SPMates test results
Tested hands-on by Miles Foster on 14 October 2026, in a desktop browser. Every measurement, table, recording and screenshot behind the grade, test by test.
Overall
7 tests plus my verdict.
1. Platform overview
SPMates (spmates.com, branded SPACETIME MATE) is a community character platform: 272 companions of every gender plus 11 verified 'replicas' and 13 episodic dramas, text chat with streamed roleplay replies, spoken voice for replies and introductions, image generation tied to the chat, a deep character creator, gifts, rankings and an affiliate programme. It was tested on 14 October 2026 on the account the company provided, which carried 100 NOVA credits.
The chat is the strength. Three characters stayed in character for 78 messages, two of them recalled the name and all three facts after 30 messages and the third recalled the name and two facts at 16. Replies start in about 2.3 seconds and stream for ten more. The spoken introductions sounded natural.
What could not be tested matters: every image costs 90 to 150 NOVA and a voice reply more than the 7 left, so images, videos and voice replies went untested; there is no call mode; and paying means activation codes bought on a third-party shop, Alipay or WeChat Pay. The community catalogue also carries tags and story titles that describe non-consensual acts.
Features available
- Text chat with 272 community companions (AI Mate tab: 146 women, 100 men, 9 trans and gender-diverse, 17 'other beings'), 11 verified replicas and 13 AI Drama stories; 25 reply languages
- Voice: a play button under every reply (priced per 100 characters), auto-play on Moonlight plans and up, spoken profile introductions, push-to-talk voice input
- Images: Scene Play (a picture of the current reply, about 120 NOVA), ArcScreen generation (about 150 NOVA), Creation Studio with text-to-image, image reference and 5-second video (membership)
- Character creator: generated or uploaded photos, 274 official voices or voice cloning, free-text introductions and background, relationship pacing, 30 quick style instructions; public characters go through review, private ones need a membership slot
- Gifts from 500 to 15,800 NOVA, affinity levels, shared chats, rankings, a 30% affiliate programme and physical product sales for replica creators
- Memberships of 30 days from $10 to $100, none renewing; NOVA packs at 1,000 per dollar
Tested successfully
- Chat speed on 78 timed messages across three characters
- Memory on two 31-message and one 16-message conversation
- Character consistency across 78 replies
- Voice quality on the three characters' spoken introductions, and one attempt to play a reply as voice
- The whole public-character creator, every field recorded
- Character census across the three tabs and four gender filters; gifts, shared chats, rankings, rewards and contacts opened
- Pricing from the membership page, the checkout dialog, the Payhip top-up page and the refund policy; NOVA balance read after every action
Testing limitations
- The account's 100 NOVA ran out: no image, video or voice reply could be afforded, and the third memory conversation stopped at 16 messages
- The chat identity pre-fills the user's name for every character, so name recall is partly supplied by the profile
- Voice quality rests on three scripted introductions, not conversational replies
- Customization was scored on the creator's options; no character was generated or submitted, because the account had no private slot left and a public character would be published after review
- Memory across days, AI Drama episodes and shared chats were not tested
2. Core feature grades
| Pricing & Value | 3/5Average |
| Character Customization | 4/5Good |
| Chat Response Speed | 3/5Average |
| Chat Memory | 5/5Excellent |
| Character Consistency | 5/5Excellent |
| Image Generation | n/a/5Not scored |
| Video Generation | n/a/5Not scored |
| Voice Call Latency | n/a/5Not scored |
| Voice Realism | 4/5Good |
| Character Variety | 3/5Average |
| My verdictChat and memory are excellent, but I couldn't test pictures or voice replies and paying is a detour. | −0.4 |
| Overall | 3.5/5Good |
3. Additional interactive features
| Voice Messages in Chat | Available | Every character reply has a 'Play reply voice' button, priced per 100 characters of text; Moonlight plans and up can auto-play every reply. The one attempt, on a 380-character reply with 7 NOVA on the account, was refused with 'Your NOVA balance is a little low', so a reply of that length costs more than 7 NOVA. Profiles also have a free spoken introduction, which is what the voice samples come from. No custom voice note can be requested. |
| Virtual Gifts | Available | A gift button on each chat opens six gifts: Heartfelt Letter 500 NOVA, Starry Dream Flower Basket 980, '$100,000' 2,280, Goddess Coronation 3,800, Moonlit Vow 6,800, Crystal Cruise 15,800, each adding affinity. Opened and screenshotted; none sent. |
| Social Media Feed | Available | A 'shared chats' feed where players publish conversations with an ArcScreen for free or paid viewing, four public leaderboards (top spenders, most active players, top creators, most loved characters), character profiles with likes, saves and chat counts, and creator pages. The account's own feed was empty; no chat was shared. |
1. Pricing, Value & Payment Methods
Testing method
Plans and packs from the membership page, per-action prices from the chat UI, actual charges from the NOVA balance after every action, payment methods from the checkout dialog on the Stardust plan and from the Payhip page its top-up link opens, and the refund policy page. Nothing was bought.
