AEON-7

NVFP4 quantizations · Abliterated LLMs · DGX Spark deployments · AI media production toolchain

I build deployment-ready open releases for next-gen GPUs — NVFP4-quantized abliterated LLMs (Gemma 4, Qwen 3.6, Nemotron 3) running on NVIDIA DGX Spark (GB10 / Blackwell / sm_121a), DFlash speculative decoding, EAGLE drafters, and a complete agent-driven AI media production toolchain.

Everything below is public, MIT/Apache-licensed, and reproducible — Docker stacks, pre-built vLLM images, deployment guides, and benchmark numbers included.

🎬 AEON Media Production

Open-source AI-driven media production toolchain. Five focused repositories — each generating one kind of media (music, radio drama, music video, cinematic video, or the base ComfyUI stack), all designed for AI agents through skill MD files and CLI scripts. No node-graph wrangling.

Repo	What it does	★
comfyui-aeon-spark	Bleeding-edge ComfyUI for DGX Spark (CUDA 13 + SageAttention v3 + NVFP4 + 14 custom-node packs + Flux 2 Dev / LTX 2.3 22B / ACE-Step v1.5 XL Turbo pre-bundled). Foundation for every other repo in this section.
aeon-music-maker	ACE Step 1.5 XL music generation with dynamics-preserving mastering chain (HPF → EQ → tape sat → LUFS gain-match → true-peak ceiling). FLAC-lossless output, auto-detected mastering presets, CLI-driven.
aeon-radio-drama	Full-pipeline radio drama / audiobook production — dialogue (Qwen3-TTS) + music (ACE Step) + SFX (MMAudio / Stable Audio Open / ACE) + sidechain mix in one command. Three-Lock voice persistence. Bundles standalone `music_maker.py` + `sfx_maker.py` for one-shot music or SFX generation.
aeon-music-video	Audio-reactive music video builder. librosa-driven beat / onset / RMS / spectral-centroid detection drives ffmpeg filter chains for synced visual effects. CPU-only — no GPU, no ComfyUI, no model downloads required.
aeon-movie-maker	Fast cinematic video via LTX 2.3 22B. Single clips, full screenplays with character continuity (last-frame carry-forward + per-character seed offsets), and sidechain-mixed final cuts. CLI-tunable LoRA strengths + saturation troubleshooting guide.

What every AEON Media Production repo ships with

File	What it is
`README.md`	Quick start, configuration table, local-vs-remote ComfyUI execution modes, env-var reference, model-installation paths
`AGENTS.md`	Step-0 execution-mode detection guide for AI agents, invocation contract, recovery patterns
`SKILL.md`	Full prompt-engineering recipes, troubleshooting decision trees, the canonical agent skill definition
`ATTRIBUTION.md`	Upstream credits — every model, library, and custom node properly attributed
`.env.example`	Verbose, self-documenting — every variable has inline instructions on where to get values (HF tokens, Civitai tokens, ComfyUI URL patterns)
`setup.sh`	First-time install — validates ComfyUI reachability, installs Python deps, inventories model files, prints download commands for missing pieces
`sync.sh`	Incremental update — diff preview, auto-stash local edits, ff-only pull, refresh deps, re-run model check. Supports `--dry-run` / `--yes` / `--no-models`
`.gitignore`	Standard — never commits `output/`, `models/`, `.env`, `__pycache__`, etc.
`LICENSE`	MIT

Lifecycle (every repo, same pattern)

git clone https://github.com/AEON-7/<repo>     →   ./setup.sh        →   start using
                                                       │
                                                       ▼
                                              copy .env.example → .env
                                              edit COMFYUI_URL etc.
                                                       │
                                                       ▼
                                          python scripts/<tool>.py ...
                                                       │
                            (later, when upstream updates) ▼
                                                ./sync.sh
                                              (preview → confirm → pull → refresh)

Local vs remote ComfyUI

Every tool that uses ComfyUI supports two execution modes, documented per-repo:

Mode A — Local: CLI runs on the same machine as ComfyUI. Just python scripts/<tool>.py ....
Mode B — Remote: ComfyUI on a GPU box (DGX Spark, headless server). Either invoke the CLI over SSH (ssh user@gpu-host 'cd repo && python ...') or hit the remote ComfyUI HTTP API directly via SSH tunnel or --listen 0.0.0.0.

aeon-movie-maker has additional constraints documented (I2V + screenplay carry-forward needs filesystem-level access — pure HTTP-only remote works for T2V single clips only).

💎 Gemma 4 Models — NVFP4 quantizations for DGX Spark

Abliterated Gemma 4 deployments at NVFP4 precision (4-bit weights + 8-bit activations) for NVIDIA DGX Spark / Blackwell GPUs. Includes EAGLE speculative-decoding drafters for both base models.

Repo	Model	Architecture	Description
Gemma-4-31B-DECKARD-HERETIC-Uncensored-NVFP4	Gemma 4 31B DECKARD HERETIC	Dense, thinking	NVFP4-quantized abliterated 31B dense reasoning model. AWQ_FULL + SVDQuant variants.
Gemma-4-26B-A4B-it-Uncensored-NVFP4	Gemma 4 26B A4B-it	MoE	NVFP4-quantized 26B MoE. 50 tok/s single, 1430 tok/s aggregate on DGX Spark.
supergemma4-26b-abliterated-multimodal-nvfp4	SuperGemma 4 26B Abliterated Multimodal	MoE, multimodal	NVFP4 AWQ Full quantization for Blackwell GPUs. Pre-built vLLM container + patches included.
Gemma-4-E4B-it-Uncensored-NVFP4	EAGLE drafter for 26B MoE	Speculative decoding	EAGLE E4B speculative-decoding drafter for the Gemma 4 26B MoE. NVFP4 AWQ.
Gemma-4-E4B-DECKARD-HERETIC-Uncensored-NVFP4	EAGLE drafter for 31B DECKARD	Speculative decoding	EAGLE E4B speculative-decoding drafter for the 31B DECKARD HERETIC.

