AuraSR
Open-source 600M-parameter GAN upscaler that quadruples images in about a quarter second.
About
AuraSR is Fal's open-source super-resolution model, built as a variation of Adobe's GigaGAN paper for image-conditioned upscaling, with a PyTorch implementation derived from the lucidrains gigagan-pytorch codebase. Where diffusion upscalers trade speed for quality, AuraSR goes the GAN route: roughly 600M parameters, 4x upscaling, and processing in about 0.25 seconds — fast enough for near-real-time pipelines and game graphics enhancement.
Fal released it openly in 2024 with model weights on Hugging Face (fal/AuraSR and the improved AuraSR-v2), a playground demo on fal.ai, and a simple pip-installable Python package. One wrinkle: no formal benchmark numbers have been published, so quality claims rest on demo impressions and community feedback rather than measured evals. The repo is CC-BY-SA-4.0, so commercial users should check the share-alike terms.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No published benchmarks found — Fal has released no formal eval numbers; quality claims rest on demo output and community feedback
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (TIER metrics, metrics-v1). Evidence: Fal's official repo (fal-ai/aura-sr) and Hugging Face pages (fal/AuraSR, fal/AuraSR-v2); press coverage (aibase) corroborates the 600M-param, 4x, ~0.25s claims. Strengths: genuinely downloadable weights, real company backing. Docked for: (1) no published benchmarks at all — quality is unverified by measurement; (2) repo license is CC-BY-SA-4.0 (share-alike), less permissive than typical ML releases. License: CC-BY-SA-4.0 (repo).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 5.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · Fal official repo fal-ai/aura-sr (525 stars); HF fal/AuraSR + fal/AuraSR-v2; aibase press corroborates 600M params, 4x, ~0.25s claims
- · Zero published benchmarks — quality unverified by measurement; GigaGAN-variant GAN upscaler
- · Real company backing (Fal); fal.ai API + playground + pip package exist