I’ve been testing Runway AI for video and image generation and my results have been pretty mixed. Some outputs look amazing, but others are low quality, glitchy, or way off from my prompts. I’m trying to decide if it’s worth paying for a subscription or if I should switch to another AI tool. Can anyone share real user reviews, tips for better results, or alternatives that might be more reliable?
I’ve had a pretty similar ride with Runway, so here’s the blunt take after a few months of on and off use.
Where it’s strong:
-
Quick concept tests
• If you want to see “does this idea work as a vibe” for a short, it helps.
• For example, I did a 10 second cyberpunk alley shot from a static image. Took like 10 minutes to iterate to something usable for a storyboard. -
Stylized, artsy stuff
• It seems to handle abstract, painterly, or “dreamy” looks better than strict realism.
• If your project leans into surreal or experimental visuals, the glitches feel less wrong and more “style”. -
Short clips, small screens
• Stuff up to about 5 to 8 seconds at social media resolution looks okay.
• Once I pushed to 1080p and did longer sequences, artifacts and weird morphs got harder to ignore.
Where it struggles:
-
Prompt accuracy
• Detailed prompts still give hit or miss results.
• Complex actions like “person turns, looks into camera, smiles, then walks away” usually fall apart somewhere in the motion.
• Faces shift across frames, outfits change, hands go weird. -
Consistency across shots
• Keeping the same character, same outfit, same face across multiple clips is tough.
• I tried to do a 30 second sequence broken into 5 clips. Each clip looked like a cousin of the previous one, not the same person. -
Motion quality
• Fast movement often gets smeared.
• Limbs and objects deform if you push longer durations.
• Camera moves like pans or dolly feels floaty and fake unless you keep them slow and simple.
Things that helped me get better results:
-
Use strong visual references
• Upload a base image instead of relying on text alone, especially for characters.
• I got better consistency feeding in a portrait and reusing it for every clip. -
Keep prompts simple and strict
• Example: “medium shot of a woman in a red hoodie, neutral expression, standing in a dim alley, slow camera push in”
• Avoid stacking too many actions. Break scenes into small beats. -
Shorter clips, then edit
• Work in 3 to 6 second chunks.
• Stitch, cut, add sound and effects in a normal editor.
• Treat Runway like a shot generator, not a full scene maker. -
Expect to iterate a lot
• For each decent clip I used, I had 5 to 10 fails.
• Plan time and credits for multiple runs of the same idea.
When it feels worth paying for:
• Previs, animatics, concept pitches.
• Music videos or experimental stuff where weirdness is fine.
• Background plates or B roll where you blur or composite it.
When it does not feel worth it:
• Client work that needs strict control and consistency.
• Long form narrative with recurring characters.
• Anything where you need precise action or lip sync.
If you are on the fence about paying, I’d do this:
- Take one real project idea.
- Try to produce a 15 to 30 second “finished” piece using only Runway and your editor.
- Track:
• How many generations do you run
• How many hours you spend
• How often you compromise your idea for what the tool gives
If after that test you feel like you fought the tool more than it helped, pause the subscription and recheck in a few months. If it saved you time vs manual animation or stock hunting, then it is worth keeping as a tool in your kit.
TLDR: It works ok as a fast concept and experimental video helper. It does not replace a real pipeline for anything that needs control and reliability.
Yeah, “mixed” is the exact word I’d use too.
I’m broadly in the same camp as @sognonotturno, but I’d split it a bit differently:
1. Ask what job you actually need Runway to do
Forget “is it good?” and ask: what specific pain are you trying to solve?
- Need ideas fast for shots / mood / angles?
Runway is decent. Not magical, but it beats staring at a blank timeline. - Need final, dependable shots for a real project?
That’s where it starts to feel like gambling with your time and sanity.
If your real use case is: “I want pixel-level control, consistent characters, and precise motion,” I’d honestly say it’s not worth a subscription yet, unless you treat the failures as “R&D time” you’re ok burning.
