[docs]: highlight Dreamverse deployment paths + add Server B200 (SSH) guide

Prep for the Dreamverse release by making the deploy story discoverable:

- apps/dreamverse/README.md: add a top-level Deployment overview table
  (Local GPU / Server B200 SSH / Docker / Modal) and a new "Server B200
  deployment (SSH)" section with native (source install) and Docker options.
  Commands mirror the existing launch / docker / ffmpeg scripts; adds a
  first-boot warmup caveat.
- README.md: add a Dreamverse highlight section and a Key Features bullet
  linking to the app, the live demo, and the deployment guide.

Author: alexzms

README.md

@@ -42,6 +42,8 @@ FastVideo has the following features:
   - Support H100, A100, 4090
   - Support Linux, Windows, MacOS
   - See this [page](https://hao-ai-lab.github.io/FastVideo/inference/support_matrix/) for full list of supported models, hardware assumptions, and optimization compatibility.
+- Realtime video generation & editing
+  - [Dreamverse](apps/dreamverse/README.md): stream and "vibe direct" video in realtime ([live demo](https://dreamverse.fastvideo.org/)), deployable on local GPU, a self-hosted B200 server, Docker, or serverless Modal
 
 ## Getting Started
 
@@ -69,6 +71,19 @@ See below for recipes and datasets:
 | [FastWan2.1-T2V-1.3B](https://huggingface.co/FastVideo/FastWan2.1-T2V-1.3B-Diffusers) | [Recipe](https://github.com/hao-ai-lab/FastVideo/tree/main/examples/distill/Wan2.1-T2V/Wan-Syn-Data-480P)       | [FastVideo Synthetic Wan2.1 480P](https://huggingface.co/datasets/FastVideo/Wan-Syn_77x448x832_600k)     |
 | [FastWan2.2-TI2V-5B](https://huggingface.co/FastVideo/FastWan2.2-TI2V-5B-Diffusers)   | [Recipe](https://github.com/hao-ai-lab/FastVideo/tree/main/examples/distill/Wan2.2-TI2V-5B-Diffusers/Data-free) | [FastVideo Synthetic Wan2.2 720P](https://huggingface.co/datasets/FastVideo/Wan2.2-Syn-121x704x1280_32k) |
 
+## Dreamverse — Realtime Video Generation & Editing
+
+[Dreamverse](apps/dreamverse/README.md) is FastVideo's realtime video generation
+and editing platform — "vibe directing" a video as it streams. It lives in the
+monorepo under [`apps/dreamverse/`](apps/dreamverse/) and ships its own backend
+(`dreamverse-server`) plus a web UI.
+
+Try the [live demo](https://dreamverse.fastvideo.org/), read the
+[blog](https://haoailab.com/blogs/dreamverse/), or run it yourself. Dreamverse
+deploys on a local GPU, a self-hosted B200 server over SSH, Docker, or
+serverless [Modal](apps/dreamverse/scripts/modal/README.md) — see the
+[Dreamverse deployment guide](apps/dreamverse/README.md#deployment).
+
 ## Inference
 
 ### Generating Your First Video

apps/dreamverse/README.md

@@ -2,6 +2,21 @@
 
 Dreamverse is the FastVideo realtime video generation & editing platform. It lives in this monorepo under `apps/dreamverse/`.
 
+## Deployment
+
+Dreamverse runs anywhere you have a CUDA NVIDIA GPU. Pick the path that fits:
+
+| Method | Best for | Guide |
+| --- | --- | --- |
+| **Local GPU (native)** | Dev boxes, quick local demos | [Quick Start: Local GPU](#quick-start-local-gpu) |
+| **Server / bare-metal B200 (SSH)** | Persistent self-hosted serving | [Server B200 deployment (SSH)](#server-b200-deployment-ssh) |
+| **Docker** | Reproducible container build/run | [`docker/README.md`](docker/README.md) |
+| **Modal (serverless B200)** | On-demand cloud GPU, no infra to manage | [`scripts/modal/README.md`](scripts/modal/README.md) |
+
+Whatever the target, the backend needs the prompt-LLM keys (`CEREBRAS_API_KEY`
+and `GROQ_API_KEY`) and serves on port `8009`: `/healthz` reports liveness and
+`/readyz` returns `200` once model load and startup warmup finish.
+
 ## Install Dreamverse
 
 You can install Dreamverse using one of the methods below.
@@ -163,6 +178,83 @@ BACKEND_HOST=localhost BACKEND_PORT=8009 npm run dev
 
 Open `http://localhost:5299`.
 
