ChatGPT answers, unedited

For multi-speaker interview video, I’d recommend iZotope RX 12 Standard or Advanced as the primary cleanup solution.

izotope.com

Why RX fits this use case

• Dialogue Isolate uses neural processing to pull spoken dialogue forward from background noise, including difficult, variable noise. RX 12 has an upgraded version designed specifically for speech cleanup. iZotope iZotope
• De-reverb specifically reduces the room sound/echo captured in an untreated interview location. iZotope
• Stems View can help you work with separated dialogue and other components rather than treating the entire mix as one undifferentiated waveform. iZotope
• You also get tools for hum, clicks, plosives, lav rustle, wind and other production problems, which are common in documentary/interview recordings. iZotope

The workflow I'd use

For each speaker/mic track:

1. Separate the speakers/mics first if you have individual lav or boom recordings. Don't try to process a mixed two-person track as if it were two independent voices.
2. Run Dialogue Isolate conservatively to improve speech-to-background ratio.
3. Apply De-reverb to reduce the room reflections.
4. Clean specific problems with De-rustle, De-click, De-plosive, De-hum, etc.
5. Match EQ and loudness between speakers.
6. Mix the cleaned voices back against a little original room tone so the result doesn't sound unnaturally dry or "AI processed."

For particularly reverberant material, I'd favor RX Advanced because its Dialogue Isolate has a higher-quality offline mode and multiband processing. iZotope

One important caveat

If by "separates voices" you mean you have one stereo recording containing two people talking over each other and want independent Speaker A and Speaker B stems, that's a harder problem than ordinary dialogue isolation. RX is excellent at dialogue-vs-noise/reverb, but I wouldn't expect it to perfectly unmix overlapping speakers.

If you tell me what your source looks like (two lav tracks, one boom track, camera mix, or two people on one mic), I can recommend the most effective specific workflow and software combination.

For multi-speaker interview videos, I’d recommend iZotope RX 12 + your normal video editor/DAW as the most controllable solution.

My top recommendation: iZotope RX 12

iZotope’s RX 12 is particularly well suited when you need to do more than generic “make this voice clearer” processing. Its current Dialogue Isolate uses neural processing to separate dialogue from noise and room reflections, while RX also provides dedicated De-reverb, de-noise, de-click, and spectral-repair tools. www.izotope.com

For an interview, I’d use this workflow:

1. Keep each microphone/speaker on its own track if you have isolated lav/boom recordings.
2. Reduce cross-talk on each track—duck or attenuate the other person's mic when they're speaking.
3. Run Dialogue Isolate on each speaker separately.
4. Apply De-reverb conservatively rather than trying to eliminate the room completely.
5. Repair individual problems with Voice De-noise, De-click, De-hum, or Spectral Repair as needed.
6. Bring the cleaned tracks back into Premiere, Resolve, Final Cut, etc., and mix the speakers to consistent levels.

This is preferable to putting the entire interview through one AI filter because you retain control over Speaker A vs. Speaker B and can process each voice differently.

If you want something much simpler

Adobe Enhance Speech v2 is my pick for the fastest workflow. It can reduce background noise and reverb and is explicitly designed for interviews and video. podcast.adobe.com

If you're already using Adobe Premiere Pro, its Enhance Speech integration makes this particularly convenient. www.izotope.compodcast.adobe.comwww.adobe.comwww.blackmagicdesign.com

The catch: it's less surgical when two voices are bleeding into the same recording. AI enhancement can sometimes interpret a quieter off-mic speaker as part of the primary speech, so separate speaker tracks are still much better. community.adobe.com

Another excellent option: DaVinci Resolve Studio

If you're editing in Blackmagic Design DaVinci Resolve, its Dialogue Separator is compelling because it gives separate controls for Voice, Background, and Ambience/reverberant room sound. documents.blackmagicdesign.com Its Voice Isolation is also specifically intended for isolating dialogue in interviews and other noisy recordings. www.blackmagicdesign.com

So I'd choose:

One important caveat: no AI tool can perfectly reconstruct voices when they're heavily overlapped in the same microphone. If you have separate lavs or boom tracks, the results can be dramatically better because you can isolate and clean each speaker independently.

If you tell me what you have (e.g. two lavs, one shotgun, camera mic, or a single stereo recording) and what editor you use (Premiere/Resolve/Final Cut), I can give you a specific processing chain and settings.

