惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
罗磊的独立博客
博客园 - 【当耐特】
M
MIT News - Artificial intelligence
月光博客
月光博客
博客园_首页
博客园 - 叶小钗
T
Tailwind CSS Blog
H
Hackread – Cybersecurity News, Data Breaches, AI and More
I
InfoQ
量子位
小众软件
小众软件
爱范儿
爱范儿
The GitHub Blog
The GitHub Blog
IT之家
IT之家
Jina AI
Jina AI
阮一峰的网络日志
阮一峰的网络日志
G
Google Developers Blog
WordPress大学
WordPress大学
人人都是产品经理
人人都是产品经理
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
J
Java Code Geeks
云风的 BLOG
云风的 BLOG
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
GitHub - freeaigit/free-scene: Multi-scene AI movie maker...
nadermx · 2026-05-14 · via Hacker News - Newest: "AI"

Multi-scene AI movie maker. Give it a story idea, get a stitched mp4.

$ freescene "A retired astronaut tends a rooftop garden in Tokyo, she finds a small alien plant. Tone: quiet, hopeful, Ghibli-esque."

─────────────────────────────── freescene ────────────────────────────────
Idea: A retired astronaut tends a rooftop garden in Tokyo, she finds a small alien plant. Tone: quiet, hopeful, Ghibli-esque.
Scenes: 3  Style: cinematic, warm, realistic  Duration/scene: 4s
LLM: qwen7b  Video: cogvideox  Backend: https://api.free.ai

⠼ Rendering scene 2/3…  ━━━━━━━━━━━━━━━━━━━━━━━━━━━╸━━━━━━━━━━ 4/5 0:01:14

──────────────────────────────── Done ────────────────────────────────────
Output: /tmp/freescene_abc123def456/abc123def456.mp4
Job ID: abc123def456  Elapsed: 167.4s

How it works

  1. Script — an LLM acts as film director and writes N scene prompts from your one-line idea.
  2. Render — each prompt is rendered as a short video clip via an OpenAI-compatible video API.
  3. Concatffmpeg stitches the clips into one mp4. Stream-copy first; falls back to libx264 re-encode if codecs/containers don't match.

Same pipeline that powers free.ai/video/movie/, extracted into a standalone, self-installable CLI.

Install

pip install freeai-scene

You also need ffmpeg on PATH:

# macOS
brew install ffmpeg

# Debian/Ubuntu
sudo apt-get install ffmpeg

# Windows
winget install Gyan.FFmpeg

Cost upfront — be honest

What you're running Tokens Anon (2,500/day) Registered (30,000/day) $5 pack (200K)
Single 2-second scene (cheapest) 5,100 🚫 ✅ 5/day ~40 scenes
Default 3-scene 3-second movie 22,600 🚫 1 full movie/day ~9 movies
3-scene 4-second movie 30,100 🚫 🚫 (100 over) ~6 movies
5-scene 6-second movie 75,100 🚫 🚫 ~2 movies

Per-scene cost is max(5,000, duration × 2,500). The CLI prints a cost estimate + affordability hint before each run; --dry-run shows it without spending.

Free tier sweet spot: a free.ai signup gets you 30,000 tokens/day — exactly enough for one default 3-scene 4-second movie every 24 hours, forever. For more, set FREE_API_KEY to your account credits, grab a $5 pack (200K tokens, ~6 movies), or run against your own GPU (see supported backends) for $0.

Quick start

No setup required — try a dry run first to see the cost:

freescene "Cyberpunk Tokyo at dawn, a lone hacker on a rooftop" --scenes 5 --style scifi --dry-run

Render it for real (needs an API key or your own backend — see above):

freescene "Cyberpunk Tokyo at dawn, a lone hacker on a rooftop" --scenes 5 --style scifi

For your full account quota:

export FREE_API_KEY=fk_live_...
freescene "Pixar squirrel learns to surf" -n 4 -s 3d -d 6

Bring your own backend (any OpenAI-compatible LLM + video API):

export FREE_API_URL=http://localhost:8080
export FREE_API_KEY=$YOUR_TOKEN
export LLM_MODEL=qwen3-30b
export VIDEO_MODEL=hunyuan-video
freescene "A drone flight over Mars colonies"

Mix and match — script-writing LLM on OpenAI, video render on a local vLLM:

export LLM_API_URL=https://api.openai.com
export LLM_MODEL=gpt-5
export VIDEO_API_URL=http://localhost:7860
export VIDEO_MODEL=cogvideox
freescene "Submarine exploring an alien reef"

CLI reference

freescene IDEA [OPTIONS]

  -n, --scenes INTEGER             Number of scene clips (1-12, default 3)
  -s, --style TEXT                 Style preset or free-form descriptor
                                   Presets: cinematic, anime, documentary,
                                   noir, 3d, vintage, scifi, fantasy
  -d, --duration INTEGER           Seconds per scene (2-12, default 4)
  -m, --video-model TEXT           Per-scene render model
      --llm-model TEXT             Script-writing model
      --api-url TEXT               OpenAI-compatible API root
      --api-key TEXT               Bearer token
  -o, --output FILE                Output .mp4 path
      --show-script                Print scene prompts before rendering
      --quiet                      Suppress progress UI (prints output path only)
      --version                    Show version
  -h, --help                       Show help

Python API

from freescene import make_movie, Config

cfg = Config(
    api_url="https://api.free.ai",
    api_key="fk_live_...",
    video_model="cogvideox",
)

result = make_movie(
    idea="A retired astronaut tends a rooftop garden in Tokyo...",
    num_scenes=3,
    style="ghibli-inspired, soft",
    duration_per_scene=4,
    config=cfg,
)

print(result.output_path)   # PosixPath('/tmp/.../abc123.mp4')
print(result.scenes)        # ['wide shot of ...', 'medium shot ...', 'close-up ...']
print(result.elapsed_s)     # 167.4

Async variant for when you're already inside an event loop:

from freescene.pipeline import make_movie_async

result = await make_movie_async(idea="...", num_scenes=3)

Environment variables

Var Default Purpose
FREE_API_URL https://api.free.ai OpenAI-compatible API root
FREE_API_KEY (none) Bearer token
LLM_API_URL inherits FREE_API_URL Override for /v1/chat/completions only
LLM_MODEL qwen7b Script-writing model id
VIDEO_API_URL inherits FREE_API_URL Override for /v1/video/generate/ only
VIDEO_MODEL cogvideox Per-scene render model

Supported backends

Anything that speaks OpenAI's chat-completions shape for scripts and a POST /v1/video/generate/ returning {video_url|output_url|url} for renders. Verified backends:

  • Free.ai (default) — cogvideox, hunyuan-video, wan-i2v for video; qwen7b, qwen3-30b, GPT, Claude, Gemini etc. for the script.
  • Your own vLLM + diffusers shim — set FREE_API_URL to your gateway.
  • OpenAI for scripts — set LLM_API_URL=https://api.openai.com, LLM_MODEL=gpt-5, plus a separate VIDEO_API_URL for render.

If your video API returns a different JSON key, file an issue with a sample response — happy to widen the parser.

Development

git clone https://github.com/freeaigit/free-scene.git
cd free-scene
pip install -e ".[dev]"
pytest

Tests cover the LLM-output parser (mocked) and the ffmpeg-concat path (real ffmpeg, auto-skipped if not installed).

License

MIT. See LICENSE.

Credits

Extracted from Free.ai's /video/movie/ pipeline. Use the hosted version at free.ai/video/movie/ if you don't want to manage your own GPU.