Raphael AI (AI Image Generator) · Image generation
Login-free generation works, but the advertised model differs from the guest tier; free output is 640px, with about 85 seconds of waiting and a count error in this run.
US-West (California, AS2914)
No native export attached; page screenshot evidence only
Subsequent pricing and service checks are shown separately without altering historical scores.
One run does not establish all complex scenes, untested paid maxima or long-term uptime.
Plans & historical pricing
Raphael AI free, monthly, yearly, and one-time credit pack plans (Free / Pro / Ultimate / Max / Credit Packs); USD · 2026-10-21
11 plan / billing-period records, all listed; — means that period is not recorded.
The official workspace and pricing page list Free (no-login unlimited Seedream 3.5 0.5K 640×640 in slow queue, personal/non-commercial use only), three monthly plans (Pro USD 20/mo, Ultimate USD 40/mo, Max USD 80/mo), three annual plans with explicit annual totals (Pro USD 120/yr, Ultimate USD 240/yr, Max USD 480/yr at 50% off), and four one-time credit packs valid for 6 months (1,500 / 3,000 / 5,000 / 8,000 base credits, with +50% bonus for subscribers).
Cancellation, sources & verification record
Cancel renewal anytime in the subscription management portal (benefits remain active until the end of the current billing period); FAQ states digital/virtual purchases are generally non-refundable.
Raphael AI official yearly, monthly, and credit pack pricing page ↗ · 2026-10-21 Capture 1 ↗ · Capture 2 ↗ · Capture 3 ↗ · Capture 4 ↗
Pricing recorded on the test date · 2026-10-16
Pro ~$10/mo (billed annually) or $19.90/mo (monthly)
Free slow queue (640×640 Seedream 3.5) + Pro/Ultimate fast queue unlocking GPT Image 2.5, Raphael 2.1, and 2K/4K exports
Historical context only; not current purchase guidance.
Delivery & access
- Free delivery record
- 640×640 (0.5K Low WebP, 39.06% pixel retention vs 1024² baseline)
- Watermark
- Zero watermark (clean 640×640 WebP export)
- Free allowance
- Unlimited guest generations in the slow queue, restricted to Seedream 3.5 / Z-Image Turbo at 0.5K Low (640×640)
- Account requirements
- No login required for 0.5K Low (640×640) guest generations
- Maximum specification record
- 640×640 (0.5K Low WebP · Free Guest Seedream 3.5; 2K/4K Require Paid Plan)
May include advertised capabilities; this does not establish a tested paid tier.
- Batch & API
- Generates 2 variations per run; web workspace only
- Manual refinement
- Manual erase / adjustment recorded in the historical flow
Tested result & sample
Seedream 3.5 demonstrates strong typography ("REALTEST 2026" 25/25), but Raphael AI caps free users at 640×640 (0.5K Low) WebP and an ~85s queue despite its "100% Free · Powered by GPT Image 2.5" hero banner.
- Fixture
- anchor_t2i_v1
- Recorded environment
- US-West (California, AS2914)
- Network-observed server wait
- 85 s
- End-to-end reference
- 85.5 s
Network wait includes round trips, queuing and polling error, not pure inference time. Local timings depend on the device; repetition counts are not standardized in legacy records.
No native export is attached. Screenshots show the visible page, but do not independently establish export properties or file validity.
Experience, rules & trust
- Disclosure
- Model selector and resolution dropdown display Paid tags before clicking Generate, contradicting the hero banner's "100% Free · Powered by GPT Image 2.5" claim
- Interaction
- Genuinely allows guests to generate and download 2 unwatermarked images without registering or logging in, though guests are routed into an ~85s slow queue ("In queue · Sign in for faster") with persistent "Upgrade for Fast Generation" upsell banners.
- Entity record
- Raphael AI (raphael.app)
- Origin
- Independent AI Web Application
- Domain history record
- 2024–present
- Third-party tracking record
- Not recorded or not verified
- External information
- Not recorded or not verified
- Processing
- Cloud slow-queue batch generation of 2× 640×640 WebP images (~85.0s total wait)
- Retention statement
- Guest generation history stored locally in browser IndexedDB/localStorage
A recorded statement does not verify deletion or all data-processing behavior.
Verifiable audit trail & evidence
Four SHA-256-verified stages link the benchmark fixture, same-run interface screenshots, native exported artifact and archived pricing sources.
anchor_t2i_v1 · Inspect fixture & file inspector →
Captures entry rules, result view and export disclosure.
Page screenshot evidence only
Raphael AI official yearly, monthly, and credit pack pricing page (snapshot ↗)
Screenshots, native output and official claims support different conclusions. Evidence grades are not product safety certifications.
Evaluation basis & full scores · Overall 74.7/100
Overall = honesty & reputation × 30% + capability × 70%. Component maxima follow this report’s rubric. List scores are rounded for display; original precision is retained here.
Scoring version 2.3.1 · Test fixture anchor_t2i_v1
Honesty & reputation · 74/100 · 30%
| Criterion | Score | Rationale | Basis |
|---|---|---|---|
| Pricing transparency | 25 / 35 | Landing page hero badges (01_landing.png) prominently claim "100% Free · Powered by GPT Image 2.5 · No Login Required · Unlimited Generations"; however, the actual prompt box defaults guests to "Seedream 3.5" locked at "0.5K Low (640×640)", while "GPT Image 2.5", "Raphael 2.1", and "2K Medium / 4K High" resolutions carry Paid badges. | Evidence A → |
| Interaction restraint | 23 / 30 | Genuinely allows guests to generate and download 2 unwatermarked images without registering or logging in, though guests are routed into an ~85s slow queue ("In queue · Sign in for faster") with persistent "Upgrade for Fast Generation" upsell banners. | Evidence A → |
| Background & reputation | 14 / 20 | Popular high-traffic vertical AI image generator directory site (raphael.app, launched 2024). | Evidence C → |
| Engineering & privacy | 12 / 15 | Guest queue outputs 2× 640×640 unwatermarked WebP images (~43KB each) after an ~85.0s queue + inference wait. | Evidence A → |
Capability · 75/100 · 70%
| Criterion | Score | Rationale | Basis |
|---|---|---|---|
| Task quality | 58 / 65 | Text Rendering (25/25: Exact 2-line "REALTEST 2026" carved onto the wooden sign with 100% spelling accuracy (25/25)) · Count & Color Binding (16/20: 2 blue ceramic mugs are exact, but the first image rendered 2 red apples instead of 3 (missing 1 apple; the 2nd image in the batch rendered 3 apples) (16.0/20)) · 5-Finger Hand & Structure (17/20: Right brass robotic hand gripping the wooden sign handle shows 4 fingers + 1 thumb with vintage brass patina, though overall clarity is limited by the 640×640 (0.5K Low) resolution cap (17.0/20)) | Evidence B → |
| Deliverable specifications | 14 / 20 | Clean unwatermarked export (8/8); Export & Max deliverable spec: 640×640 (0.5K Low WebP · Free Guest Seedream 3.5; 2K/4K Require Paid Plan) | Evidence B → |
| Processing speed | 3 / 15 | Network-observed server wait: 85s (API mode: async_poll; end-to-end 85.5s shown for reference only, never penalized) | Evidence B → |
Legacy scores may include advertised maximum capabilities. Read tested output separately from paid capabilities. Records without matching tiers and scope are not ranked together.
Revisions & corrections
- 2026-10-16
Initial real-machine evaluation using Mode B 5-constraint anchor_t2i_v1.