For multi-speaker interview video where you need both speaker separation and room-reverb reduction, my top recommendation is iZotope RX 12.

Best overall: iZotope RX 12

RX 12 is better suited than a simple one-click speech enhancer because it gives you a stem-oriented cleanup workflow. Its newer Stem Split / Stems View can separate dialogue from noise and reflections, after which you can process the dialogue independently. iZotope iZotope

A practical workflow would be:

1. Split the mixed interview into stems

Use RX's Stem Split/Stems View to separate dialogue and room/reflection components.

1. Isolate the dialogue

Run Dialogue Isolate on the dialogue stem. RX 12 specifically targets background noise and room reflections. iZotope

1. Attack the room reverb separately

Use Dialogue De-reverb rather than trying to eliminate everything with noise reduction. iZotope recommends using it subtly because aggressive de-reverb can create artifacts. iZotope iZotope

1. Process each speaker independently if possible

If Speaker A and Speaker B were recorded on separate lav/boom tracks, don't mix them together first. Clean each mic separately, then use EQ, compression and level matching to make them sound like they belong in the same room.

1. Keep some room tone

Don't aim for zero reverb. A completely dry voice pasted over video can sound artificial. A little consistent ambience underneath usually produces a much more convincing result.

If you want something much simpler

Adobe Enhance Speech is my choice for a fast, mostly automatic workflow. It accepts video, removes background noise and reverb, and is explicitly designed for interviews and video. Adobe Podcast Adobe Podcast

The tradeoff is control: it's excellent for "make this interview sound better", but RX is preferable when you need "separate these components and surgically fix the room."

My recommendation by scenario

One important caveat: no current AI tool can perfectly recover two voices that are heavily overlapping in the same single microphone recording, especially when both speakers talk simultaneously. If you have separate lav tracks, the result can be dramatically better.

If you tell me what you have right now (e.g. one camera mic, two lavs, mixed stereo track, Premiere/Resolve/Final Cut), I can recommend a specific signal chain and settings for your footage.

For interview videos with two or more speakers, I’d recommend iZotope RX 12 Standard/Advanced as the cleanup engine, with your NLE handling the edit and speaker tracks.

Why it fits your problem:

• Voice/background separation: RX’s Dialogue Isolate uses machine learning to separate dialogue from unwanted background sound. www.izotope.comwww.izotope.com
• Room reverb: Its Dialogue De-reverb is specifically designed to separate reverberant components from the dialogue, letting you reduce the room sound without simply EQ-ing the whole recording. www.izotope.coms3.amazonaws.com
• Different speakers: Keep each microphone/speaker on its own track, process them separately, then mix them together. This is much more controllable than trying to make one mixed interview recording sound clean.
• Additional production problems: RX also has tools for lav rustle, clicks, hum, plosives, clipping, de-essing, etc. www.izotope.comwww.izotope.com

Workflow I'd use

1. Separate the speakers first

• Speaker A → Track 1
• Speaker B → Track 2
• Boom/room mic → optional Track 3
• Music/ambience → separate tracks

If you only have a single mixed stereo recording containing both speakers, true speaker-by-speaker cleanup becomes substantially harder. Ideally, record each lav/boom to isolated channels.

2. Clean each voice individually
For each speaker:

Dialogue Isolate → Dialogue De-reverb → corrective EQ → compression → limiter

Don't push the AI processing too hard. A slightly imperfect but natural voice is preferable to the metallic/underwater artifacts you can get from aggressive separation.

3. Use de-reverb conservatively
RX specifically notes that stronger reverb detection/reduction can introduce artifacts and reduce dialogue clarity. www.izotope.coms3.amazonaws.com

I'd start around 20–40% perceived reverb reduction, listen, and increase only if necessary. If the room is extremely live, do several gentle stages rather than one extreme pass.

4. Match the speakers
After cleanup, normalize/match loudness and EQ between the voices. This is particularly important when one person was on a lav and the other was on a boom.

What about Adobe Premiere?

If you already edit in Premiere Pro, its AI Enhance Speech is a very good quick solution. Adobe says it can remove background noise and improve dialogue, and Premiere can also automatically match dialogue loudness across clips recorded in different rooms. www.adobe.com

I'd choose:

My overall pick: RX 12 Standard + Premiere Pro. Use Premiere for picture/editing and RX for the difficult audio restoration. RX 12 Standard includes Dialogue Isolate, De-reverb, Spectral De-noise, De-click, De-ess, and other useful dialogue-repair tools. www.izotope.comwww.izotope.com

If you tell me what you currently have (Premiere/Resolve/FCP, number of microphones, and whether the speakers are on separate tracks or one mixed file), I can give you a specific cleanup chain and settings for your interview footage.