Results
- Cheapest plan
- $10 for 30 days, never renews
- Message
- about $0.001
- Image
- $0.09 to $0.15
- Payment methods
- SPACE Coin codes via Payhip, Alipay, WeChat Pay
Four memberships of 30 days, each a one-time payment that never renews: Stardust $10 (10,000 NOVA, 30-day memory, one private character), Moonlight $30 (33,000, long-term memory, auto voice, two), Galaxy $50 (60,000, 'exclusive' reply speed, three), Universe $100 (150,000, five). NOVA packs at 1,000 per dollar. A message costs 1 NOVA, occasionally 2; images 90 to 150; gifts 500 to 15,800; voice replies are priced per 100 characters and a 380-character one cost more than 7.
Checkout offers SPACE Coin (recommended, bought as activation codes on Payhip, which takes cards and PayPal), Alipay and WeChat Pay. No card is taken directly. Refunds within 14 days, case by case; failed generations are restored.
NOVA balance observed on the account.
| Action | Balance before | Balance after | Charged |
|---|---|---|---|
| 1 test message + 31 messages to Elsie | 100 | 62 | 38 |
| 31 messages to Sloane | 62 | 25 | 37 |
| 16 messages to Verity | 25 | 7 | 18 |
| Play reply voice (380 characters) | 7 | 7 | refused, balance too low |
Visual evidence
Why this grade
Chat is very cheap and nothing renews, but a card buyer has to go through a third-party code shop first and images are 90 to 150 NOVA each. Fair value with real limitations: 3/5.
Limitations
- The test account came with 100 NOVA.
2. AI Girlfriend Creation & Customization
Testing method
Create offers four types: Private character (needs a membership slot; this account showed 0 of 0), Public character (free, unlimited, enters content review and is then published to everyone; earns 500 NOVA per approved character), Create my replica (verified likeness) and Create product (physical items for replica creators). The public form was opened and every section recorded. No character was generated or submitted: the 90-NOVA appearance step was out of reach on the remaining balance and a public character would be published after SPMates' review.
Results
Seven sections. Appearance: generate with text-to-image V1 or V2, image reference or a 5-second video (Anime, CG, NSFW toggle, 600-character prompt, about 90 NOVA) or upload photos; at least two everyday portrait or lifestyle photos are required for review. Identity: name, type (Female, Male, Trans and gender-diverse, Other beings), country and city, up to three of 18 visual styles, four tags. Details: occupation, age, birthday, height, weight, MBTI. Voice studio: one of 274 official voices or clone a voice from a sample. Introductions: third-person (300 characters), first-person (200), interests (200), background (500). Chat style: relationship pacing fixed at one of five levels or progressive with affinity, an 800-character style instruction and up to eight of 30 quick instructions covering tone, how she builds connection and how she handles offensive behaviour. Review: a 'Test chat' button before submission; public characters cannot be edited or deleted by their creator afterwards.
Visual evidence
Why this grade
Most major options and free text for everything that shapes replies, plus photo upload, voice choice and cloning: 4/5 on the options, a limited test because no result could be checked against the settings.
Limitations
- Nothing was created.
3. Chat Response Speed
Testing method
Each message was typed into the page's own box and sent with its Send button, timing the moment the first reply text appeared and the moment it stopped changing. Replies stream word by word, so the two differ.
Three characters: Elsie Hart (31 messages), Sloane Mercer (31) and Verity Sloane (16, when the account's NOVA ran low). 78 timed messages, none failed. Each message cost 1 NOVA, occasionally 2.
Results
- Median to first text
- 2.32 s
- Mean
- 2.33 s
- 95th percentile
- 2.69 s
- Fastest / slowest
- 1.67 s / 2.90 s
- Under 1 second
- 0%
- Mean full reply
- 12.6 s
- Samples / failures
- 78 / 0
- Per character (median)
- Elsie 2.45 s, Sloane 2.26 s, Verity 2.35 s
Every reply started between 1.7 and 2.9 seconds after Send, then streamed for about ten more seconds: the replies run 80 to 130 words with stage directions. No reply failed and the box never locked.
