Credit based usage model
Credits purchase generation capacity, but reviewers report unexplained losses and no automatic refunds on failure.
Credit based AI image and video generation platform aimed at creators needing fast visuals, but current reliability, model availability, and refund handling make it risky for any production dependent workflow.
Independent review — we test tools ourselves and analyze public user reviews. How we test.
Focalml offers multi model image and video generation in a credit based workspace that once satisfied some long term users. Recent reviews describe widespread model outages, repeated generation errors, and vanishing credits. Billing issues, non canceling subscriptions, and near total customer support silence are recurring themes. Given the instability and unresolved refunds, it currently suits only low stakes experimentation, not serious creative or business projects.
Focal is an agentic AI video platform built for the job most generators skip: producing full TV episodes and movies rather than isolated clips. It came out of Y Combinator’s Winter 2024 batch, run by a small San Francisco team.
The workflow begins from either an idea or a finished script. Hand it a premise and it writes a structured story with scenes and characters. Bring your own screenplay and it adapts the text into shots instead of rewriting your intent. Each shot is then routed to a specialised model, drawing from Veo, Seedance, Kling and Hailuo for motion, with GPT Image, Nano Banana, Flux and Seedream handling stills. ElevenLabs and InWorld supply the voices and sound.
What sets it apart from a single-model tool is continuity. Characters and locations can be generated once and reused across an entire project, which is the hardest problem in AI filmmaking. Editing works two ways, through plain-language chat commands or a traditional timeline, with video extension and frame interpolation for longer, smoother sequences.
Credits purchase generation capacity, but reviewers report unexplained losses and no automatic refunds on failure.
Enables creating narrative videos, though many attempts fail with errors that still consume credits.
Multiple named models exist, yet several return deprecation errors on every request according to reviews.
Catalog lists Gemini and other models, but maintenance gaps and missing replacements frustrate users.
Plans promise daily credits, although some subscribers report never receiving the advertised allotment.
Browser workflow supports prompts and asset stages, hindered by pipeline failures and unreliable processing.
Support channel exists but is widely criticized for non responses and unresolved tickets.
I signed up without paying anything and landed on a clean prompt box. Focal presented itself as an agentic video platform, with the model picker already set to GPT Image 1.5 for stills and Seedance 1.5 Pro for motion. My free account came with 40 credits, enough to run a short test.
One thing caught my eye straight away. A tooltip told me the Pro models sit behind a paid plan and nudged me to upgrade or switch to a free model. The onboarding looked current, with a banner at the top pushing a new Seedance 2.0 release.

For an honest read on quality I gave it a history prompt and let it render rather than judging the gallery.
Prompt used: “A short history of Isoroku Yamamoto's warning about Pearl Harbor”
The clip came back in an anime style, a naval commander leaning over a map table with battleships and drifting smoke behind him. Motion held steady across the shot, the face stayed consistent frame to frame, and narration subtitles were baked into the video. For a free-tier render on the basic model pairing, the output landed above what I expected at this price point, which is to say zero.

Signing in told a different story from the homepage.
A banner across the top read “Focal is sunsetting.” A second bar steered me to the All Projects page to export my work. A popup repeated the message and offered a Learn more link next to an Export your projects link. The sidebar still carried an Upgrade to Pro button and a Refer a friend offer for free credits, sitting beside my balance of 40 and speed tiers that priced generation at roughly 18 credits per minute on Fast and about 600 on Max. The model picker had switched to Flux Klein 9B with LTX2.
The popup would not close, and the page behind it would not scroll.

Opening Learn more loaded a blog post. Focal confirmed it had signed an acquisition offer and would wind down the existing product, thanking a community it put at over a million users. No shutdown date appeared anywhere on that page. The same locked popup sat over this post as well, so again I was stuck reading through a box I could not move.

The wind-down banner told me to export, so I tried. The modal offered a few scopes for the download, from a full copy of each project down to lighter picks that saved only the finished video or only the source media.
I selected the finished-video option and pressed the action button. It stayed greyed out. “Choose folder & start export” did not respond to a single click. The one function that matters most during a shutdown, getting my work off the platform, did not complete for me.

Here is where it turns strange. Logged out, the pricing page was fully live and taking money. Free sat at $0 with 30 starting credits and 10 daily once those run out. Starter ran $10. Standard, at $30, wore the Most Popular tag, and Pro topped the table at $100. Every tier had an active Subscribe button. A product that tells its signed-in users it is shutting down was still signing new ones up at the front door.

A sister domain made the gap impossible to miss. askfocal.com was already gone, showing a bare “Focal has been sunset” screen flagged as Unavailable.
So inside one brand I ran into three states at the same time. askfocal.com said it had already been sunset. The focalml.com app said it was sunsetting now and told me to export before it was too late. The logged-out focalml.com homepage said nothing at all and kept inviting new sign-ups while charging for plans.

| Dimension | Our test | User signal | Verdict | Composite |
|---|---|---|---|---|
| Generation Reliability Frequency of successful outputs | 2 | 1.5 | Weak | |
| Customer Support Speed and helpfulness of assistance | 2.5 | 2 | Weak | |
| Value for Money Perceived fairness of cost | 2 | 1.5 | Weak | |
| Billing and Refunds Accuracy and refund handling | 1.5 | 1 | Weak | |
| Model Availability Working models and uptime | 2 | 1.5 | Weak | |
| Long term Stability Consistency over many months | 3.5 | 3 | Weak |
Toolscopia uses cookies to keep you signed in, remember your preferences, and understand how our reviews get read. Analytics are anonymous. See our privacy policy for the full detail.
Comments 0
Join the discussion and share your perspective.
Sign in to post a comment and reply to other readers.
No comments yet
Be the first to share your perspective on this tool.