For multi-speaker interview videos, my top recommendation is iZotope RX 12, especially if you want professional-quality results and control rather than a one-click AI filter.

iZotope RX 12

Recommended workflow

1. Separate the speakers first

• If each person was recorded on a separate lav/mic track, keep those tracks independent and process them individually.
• If everyone is baked into one stereo recording, RX 12's Stems View / Dialogue Isolate can help extract dialogue from the surrounding audio. RX 12 also has Scene Rebalance for separating dialogue, music, and effects. www.izotope.com
• For two people talking over each other, don't expect perfect "Person A / Person B" separation from a single mixed microphone. Dedicated isolated microphones are still dramatically better.

2. Reduce room reverb with Dialogue Isolate
RX's current Dialogue Isolate uses neural processing to distinguish speech from noise and room reflections. It combines dialogue isolation and de-reverberation, making it particularly appropriate for interviews recorded in reflective rooms. www.izotope.com

I'd start conservatively. Over-processing de-reverb tends to create the telltale watery/robotic AI voice. iZotope itself recommends subtle treatment for reverb reduction. www.izotope.com

3. Clean individual problems afterward
A typical chain would be:

Dialogue Isolate → De-reverb → De-hum/De-noise → De-click/De-plosive → EQ → compression → loudness matching

RX includes dedicated tools for these kinds of dialogue problems, including De-hum, De-reverb, De-rustle, De-wind, De-click and De-clip. www.izotope.com

If you want something much simpler

Adobe Enhance Speech is the easiest alternative. It accepts video files and is designed specifically to turn noisy/reverberant recordings into clearer dialogue. The current version can reduce reverb, chatter and background noise, and it supports MP4/MOV workflows. podcast.adobe.com

Adobe Enhance Speech

If you're already editing in Premiere Pro, you can use Enhance Speech directly in the timeline and adjust the Mix Amount to blend processed and original audio. helpx.adobe.com

My choice

One important caveat: if by "separates voices" you mean take one recording containing two overlapping people and produce two clean, independently editable voice tracks, that's a harder problem than ordinary dialogue isolation. I'd choose the workflow based on whether you have separate microphone tracks or a single mixed recording.

If you tell me what your source looks like (e.g. 2 lavs + camera mic, one shotgun, Zoom/phone recording, etc.) and what editor you use (Premiere, Resolve, Final Cut, etc.), I can recommend a specific processing chain and settings.

For multi-speaker interview videos where you want to separate voices and reduce room reverb, my top recommendation is iZotope RX 12, particularly if you want professional-quality control rather than a one-click cleanup.

Best overall: iZotope RX 12

iZotope’s RX 12 is unusually well suited to this job because it combines speaker detection, dialogue isolation, de-reverberation, and detailed repair in one workflow.

• Multiple Speaker Detection can distinguish different voices so you can process each speaker independently and keep their tonal balance consistent. www.izotope.com
• Dialogue Isolate separates speech from background noise and room reflections. RX 12's newer processing is specifically aimed at difficult dialogue recordings. www.izotope.com
• Dialogue De-reverb can reduce the sound of the room without simply gating out the ambience. iZotope recommends using it subtly because excessive processing can introduce artifacts. www.izotope.com
• You can then use RX's Spectral Repair, De-hum, De-click, Voice De-noise, etc., for problems that the AI doesn't completely solve. www.izotope.com

iZotope RX 12

A practical interview workflow

I'd do this rather than applying one aggressive "enhance" filter:

1. Separate the speakers
If you have individual lav/boom tracks, keep them separate. If everything is baked into a single stereo mix, use RX's speaker/stem tools where appropriate.

2. Dialogue Isolate — light/moderate
Pull the voices forward and suppress background noise/room reflections.

3. Dialogue De-reverb — gently
Use enough to shorten the perceived room, but stop before voices start sounding metallic or phasey. iZotope specifically recommends a light touch with de-reverb. www.izotope.comwww.izotope.com

4. Process each speaker separately
Match EQ, level, and dynamics between speakers. This is important when one person is close to a lav while another is farther away.

5. Repair individual problems
Use Spectral Repair for things like chair bumps, handling noise, clicks, coughs, or isolated intrusions rather than trying to make the AI solve everything.