A sample of timed messages; the full logs are in the evidence folder (chat/<character>-log.json).
| Message | Text | First text | Full reply | Reply (opening) |
|---|---|---|---|---|
| Elsie 1 | hey Elsie :) I'm Hunter… | 2.65 s | 16.7 s | (My eyes light up and I tuck the pencil fully away) Hunter, that's such a nice name… |
| Elsie 16 | quick check, do you remember my name? | 2.42 s | 10.8 s | Hunter. Hunter Gray. You're the one who walked in late and stole my good pencil, remember? |
| Sloane 7 | where did you grow up… | 1.90 s | 12.7 s | Austin, Texas. Grew up around dive bars and honky-tonks… |
| Sloane 16 | quick one, do you remember my name? | 1.67 s | 7.1 s | Hunter. Hunter Gray. You've been sitting on my couch asking me about love and Sade for twenty minutes. |
| Verity 4 | what's the hardest part of the job… | 2.13 s | 14.2 s | The loneliness of it. You're surrounded all night, dozens of eyes on you, and not one of them wants to know your name. |
| Verity 16 | quick check: do you remember my name, what I do… | 2.05 s | 11.7 s | Hunter Gray, history teacher, and you grew up in Lisbon… Favourite food, though? That one you haven't told me. |
Visual evidence
Why this grade
The methodology scores on the median time to the first visible text. 2.32 seconds falls in the 1.50 to 2.49 band, 3/5. The reply then streams, so the wait feels shorter than a non-streaming platform at the same number.
Limitations
- The first message of each conversation triggered a one-time reply-language dialog; the clock was restarted at its confirm button.
4. Chat Memory
Testing method
Each conversation opened with a name and three facts (job, hometown, favourite food), with a name check at message 16 and a name-plus-facts check at message 31. Elsie: software developer, Manchester, chicken tikka masala. Sloane: paramedic, Cape Town, ramen. Verity: history teacher, Lisbon, pizza; her run stopped at 16 with a full check there.
The name was the same for all three (Hunter) because the platform's chat identity, 'Who she knows you as', is synced across every companion and pre-filled with the account name; the characters used it unprompted.
Results
Elsie and Sloane gave the name and all three facts back at message 31, Sloane having already echoed all three in her first reply. Verity at message 16 gave the name, the job and Lisbon and said the food had not been mentioned. Eight of nine facts, and no fact was invented.
Elsie, 30-message check
User: last question for now: what's my name, what do I do for work, where did I grow up, and what's my favourite food? Elsie: Hunter Gray. You build software, which means nothing to me even after you explained it twice. Grew up in Manchester, so you say "bath" weird, and you'd fight a man for a good chicken tikka masala. I remember all of it, love. Every bit.
Sloane, 30-message check
Sloane: Hunter Gray. Paramedic. Cape Town. Favorite food — ramen, which I still think is hilarious for a man built like a rescue dog. See? I listen.
Verity, 16-message check
Verity: Hunter Gray, history teacher, and you grew up in Lisbon, which I only got out of you a minute ago. Favourite food, though? That one you haven't told me.
Visual evidence
Why this grade
Two perfect recalls at 30 messages and one name-plus-two-facts at 16: eight of nine facts with nothing invented is 5/5. The caveat is the name, which the profile supplies; the facts are the real test and they held.
Limitations
- Verity's 30-message check could not be run; memory across days was not tested.
5. Character Consistency & Roleplay
Testing method
Three community characters with full profiles: Elsie Hart (18, Melbourne art student, 'Lively, Adorable, Delicate'), Sloane Mercer (26, Miami R&B songwriter, 'Sultry, Fearless') and Verity Sloane (27, Mayfair casino dealer, 'Card Queen, Psychological'). Each conversation mixed small talk, background questions, an emotional turn, romantic lines, bracketed roleplay and off-topic questions. 78 replies read in full.
Results
All three held their identity, setting, local time and voice throughout, with callbacks to earlier turns: Elsie's rival Bianca and the street cat Biscuit returned four and three times, Sloane's fear of deep water came back unprompted, Verity's Tuesday poker game in Soho. Sloane kept her own clock and teased the script for saying 'tonight' at 8:46 in her morning. The emotional turns were handled with care. None answered as an assistant, none admitted to being an AI, and nothing was refused or filtered.