2. The “quality whiplash” is part of the product right now
You’re not imagining the wild swings:
- One run: cinematic, sharp, moody, almost looks like a real production still.
- Next run with nearly the same prompt: melted faces, stiff motion, random object flicker.
That inconsistency is the key thing to decide on. Not “can it make something cool?” because clearly it can. The question is: can it make something cool on demand when you actually need it?
If you find yourself thinking “this is fun when I’m just playing, but terrifying when I’m on a deadline,” that’s your answer.
3. Where I’d disagree a bit with @sognonotturno
They’re right that it shines for surreal / experimental stuff. I actually think it’s worse there if you already have a strong artsy vision. The “cool glitches” often feel like the same type of glitch over and over: warped limbs, jittery morphs, flickery frames. After a while it all has that “AI video” smell.
Weirdly, I’ve had slightly better mileage when I keep things very grounded and simple:
- Static or almost static camera
- Subtle motion only
- Clean lighting, no crazy contrast
It’s boring, yeah, but it tends to produce more “usable” clips that you can then spice up in post. So I’d say: don’t fully lean into chaos unless you actually want it to look AI generated.
4. A more brutal way to test if it’s worth paying
Instead of just doing a 15–30s test like they suggested, do this:
- Take a scene or shot you already know how to make with your normal tools.
- Recreate it in Runway as close as possible.
- Then compare:
- Time spent
- Level of compromise from your original idea
- How much “cleanup” you still need (sound, compositing, masking, etc.)
If Runway did not:
- reduce your total hours
- or significantly boost the visual impact
then it’s basically a toy for you right now, not a tool.
5. When I think the sub actually makes sense
Keeping it real:
- You’re in heavy preproduction / moodboarding mode a lot.
- You do pitches where “wow factor” matters more than technical perfection.
- You’re cool with wasting 60 to 70% of outputs because the 30% you keep sparks new ideas you would not have thought of.
If you’re in delivery-focused work (clients, tight timelines, strict look), I’d lean no for now and maybe just hop on monthly for short bursts when you know you’ll use it.
6. Honest answer to your implied question
If your main thought after testing is “this is cool, but I can’t trust it,” that’s the correct instinct. Trust is the product here, not just visuals.
So if you’re on the fence:
Pause the sub, keep your projects list handy, and revisit in a few months. The tech is evolving stupidly fast. The feeling of “almost there but not quite” you have now is very likely to shift, but you don’t need to pay monthly just to wait for that moment.
Short version: keep Runway if it fits one of these very specific lanes, drop it for anything else.
Where Runway AI is actually worth it
Pros:
- Fast “what if?” exploration for shots and concepts
- Great for pitch decks, proof of concept, and mood videos
- Occasionally produces frames that look shockingly high end
- Easy web UI compared to some local models
Cons:
- Inconsistent output quality, exactly like you described
- Very weak at character consistency and precise control
- Iteration can feel like a slot machine that eats your time
- Not ideal for deadline driven, client facing work
I line up with a lot of what @sognonotturno said, but I’m slightly more optimistic if you treat Runway as a side tool, not the core of your workflow.
A different way to test it:
- Pick one real project you’re doing this month.
- Decide one job for Runway:
- generate B‑roll style abstraction over dialogue, or
- create only backgrounds you will composite over, or
- make rough animatics instead of storyboard frames.
- Cap yourself to a hard time budget, like “2 hours per project with Runway.”
If, inside that constraint, Runway AI:
- gives you usable clips or boards that actually make it into the edit
- or meaningfully upgrades the look of something you’d ship anyway
then the subscription is earning its keep. If not, it’s currently just an experimental toy for your use case.
Where I slightly disagree with the other replies: I wouldn’t only keep Runway for surreal or glitchy stuff. Used surgically for backgrounds, abstract textures, or concept tests, it can quietly save time even on grounded projects, as long as you accept it will not behave like a dependable DCC tool.
If your gut says “I can’t trust this for anything critical,” follow that and pause the sub. Revisit when you either have a project that fits those narrow lanes or the models clearly level up in consistency.