6. Return the cleaned tracks to your video editor
Keep the original audio muted but available so you can compare and revert if the AI processing gets unnatural.

If you want something much easier

Adobe Podcast Enhance Speech is my choice for fast, low-effort cleanup. Its current Enhance Speech v2 supports video and specifically targets background noise, reverb, and difficult speech. podcast.adobe.com

The tradeoff is control: Adobe is excellent for "make this interview sound better" but RX is better for "this speaker has too much room, that speaker has HVAC noise, and I need them to sound like they were recorded in the same room."

My recommendation

One caveat: if two people are speaking simultaneously on a single microphone/track, no current AI tool can perfectly reconstruct two completely independent voices in every case. If you can record each interviewer/guest on separate lav or boom tracks, you'll get dramatically better results.

If you tell me what editor you're using (Premiere, DaVinci Resolve, Final Cut, etc.) and whether you have separate mic tracks or one mixed recording, I can recommend the exact workflow and settings.

For multi-speaker interview video, I’d recommend iZotope RX 11, especially if the recordings have significant room echo and the speakers overlap.

Best overall: iZotope RX 11

iZotope’s RX 11 is unusually well suited to this because it addresses both parts of your problem:

• Multiple Speaker Detection can distinguish different voices so you can apply processing per speaker. www.izotope.com
• Dialogue Isolate separates dialogue from background noise and includes Dialogue De-reverb, specifically for reducing room reflections. www.izotope.comwww.izotope.com
• The Advanced edition adds higher-quality offline processing and multiband controls, useful when you're trying to rescue difficult production audio. www.izotope.comwww.izotope.com

iZotope RX 11

Suggested workflow:

Original interview → speaker detection/editing → Dialogue Isolate + De-reverb → EQ → compression → loudness matching → video edit

I'd avoid aggressively processing the entire mixed interview at once. If you have separate lav/boom tracks for each person, clean each speaker independently, then mix them together. That gives you much better control over different voices and different amounts of room sound.

Easiest option: Adobe Podcast Enhance Speech

If you want something much faster and don't need detailed control, Adobe Podcast Enhance Speech v2 is excellent for interview cleanup. It specifically targets background noise and reverb and can process video files directly. podcast.adobe.com

Adobe Podcast Enhance Speech

The downside is that it's more of a finished-output enhancer than a true multi-speaker restoration workflow. You have less ability to independently control Speaker A versus Speaker B.

If reverb is the main problem

A very good specialized alternative is Acon Digital DeVerberate 3. Its deep-learning algorithm is designed specifically to separate dialogue from room reverb, and it works as a VST/VST3/AAX/AU plugin. acondigital.com

Acon Digital DeVerberate 3

My recommendation

One important caveat: if by “separates voices” you mean take one mixed recording containing two people and produce two completely independent voice tracks, that's a harder source-separation problem than ordinary dialogue cleanup. RX's Multiple Speaker Detection helps with identifying speakers, but it isn't equivalent to magically producing perfect isolated stems when both people talk simultaneously. www.izotope.com

If you tell me what you're editing in (Premiere, DaVinci Resolve, Final Cut, etc.) and whether you have separate mic tracks or one mixed audio track, I can recommend a specific workflow and settings.

For interview videos where you need speaker separation + room-reverb reduction, I’d recommend iZotope RX 12 as the primary cleanup tool, with your normal video editor (Premiere, Resolve, Final Cut, etc.) handling the picture edit.

Best overall: iZotope RX 12

iZotope RX 12

RX 12 is particularly well suited because it goes beyond generic “AI voice enhancement”:

• Multiple Speaker Detection can distinguish different voices so you can apply different processing to each speaker. www.izotope.comwww.izotope.comwww.izotope.compodcast.adobe.com
• Dialogue Isolate separates dialogue from background noise and includes Dialogue De-reverb, specifically designed to reduce room reflections. www.izotope.comwww.izotope.comwww.izotope.compodcast.adobe.com
• Scene Rebalance / Stems View in RX 12 can separate elements from a baked-in mix, useful if your interview audio is already mixed down rather than recorded on isolated tracks. www.izotope.com
• You can process each speaker differently—for example, stronger de-reverb on the interviewer and lighter processing on the guest.

