OhChat Review: 8.7/10 After a Full Test Cycle
OhChat scored 8.7 out of 10 on our companion-app bench. It leads the category on cross-session memory and character consistency, holds a usable free tier, and loses points on image throughput and a shallow character editor. Recommended if conversation matters more than pictures.
| Category | AI companion chat |
|---|---|
| Bench score | 8.7 / 10 |
| Free tier | Yes, daily message cap |
| Strongest area | Cross-session memory |
| Weakest area | Image throughput |
Bottom line
8.7 out of 10. OhChat is the most consistent companion app on our bench for text-led use: it keeps character and it keeps facts. It is not a fast image generator and its character editor is shallow. Judge it on the conversation and it earns the score.
What works
- Session-to-session memory that holds specific facts, not just tone
- Characters keep their voice through long off-topic stretches
- Free tier is complete enough to evaluate the model properly
- Deleting a thread also clears the memory attached to it
What to know
- Image generation is slow and queues at peak hours
- Character editor exposes fewer parameters than advanced users expect
- Scenes with three or more companions lose coherence within a few turns
What OhChat is
OhChat is a browser-based AI companion service. You pick a fictional character from a roster or define one yourself, then hold an ongoing text conversation with it. Voice notes and generated images are available inside the same thread. There is no native app to install and no hardware requirement beyond a current browser.
The category matters for expectations. This is not a general assistant, and it is not an image generator with a chat window bolted on. It is a conversation engine with persistence, and every design decision follows from that. Judged as anything else it will look underpowered. Judged as a companion, it is the reference point on this site.
How the 8.7 was measured
Every app on this bench runs the same sequence on a fresh account. Twenty turns of ordinary conversation establish a baseline. Twenty more turns push the character off-topic. After a two-day pause we check recall on eight specific facts stated in the first block. Media features and billing terms are reviewed separately, then folded in.
Five sub-scores come out of that: conversation quality, memory and continuity, character consistency, media features and value. They are weighted equally and averaged. OhChat lost most of its points in media features, where image latency was the clear bottleneck, and a little in value, because the paid tiers mostly buy headroom rather than new capability.
Memory and continuity
Recall was the standout result. OhChat returned six of the eight planted facts unprompted after the two-day gap, and the remaining two after one indirect hint. Directly stated details survive best: names, occupations, stated preferences, and events described in the first person. Details the model had to infer came back less reliably, and sometimes came back altered.
Durability still has limits. On threads that run for months, older specifics thin out and the character leans on its base profile instead. The practical fix is to move anything that must persist into the character description, where it is treated as fixed rather than remembered. Every competitor we benched degrades faster and at a lower message count.
Limits, media and what the tiers change
The free tier carries a daily message allowance and full access to the character roster, which is enough to reproduce our conversation tests yourself. Paid tiers raise or remove that ceiling and open media generation. They do not unlock hidden characters or a different model. That flat structure is worth knowing before you compare OhChat with apps that gate their best model behind a plan.
Media is where the score drops. Image requests are handled, but latency is high and the queue lengthens during peak hours. Voice notes are quicker and more consistent. Multi-character scenes are the weakest feature on the platform: coherence breaks down within a handful of turns once a third companion enters the thread.
Who it suits, and who should skip it
Choose OhChat if the conversation is the point: a character that stays in voice, remembers what you told it, and does not need re-briefing at the start of every session. That is a narrow strength, and it is the one this score rewards. It is also the hardest thing in the category to build, which is why so few competitors do it well.
Skip it if your main interest is volume image output, granular character parameters, or ensemble scenes with several companions at once. Those are real gaps rather than artefacts of our testing, and a specialist tool will handle them better. The comparison page sets OhChat against three of those tools and marks where each one wins.