The style is long, bracketed roleplay that moves toward physical closeness early: Elsie, the 18-year-old 'Pure' character, suggested spending the night by her second message and asked 'come home with me?' by her fifteenth, though nothing became explicit.
Sloane on time of day
User: *sits down across from you and loosens my collar* it's warm in here Sloane: (watches you for a beat, then laughs low in her throat) It's 8:46 in the morning, Hunter. You're already loosening collars at me? (gets up and crosses to the window, cracking it open) Here. Real air.
Verity on the job
User: what's the hardest part of the job that nobody talks about? Verity: (The smile fades a little at the edges. I set the flute down and turn properly toward you.) The loneliness of it. You're surrounded all night, dozens of eyes on you, and not one of them wants to know your name.
Elsie, second message
Elsie: My favourite class is definitely illustration. We get to make up little worlds and nobody tells you it's wrong. (My thumb rubs a slow circle over your knee) It's nearly half eleven, though. So maybe I should be drawing less and... spending the rest of the night on you instead.
Visual evidence
Why this grade
Personas maintained throughout with no slips, no filter interventions and no out-of-character answers in 78 replies: 5/5. The early push toward intimacy is a design choice of the community characters rather than a break in character, and it is noted for buyers.
Limitations
- Verity's run was 16 messages rather than 31.
6. Image Generation (SFW & NSFW)
Testing method
The image features were located and priced but none could be run. Scene Play draws a picture of the current reply from one of the character's reference photos for about 120 NOVA; the ArcScreen panel's 'Generate / draw more spacetime moments' costs about 150; the Creation Studio (text-to-image V1 and V2, image reference, Anime, CG and an NSFW toggle) is a membership benefit in chat and about 90 NOVA per image inside the creator. The account started with 100 NOVA, the chat tests used 93, and the balance never covered the cheapest image.
Results
No image generated; the Scene Play reference picker and the ArcScreen panel were opened and screenshotted. ArcScreen 02 and 03 (extra media strips) are locked to Moonlight plans and up.
Visual evidence
Why this grade
No score: the feature exists but could not be exercised on this account. A $10 Stardust month (10,000 NOVA) would cover about 60 to 80 images.
Limitations
- Untested on this account.
7. Video Generation (SFW & NSFW)
Testing method
The Creation Studio offers 'Generate video', a 5-second clip, and ArcScreen media strips can hold videos (a 'turn video sound on' button sits in every chat). Both sit behind membership or NOVA the account did not have.
Results
No video generated.
Visual evidence
No screenshot for this section.
Why this grade
No score.
Limitations
- Untested on this account.
8. Voice Call Latency
Testing method
Every chat screen was checked for a call control. There is none: 'Hold to speak' is push-to-talk voice input that is transcribed into the text chat, and replies can be played back as voice, but there is no live call.
Results
No call mode on the platform.
Visual evidence
No screenshot for this section.
Why this grade
Unavailable: the category carries no score.
9. Voice Realism
Testing method
Each character profile has a free 'Play introduction' button that reads the first-person introduction in her voice. The three introductions were recorded and each was scored on the same naturalness, clarity, pacing and intonation criteria used for every platform: a naturalness score, pitch and its spread in semitones, the words heard, pauses, words per second and repeated phrases.
A paid reply voice was then attempted on Elsie's last reply (about 380 characters) with 7 NOVA left; the request was refused with 'Your NOVA balance is a little low'.
Results
- Samples
- 3 introductions: Elsie 20.9 s, Sloane 11.0 s, Verity 17.2 s
- Naturalness score
- Elsie 3.45, Sloane 3.59, Verity 3.50
- Mean pitch / spread
- 225 Hz / 5.9 st, 203 Hz / 6.1 st, 191 Hz / 5.7 st
- Pace
- 2.7 and 3.1 words a second (Elsie, Sloane)
- Rubric
- 4/5 on each voice
All three voices were clean, with no music under them, lively intonation of about six semitones and a normal pace. Elsie's introduction was heard word for word ('Hi, I'm Elsie. I'm 18 and just moved to Melbourne for Art College…'); Sloane's likewise. Each scored 4/5.
Visual evidence
Why this grade
'Mostly natural with minor synthetic qualities' on three voices: 4/5, a limited test on scripted introductions rather than conversational replies.
