MoneyPrinterTurbo g4f Not Working? Fix & Free LLM Alternatives

What You're Seeing

You configured g4f in config.toml, started MPT, and tried to generate a video. One of these happened:

RetryProviderError: RetryProvider failed after 3 attempts
No provider found or individual provider errors for each one MPT tries
The backend starts, but video generation hangs at "Writing script..." forever
ImportError for curl_cffi, undetected_chromedriver, or platformdirs

g4f promised free GPT access by routing through third-party APIs. That promise is now broken.

Why g4f Is Dead

g4f (GPT4Free) works by scraping free tiers of various AI providers — reverse-engineered APIs that were never meant for programmatic access. This was always fragile. In 2026, every single one has broken:

Providers constantly go offline — free API endpoints get shut down, rate-limited, or patched
Missing dependencies — g4f needs curl_cffi, undetected_chromedriver, platformdirs that aren't in MPT's requirements.txt
No maintainer response — the g4f repo's free providers are community-maintained and frequently broken for weeks
Quality is terrible — even when it works, responses are slow, truncated, or nonsensical

Don't bother debugging g4f. Installing missing deps, switching providers, or downgrading versions might get it working for one session — then it breaks again tomorrow. The time you spend debugging g4f is better spent setting up a real provider.

Fix 1: Qwen via DashScope (Recommended)

Best option for most users. Qwen is Alibaba's LLM. DashScope (also called Bailian/百炼) gives you free quota — no credit card, no VPN needed, works direct from China.

Step 1: Get a free API key

Go to dashscope.console.aliyun.com
Sign up with Aliyun account (or register — phone number required)
Go to API Key management → Create API Key
Copy the key. Free quota: ~1 million tokens

Step 2: Configure MPT

Open config.toml and set:

llm_provider = "qwen"
qwen_api_key = "sk-your-api-key-here"
qwen_model = "qwen-plus"

Available models: qwen-turbo (fastest, cheapest), qwen-plus (balanced), qwen-max (best quality). For MPT video scripts, qwen-plus is ideal.

Why Qwen wins for MPT: Free quota resets monthly. China-direct — no VPN needed for API calls. Script quality is solid — comparable to GPT-3.5 for the kind of short-form video scripts MPT generates. And the config change is one line in config.toml.

Fix 2: Moonshot (Kimi)

Moonshot's Kimi model is another good option, especially for Chinese-language video scripts.

Register at platform.moonshot.cn
Create API key (new accounts get free credits)

llm_provider = "moonshot"
moonshot_api_key = "sk-your-key-here"
moonshot_model = "moonshot-v1-8k"

Heads-up: Moonshot free credits are limited. Some users report Error 429 — account suspended due to insufficient balance after the initial quota runs out. If you hit this, switch to Qwen — their free tier is more generous and renews monthly.

Fix 3: Any OpenAI-Compatible Provider

MPT v1.1.0 supports llm_provider = "openai" natively. This works with any service that exposes an OpenAI-compatible endpoint — not just OpenAI itself.

llm_provider = "openai"
openai_api_key = "your-key-here"
openai_api_base = "https://your-provider.com/v1"
openai_model = "gpt-4o-mini"

Services that work with this pattern:

DeepSeek — api_base = "https://api.deepseek.com", free credits available
OpenRouter — api_base = "https://openrouter.ai/api/v1", pay-as-you-go
Groq — api_base = "https://api.groq.com/openai/v1", free tier with rate limits
Any local LLM via Ollama/LM Studio — api_base = "http://localhost:11434/v1"

Which One Should You Use?

Provider	Cost	China Access	Script Quality	Best For
Qwen (DashScope)	Free (1M tokens/mo)	Direct, no VPN	Good	Most users — best default
Moonshot	Free credits (limited)	Direct, no VPN	Good (Chinese)	Chinese-language videos
DeepSeek	Free credits	Direct, no VPN	Excellent	Quality-focused users
OpenAI	~$0.01/video	VPN required	Excellent	Best quality, budget OK
g4f	Free	VPN required	Unreliable	Don't use

Bottom line: Switch to Qwen. It takes 5 minutes to get an API key, the config change is one line, and you'll never see RetryProviderError again. The free quota is enough for hundreds of videos per month.

Related Fixes

No Web UI after install? → Switch to v1.1.0
moviepy.editor not found? → moviepy fix
Edge TTS 403 error? → edge-tts fix
Full setup walkthrough → 10-minute setup guide
Ready to use MPT? → How to make your first video

MoneyPrinterTurbo: g4f Not Working? Fix & Free Alternatives