip install -e .

# For Qwen3-TTS models, also install:
uv pip install onnxruntime

Configure in the UI:

• Run Mode: Subprocess
• Venv Path: ~/.venv-vllm-omni

Container Mode

Use official vLLM-Omni container images:

Platform	Image
NVIDIA (CUDA)	docker.io/vllm/vllm-omni:v0.14.0rc1
AMD (ROCm)	docker.io/vllm/vllm-omni-rocm:v0.14.0rc1

Pre-download with:

vllm-playground pull --omni

---

Recipes

New vLLM-Omni Audio Models recipe category:

Recipe	Model	Description
Stable Audio Open 1.0	stabilityai/stable-audio-open-1.0	Music/SFX generation
Qwen3-TTS Custom Voice	Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice	Voice cloning TTS
Qwen3-TTS Voice Design	Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign	Voice design TTS
Qwen3-TTS Base	Qwen/Qwen3-TTS-12Hz-0.6B-Base	Lightweight TTS

Access via the Browse Recipes button in the vLLM-Omni configuration panel.

---

Developer Experience

Pre-commit Hooks

New .pre-commit-config.yaml for code quality:

repos:
  - repo: https://github.com/pre-commit/pre-commit-hooks
    rev: v5.0.0
    hooks:
      - id: trailing-whitespace
      - id: end-of-file-fixer
      - id: check-added-large-files
      - id: check-yaml

  - repo: https://github.com/astral-sh/ruff-pre-commit
    rev: v0.14.11
    hooks:
      - id: ruff-format

Install with:

pip install pre-commit
pre-commit install

---

API Endpoints

New vLLM-Omni API endpoints:

Endpoint	Method	Description
/api/omni/start	POST	Start vLLM-Omni server
/api/omni/stop	POST	Stop vLLM-Omni server
/api/omni/status	GET	Get server status
/api/omni/health	GET	Health check
/api/omni/generate	POST	Generate image
/api/omni/generate-tts	POST	Generate speech
/api/omni/generate-audio	POST	Generate music/SFX
/api/omni/models	GET	List supported models

---

Installation

Release Candidate (Testing)

pip install --pre vllm-playground

Stable Release

pip install vllm-playground

Pre-download vLLM-Omni Container

vllm-playground pull --omni

---

Documentation

• vLLM-Omni Guide - Full setup and usage guide
• Changelog - Version history

---

Acknowledgments

• The vLLM Project team
• The vLLM-Omni contributors
• The Stable Audio team at Stability AI
• The Qwen team at Alibaba
• Community contributors and testers

---

Links

• GitHub: https://github.com/micytao/vllm-playground
• PyPI: https://pypi.org/project/vllm-playground/
• vLLM-Omni: https://github.com/vllm-project/vllm-omni
• vLLM Docs: https://docs.vllm.ai

---

Made with ❤️ for the vLLM community