RUNLOCALAIv38
->Will it run?Best GPUCompareTroubleshootStartLearnPulseModelsHardwareToolsBench
Run check
RUNLOCALAI

Independently operated catalog for local-AI hardware and software. Hand-written verdicts. Source-cited claims. Reproducible commands when we have them.

OP·Fredoline Eruo
DIR
  • Models
  • Hardware
  • Tools
  • Benchmarks
TOOLS
  • Will it run?
  • Compare hardware
  • Cost vs cloud
  • Choose my GPU
  • Prompting kits
  • Quick answers
REF
  • All buyer guides
  • Learn local AI
  • Methodology
  • Glossary
  • Errors KB
  • Trust
EDITOR
  • About
  • Author
  • How we make money
  • Editorial policy
  • Contact
LEGAL
  • Privacy
  • Terms
  • Sitemap
MAIL · MONTHLY DIGEST
Get monthly local AI changes
Monthly recap. No spam.
DISCLOSURE

Some links on this site are affiliate links (Amazon Associates and other first-class retailers). When you buy through them, we earn a small commission at no extra cost to you. Affiliate links do not influence our verdicts — there are cards we rate highly that we don't have affiliate relationships with, and cards that sell well that we refuse to recommend. Read more →

© 2026 runlocalai.coIndependently operated
RUNLOCALAI · v38
Glossary / Generative AI / Deepfake
Generative AI

Deepfake

A deepfake is a synthetic media (image, video, or audio) generated or manipulated by a deep learning model, typically an autoencoder or GAN, to depict a person doing or saying something they never did. Operators encounter deepfakes when using local AI tools like Stable Diffusion or audio cloning models (e.g., RVC) to swap faces or voices. The term matters because deepfakes raise ethical and legal concerns, and local AI operators must be aware of misuse risks and detection methods.

Deeper dive

Deepfakes emerged from academic research on autoencoders and generative adversarial networks (GANs). A typical deepfake pipeline involves training an encoder-decoder pair on source and target faces, then swapping latent representations. Modern approaches use diffusion models (e.g., Stable Diffusion with inpainting) or neural radiance fields for higher quality. On local hardware, face-swapping tools like Roop or FaceFusion run on consumer GPUs (e.g., RTX 3060) at real-time speeds for 720p video. Audio deepfakes use text-to-speech models (e.g., Tortoise TTS) or voice conversion (e.g., RVC) to clone a speaker's timbre. Detection relies on artifacts like inconsistent blinking, lighting mismatches, or frequency analysis. Operators should watermark generated media and avoid non-consensual use.

Practical example

An operator runs FaceFusion on an RTX 4070 to swap a celebrity's face into a video clip. The tool loads a pre-trained InsightFace model (~200 MB) and processes frames at ~30 FPS for 1080p. VRAM usage peaks at ~4 GB. The operator must ensure they have consent from the person whose face is used, as non-consensual deepfakes can violate laws.

Workflow example

In LM Studio, an operator loads a voice cloning model like RVC. They record a 30-second sample of a target voice, then run inference to convert their own speech into the target's voice. The workflow involves selecting the model file (e.g., rvc_v2.pth), adjusting pitch and format parameters, and exporting the output as a WAV file. The operator must verify the output for artifacts and avoid impersonation without permission.

Reviewed by Fredoline Eruo. See our editorial policy.

Buyer guides
  • Best GPU for local AI →
  • Best laptop for local AI →
  • Best Mac for local AI →
When it doesn't work
  • CUDA out of memory →
  • Ollama running slowly →
  • ROCm not detected →