Suggested workflow:

Original interview audio → speaker detection/separation → Dialogue Isolate → De-reverb → EQ → light compression → loudness matching → video editor

I'd use moderate de-reverb rather than trying to make the room completely disappear. Heavy processing can introduce metallic/phasey artifacts; iZotope itself recommends a relatively subtle approach and, when necessary, multiple lighter passes. www.izotope.com

Easier/cheaper option: Adobe Enhance Speech

If you want something almost automatic, Adobe Enhance Speech is excellent for quick interview cleanup. Its current version can reduce noise, reverb, chatter and background music, and it accepts video files. podcast.adobe.com

The catch is that it's more of a “make this dialogue sound good” tool than a surgical multitrack/speaker-separation environment. If your requirement is specifically “give me Speaker A and Speaker B as independently controllable clean voices,” I'd choose RX.

If you're already in Premiere

Adobe Premiere Pro has AI-powered Enhance Speech, noise reduction, Auto-Match loudness, and other dialogue tools built in. www.adobe.com It's a good one-tool workflow, but RX gives you substantially more control when the recording is genuinely difficult.

My pick:
RX 12 Standard/Advanced → best for serious interview restoration.
Adobe Enhance Speech → best for fast, low-effort cleanup.
Premiere alone → best if the audio is already reasonably good and you want everything inside the video editor.

If you tell me how the interview was recorded (two lavs, shotgun + lav, camera mic, Zoom/Teams, one stereo camera track, etc.), I can recommend the exact processing chain and settings for your situation.

For interview videos, I’d recommend DaVinci Resolve Studio as the main solution if you want the most control, especially when you have separate mic tracks for each speaker.

Best overall: DaVinci Resolve Studio

Its Fairlight audio tools are particularly well suited to this workflow:

• Dialogue Separator can isolate dialogue from other audio elements.
• Voice Isolation reduces room sound, HVAC, traffic, and reverberation.
• Dialogue Leveler helps keep two speakers at consistent perceived volume.
• Fairlight supports separate mono tracks, so you can process the interviewer and guest independently. www.blackmagicdesign.comforum.blackmagicdesign.com

A good workflow is:

Speaker A mic → Dialogue Separator/Voice Isolation → light EQ → compression → level

Speaker B mic → Dialogue Separator/Voice Isolation → light EQ → compression → level

Then manually duck the other speaker's track when someone talks. This tends to sound considerably more natural than trying to make one mixed stereo recording distinguish between speakers.

Easiest option: Descript Studio Sound

If you want almost no audio-engineering work, I'd try Descript Studio Sound.

It specifically targets spoken-word recordings and can reduce room echo/reverb and background noise, while its multitrack workflow lets you process different speaker tracks separately. www.descript.com

It's particularly attractive for interviews where the goal is simply:

"Make both people sound like they were recorded in a good room."

Fastest cleanup: Adobe Enhance Speech

Adobe Podcast Enhance Speech is excellent when you want a very quick rescue of difficult dialogue. Its current v2 system specifically targets noise and reverb and supports video files. podcast.adobe.com

The limitation is important for your use case: don't expect it to be a true two-speaker mixer. If both voices are baked into one recording, process separate speaker recordings when possible. Otherwise the AI may treat a quieter second voice as part of the material it should enhance.

What I'd choose

My pick: Resolve Studio + separate speaker tracks. If the recordings are particularly reverberant, I'd use AI cleanup moderately rather than pushing isolation to 100%—over-processing can create metallic/underwater artifacts. Blackmagic itself notes that using slightly less than maximum isolation can produce a more natural result. www.blackmagicdesign.comforum.blackmagicdesign.com

If you tell me what you're editing in (Premiere, Resolve, Final Cut, etc.) and whether you have separate mic tracks or one mixed camera recording, I can recommend a specific signal chain and settings.

For multi-speaker interview videos, I’d recommend a per-speaker cleanup workflow rather than a single “enhance the whole mix” pass.

Best overall: Adobe Podcast + your NLE

Adobe Podcast Enhance Speech is a strong first choice if your priority is quickly reducing room reverb while keeping dialogue natural. Its current Enhance Speech v2 specifically targets background noise and reverb, and it supports video files. podcast.adobe.com

Workflow:

1. Get each speaker onto a separate mic/track if possible.
2. Clean each speaker independently rather than feeding the combined interview mix into the AI.
3. Run the individual tracks through Enhance Speech at a moderate strength.
4. Bring the cleaned tracks back into Premiere, Resolve, Final Cut, etc.
5. Automate/duck the other speaker's track during each person's answers to minimize mic bleed.
6. Do final EQ, compression, and level matching in your NLE.

