Open-source XTTS-v2 Local Voice-Over Order System: Making 10,000 RMB/Month on Web Novel Audiobooks with Zero Subscription Costs
Workflow: The daily workflow involves: first downloading web novel texts and character voice notes to be dubbed from platforms lik
Key Fields
FIELD STAMPS🔧 Workflow
The daily workflow involves: first downloading web novel texts and character voice notes to be dubbed from platforms like Fanqie Novel and Qimao Chinese Network; splitting the text by chapter and feeding it into the locally deployed XTTS-v2 model; pre-cloning dedicated voice tones for 3-5 core characters; batch-generating segmented audio for corresponding characters; using tools like Audacity to splice and synthesize complete chapter audio; and finally conducting manual reviews to correct pitch errors or mismatched emotional segments, delivering finished audiobook files that meet the client's requirements.
🛠 Setup Requirements
Setup requires basic Python environment configuration skills, downloading the Coqui XTTS-v2 open-source model and wrapper projects like novel2voice from GitHub, and deploying them on a local computer equipped with at least 4GB of VRAM (or renting a GPU server on platforms like Alibaba Cloud). The entire setup and debugging cycle takes about 3-5 days to run through the basic inference pipeline, with no professional algorithmic background required.
🧰 Toolchain
- 🔧 Coqui XTTS-v2
- 🔧 Python
- 🔧 GitHub
- 🔧 Audacity
- 🔧 Fanqie Novel Author Backend
💰 Revenue
① Web Novel Audiobook Outsourcing (Main Revenue): Clients pay by finished hour, with unit prices ranging from 200-500 RMB/hour. Delivering 4-8 short works of 100,000-200,000 words per month yields a monthly income of 10,000-15,000 RMB (10k-15k RMB). Once stable clients are accumulated, taking on long-form works can exceed 20,000 RMB (20k RMB). The exact share of web novel outsourcing in the total mix is unspecified (merchant claim, unverified independently as of 2026); ② Episode-based Voice-Over: Short drama producers pay per episode, with 80-150 RMB for a 10-minute episode, yielding a net profit of 3,000 RMB for a 30-episode short drama. The share of episode-based voice-overs is unspecified (merchant claim, lacks independent verification); ③ Monthly Maintenance: Clients pay a monthly maintenance fee of 200-500 RMB; binding 10 clients creates stable cash flow, though the contribution share of monthly maintenance is unknown (merchant claim, no third-party verification); ④ Opportunity Item: Ultra-low-cost contract manufacturing for quick order snatching, charged per episode or per word, with pricing as low as 1/3 of the market rate (the total cost for a multi-cast 10,000-word package is only 7-8 RMB). The proportion of ultra-low-price order snatching is not detailed.
💸 Cost
Local deployment has no software subscription costs. If choosing to rent a cloud GPU server, the monthly fee is about 500 RMB. Running it locally incurs only electricity and graphics card depreciation costs, totaling about 200-300 RMB. Overall costs are just 1/5th of commercial AI voice platforms.
⏱ Time Investment
Dedicate about 4-5 hours daily to batch synthesis and manual review/proofing.
🚀 Getting Started
Beginners should first search GitHub for the XTTS-v2 open-source project, download the packaged deployment bundle to run a basic demo inference on a local computer, then join author communities on Fanqie Novel and Qimao Chinese Network, publish a free 3-minute voice audition sample to attract author collaborations, and take on short projects under 100,000 words at a low price for the first order to build a reputation.
🔑 Keys to Success
- ✅ Open-source model selection and voice cloning debugging capability
- ✅ Multi-role voice library accumulation and emotional expression tuning
- ✅ Network expansion within web novel platform publishing circles
- ✅ Post-processing capability for long-audio splicing and noise reduction
- ✅ Platform copyright rule avoidance and compliance audit capability
⚠️ 风险
- ⚠️ Platform policy restrictions and copyright risks regarding purely AI-synthesized audio
- ⚠️ Cloned voice similarity failing to meet client requirements, leading to rework
- ⚠️ Quality issues such as phoneme loss and noise during batch synthesis of long audio, leading to compensation liabilities
- ⚠️ Order fluctuations and unstable income caused by relying on a single order channel
📌 Real Cases
- 📌 Sesame Pie AI reported that an individual creator used XTTS-v2 open-source voice cloning technology to customize multi-cast audiobook voice-overs for novels on web novel platforms, achieving a monthly income of over 10,000 RMB.
- 📌 Following the public beta test of Wanxiang Audio in March 2026, an individual studio used XTTS-v2 to batch-produce multi-cast web novel audiobooks, listing them on the Lanren Tingzhu platform and generating monthly revenue shares exceeding 8,000 RMB.
- 📌 A public case study by the maintainer of the GitHub open-source project AI-NovelSpeaker-V2 showed that they used XTTS-v2 to handle outsourced voice-over orders from Fanqie Novel authors, securing a stable monthly income of 12,000 RMB.