🐉 Qwen 3.5 Models

Reserved for upcoming Qwen 3.5 abliterated NVFP4 deployments. No releases yet — coming soon.

🐉 Qwen 3.6 Models — NVFP4 + DFlash on DGX Spark

Lossless abliteration of Qwen 3.6 with hardware NVFP4 quantization, optionally combined with DFlash speculative decoding for higher single-stream throughput.

Repo	Model	Architecture	Description	★
Qwen3.6-27B-AEON-Ultimate-Uncensored-DFlash	Qwen 3.6 27B AEON Ultimate Uncensored	Dense	Lossless abliteration with NVFP4 hardware quantization. BF16 (51 GB) + NVFP4 (26 GB) deployment guide, docker-compose, QuickStart.
Qwen3.6-NVFP4-DFlash	Qwen 3.6 35B-A3B-heretic	MoE	NVFP4 + DFlash speculative decoding on DGX Spark (GB10 / sm_121a). Source-built vLLM image + 7 patches + comprehensive deployment guide.

🌌 Nemotron Models

NVIDIA Nemotron deployments for Blackwell-class hardware.

Repo	Model	Architecture	Description	★
Nemotron-3-Nano-Omni-AEON-Ultimate-Uncensored	Nemotron 3 Nano Omni	12-D abliterated multimodal	BF16 + NVFP4 multimodal reasoning model for DGX Spark / Blackwell. Source-built vLLM v0.20.0 image + 4 patches + benchmark + deployment guide.

🛠️ Inference & Optimization Tools

Lower-level building blocks: speculative decoding, KV-cache compression, quantization tooling, and vLLM image builds.

Repo	What it does	★
vllm-dflash	DFlash vLLM for DGX Spark — Plug & Play Block-Diffusion Speculative Decoding. Pre-built Docker image with NVFP4, sm_121a kernels, and Qwen-targeted optimizations.
turboquant	Near-optimal KV cache quantization for LLM inference (3-bit keys, 2-bit values) with Triton kernels + vLLM integration.
Model-Optimizer	Unified library of SOTA model-optimization techniques — quantization, pruning, distillation, speculative decoding — for TensorRT-LLM / TensorRT / vLLM deployment.
modelopt-fast-moe	MoE-targeted quantization + AWQ calibration tooling. NVFP4 routing, expert-aware modelopt.

🧪 Apps & Utilities

Side projects, tools, and infrastructure that aren't model deployments but might be useful.

Repo	What it does	★
matrix-voip-agent	Headless Matrix WebRTC voice agent — auto-answers VoIP calls and bridges audio to any AI agent via PipeWire.
cosmic-mind	Security-and-resiliency-focused deployment of the Quartz web app. A place to build your second mind and share it.
regex-builder	Simple and elegant RegEx builder.
quartz	Fast batteries-included static-site generator that transforms Markdown into fully functional websites.

📊 Stats

☕ Support the work

If any of these releases have been useful to you, tips are deeply appreciated — they go directly toward more compute, more models, and more open releases. Scan a QR with your wallet, or click any address below to copy.

₿ Bitcoin (BTC) _{bc1q09xmzn00q4z3c5raene0f3pzn9d9pvawfm0py4}	Ξ Ethereum (ETH) _{0x1512667F6D61454ad531d2E45C0a5d1fd82D0500}
◎ Solana (SOL) _{DgQsjHdAnT5PNLQTNpJdpLS3tYGpVcsHQCkpoiAKsw8t}	ⓜ Monero (XMR) _{836XrSKw4R76vNi3QPJ5Fa9ugcyvE2cWmKSPv3AhpTNNKvqP8v5ba9JRL4Vh7UnFNjDz3E2GXZDVVenu3rkZaNdUFhjAvgd}

Ethereum L2s (Base, Arbitrum, Optimism, Polygon, etc.) and EVM-compatible tokens can be sent to the same Ethereum address.

🤝 Get in touch

🌐 Open an issue on any repo for questions, bug reports, or feature requests
📜 Most releases include a deployment guide + benchmark numbers — start there

Built for the open source community on NVIDIA DGX Spark, RTX 5090, and Blackwell-class GPUs.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

AEON-7

Achievements

Achievements

Block or report AEON-7

AEON-7

🎬 AEON Media Production

What every AEON Media Production repo ships with

Lifecycle (every repo, same pattern)

Local vs remote ComfyUI

💎 Gemma 4 Models — NVFP4 quantizations for DGX Spark

🐉 Qwen 3.5 Models

🐉 Qwen 3.6 Models — NVFP4 + DFlash on DGX Spark

🌌 Nemotron Models

🛠️ Inference & Optimization Tools

🧪 Apps & Utilities

📊 Stats

☕ Support the work

🤝 Get in touch

Popular repositories Loading

Uh oh!