+## Server B200 deployment (SSH)
+
+For a persistent, self-hosted deployment, SSH into a CUDA-capable GPU host (for
+example a B200 box) and bring Dreamverse up there. Both paths below assume the
+host has recent NVIDIA drivers and CUDA-visible GPUs.
+
+### Option A: Native (source install)
+
+Best when you want to run and iterate on the checkout directly.
+
+```bash
+ssh <user>@<b200-host>
+
+# 1. Clone (or pull) FastVideo and install the Dreamverse extra
+git clone https://github.com/hao-ai-lab/FastVideo.git
+cd FastVideo
+pip install --upgrade pip && pip install uv
+uv venv .venv --python 3.12 && source .venv/bin/activate
+uv pip install -e ".[dreamverse]"
+
+# 2. (Recommended) build the native FFmpeg for full streaming performance
+bash apps/dreamverse/scripts/install_native_ffmpeg.sh
+source apps/dreamverse/scripts/ffmpeg-env.sh
+
+# 3. Export the prompt-LLM API keys (direct launches do not source ~/.env)
+export CEREBRAS_API_KEY=...   # default prompt provider
+export GROQ_API_KEY=...       # fallback provider
+
+# 4. Start the backend; binds 0.0.0.0:8009 and starts one GPU worker
+dreamverse-server --host 0.0.0.0 --port 8009
+```
+
+Serve the web UI from the same host (or point an external frontend at the
+backend):
+
+```bash
+cd apps/dreamverse/web && npm ci
+BACKEND_HOST=<b200-host> BACKEND_PORT=8009 npm run dev
+```
+
+Keep the backend alive across SSH sessions with your process manager of choice
+(`tmux`, `systemd`, `nohup`, ...), and wait for warmup before serving traffic:
+
+```bash
+curl http://<b200-host>:8009/healthz   # process is up
+curl http://<b200-host>:8009/readyz    # 200 once model load + warmup finish
+```
+
+### Option B: Docker (on the server)
+
+Best for reproducible, isolated runs. Builds the image (which bundles the native
+FFmpeg build) and runs it with the GPUs exposed.
+
+```bash
+ssh <user>@<b200-host>
+git clone https://github.com/hao-ai-lab/FastVideo.git
+cd FastVideo
+
+# Build the image (defaults to dreamverse:dev; add BUILD_DREAMVERSE_UI=1 for the UI)
+apps/dreamverse/docker/docker_build.sh
+
+# Run it: passes the keys, mounts the HF cache + outputs, maps host port 8009
+CEREBRAS_API_KEY="<your-key>" \
+GROQ_API_KEY="<your-key>" \
+apps/dreamverse/docker/docker_run.sh
+```
+
+Pin a specific GPU with `DREAMVERSE_DOCKER_GPUS=device=<idx> FASTVIDEO_GPU_COUNT=1`.
+See [`docker/README.md`](docker/README.md) for build args, the UI image, GPU
+pinning, and the smoke script.
+
+> **First-boot warmup:** with `torch.compile` + startup warmup enabled, the
+> first boot compiles the segment 1 and segment 2 inference paths and can take
+> tens of minutes on a cold cache; `/readyz` stays `503` until it finishes. For
+> a faster (uncompiled) startup while testing, set
+> `FASTVIDEO_ENABLE_STARTUP_WARMUP=0` before starting the backend.
+
 ## Quick Start: Mock Backend (For UI development)
 
 The mock server emulates the Dreamverse backend protocol and streams a