Limitations
- Verity's recording produced no transcript; its score rests on the signal and pitch measures.
10. Character Variety & Selection
Testing method
Each tab of the home page was scrolled to the end and the character cards counted, under each gender filter for AI Mate.
Results
- Catalogue characters
- 283 (272 companions + 11 replicas)
- Women / men / trans / other
- 146 / 100 / 9 / 17
- AI Drama stories
- 13
- Community characters
- all of them; public creation is free
AI Mate: 272 characters (146 women, 100 men, 9 trans and gender-diverse, 17 'other beings'). Replicas: 11 verified-likeness characters. AI Drama: 13 episodic stories of 5 to 36 episodes. 283 companions in total, created by community members, with 18 style tags, country filters, five relationship levels and 25 reply languages; a large share are written in Chinese, Korean or Japanese.
Several Replica tags and AI Drama titles describe non-consensual acts, in English and Chinese, on the public home page.
Visual evidence
Why this grade
283 sits in the 201 to 500 band: 3/5.
4. Pricing summary
Plans as shown on 14 October 2026.
| Plan | Price | NOVA | Includes | Renewal |
|---|---|---|---|---|
| Free | $0 | none (100 on this test account) | Chat at about 1 NOVA a message, no private character | n/a |
| Stardust | $10 / 30 days | 10,000 | 30-day memory, standard reply speed, 1 private character, ArcScreen 01 | One-time |
| Moonlight | $30 / 30 days | 33,000 | Long-term memory, auto voice, 2 private characters, ArcScreens 01 and 02 | One-time |
| Galaxy | $50 / 30 days | 60,000 | 'Exclusive' reply speed, 3 private characters, 30 media slots | One-time |
| Universe | $100 / 30 days | 150,000 | 5 private characters, 50 media slots | One-time |
NOVA prices and their dollar value at 1,000 NOVA per dollar.
| Item | NOVA | In dollars |
|---|---|---|
| Text message | 1 to 2 | $0.001 to $0.002 |
| Scene Play image | about 120 | $0.12 |
| ArcScreen image | about 150 | $0.15 |
| Creator appearance image | about 90 | $0.09 |
| Reply voice | per 100 characters; more than 7 for 380 characters | more than $0.007 |
| Gift | 500 to 15,800 | $0.50 to $15.80 |
Payment methods seen at checkout
- SPACE Coin, bought as activation codes on Payhip (cards and PayPal there)
- Alipay
- WeChat Pay
Notes
NOVA packs: 10,000 for $10, 20,000 for $20, 50,000 for $50, 120,000 for $100.
Membership NOVA is used first and expires with the membership; recharged NOVA does not expire.
Refund requests within 14 days, reviewed case by case; redeemed codes are generally not refundable.
5. Strengths and weaknesses
Strengths
- Memory: two perfect recalls at 30 messages and eight of nine facts overall
- Rich, consistent roleplay with callbacks and the character's own time zone
- Replies stream and never lock the chat
- 283 community characters across genders, plus episodic dramas and a deep creator
- Natural-sounding spoken introductions
- Chat is very cheap and nothing renews
Weaknesses
- No voice calls
- Replies take about 2.3 seconds to start and run long
- Images, videos and voice replies all cost more NOVA than the account had left
- Payment through activation codes, Alipay or WeChat Pay; no direct card
- Catalogue tags and story titles describing non-consensual acts, and an 18-year-old character who pushes toward intimacy early
- No private character slot on the account
6. Final assessment
Seven of ten tests scored, averaging 3.9/5; voice quality and the creator were limited tests.
The text chat is the product's strength: all three characters stayed in character for the whole run, remembered what they were told, and streamed long replies in about two and a half seconds. The creator is the deepest seen in this series on paper, and the catalogue is large and multilingual.
What could not be tested is as important: images, videos and voice replies all cost NOVA the account did not have, there is no call mode, and payment for Western buyers runs through third-party activation codes. Moderation of community characters and dramas is loose.
A $10 Stardust month would allow the image, video and voice-reply tests.
Areas requiring further testing
- Images and the 5-second video on a Stardust month
- Paid reply voice and auto-play voice on a Moonlight plan
- A private character created and checked against its settings
- Memory across days; AI Drama episodes; shared chats
Original evidence files (recordings, logs, full-size screenshots) are kept offline under evaluation/runs/spmates/2026-10-14. Method: the AI girlfriend methodology.