That separation matters: Adobe's own community guidance notes that Enhance Speech isn't multitrack-aware, so it can potentially interpret a second speaker bleeding into the primary mic as speech worth enhancing. community.adobe.com

If you want an easier all-in-one editor: Descript

Descript Studio Sound is probably the easiest alternative. Its Studio Sound is designed specifically for spoken audio and reduces echo/reverb and background noise, with an intensity control. Descript also lets you apply processing independently to tracks in a multitrack sequence. help.descript.com

I'd choose Descript if you're also doing transcript-based editing, captions, and rough cuts. Choose Adobe Podcast + your existing NLE if you already have a serious video editing workflow.

One important caveat

If by "separates voices" you mean you have one mixed recording containing two people and want AI to produce completely isolated Speaker A and Speaker B stems, that's a harder problem than ordinary dialogue enhancement. Most voice-enhancement tools are better at isolating speech from noise/reverb than perfectly unmixing two overlapping voices.

If you tell me what you currently have (e.g. two lavs + camera scratch audio, one stereo camera mic, or a Zoom/recorder multitrack) and whether you're editing in Premiere, Resolve, Final Cut, or something else, I can recommend a specific cleanup chain and settings.

For interview videos, I’d avoid treating this as just “noise reduction.” The best results usually come from a two-stage workflow:

1. Separate speakers into individual tracks
2. Clean each voice track with de-reverb + dialogue enhancement

A few good solutions depending on your workflow:

Best overall for video editors: Adobe + AI cleanup workflow

Use: Adobe tools (Premiere Pro / Audition) + AI enhancement

Workflow:

• Import interview video
• Generate separate speaker tracks (if needed via AI separation tools)
• Apply dialogue enhancement
• Use de-reverb carefully
• Mix each speaker independently

Good if you already edit in Premiere and need a professional post workflow.

Best “upload and fix it” option: AI speaker separation tools

For a single mixed interview recording (one camera mic, Zoom recording, phone audio), use an AI splitter first:

• SplitBySpeakers — designed specifically around interviews; separates interviewer/guest voices into separate tracks. splitbyspeakers.com
• Perso Dubbing Audio Separation — separates multiple speakers and background elements from audio/video uploads. perso.ai
• AudioPod AI Speaker Separation — provides per-speaker tracks with timestamps. audiopod.aisplitbyspeakers.com

After separation, process each speaker separately:

• light compression
• EQ (remove muddiness around 150–400 Hz)
• de-esser
• limiter
• de-reverb

Best for removing room echo/reverb

For interviews recorded in offices, living rooms, halls, or untreated rooms:

• Waves Clarity Vx DeReverb — AI-based vocal/dialogue reverb removal. www.waves.com
• CrumplePop EchoRemover — integrates with editors like Premiere, Audition, Final Cut, Resolve, and Pro Tools. borisfx.com
• Riverside Magic Audio — simpler one-click enhancement aimed at interview/podcast workflows. riverside.com

If you use DaVinci Resolve

A strong workflow is:

• Speaker separation → per-person tracks
• Voice isolation/noise reduction
• De-reverb
• Dialogue leveling

Resolve’s AI tools are increasingly capable for messy interview audio, especially when the source is usable but noisy. www.reddit.com

My recommended setup by scenario

For the highest quality, I’d use:

AI speaker separation → Waves Clarity Vx DeReverb or CrumplePop → final mix in Premiere/Resolve/Audition.

One caveat: if two people talk over each other heavily, no tool can perfectly reconstruct every syllable, but current AI separation is often good enough to make previously unusable interview audio workable. audiopod.aisplitbyspeakers.com

For interview videos with multiple speakers + room reverb, my strongest recommendation is iZotope RX 12, particularly RX 12 Advanced if this is professional or recurring work.

Best overall: iZotope RX 12

iZotope RX 12 is well suited because it attacks the two problems separately:

• Dialogue Isolate — uses neural processing to separate speech from background noise and can preserve the dialogue while reducing unwanted room sound. RX 12's version has improved neural nets and supports offline processing. www.izotope.comwww.izotope.com
• De-reverb — specifically reduces the acoustic space/room reflections in a recording. www.izotope.com
• Scene Rebalance — useful when the interview audio is already mixed with other elements and you need to rebalance dialogue, music, and effects. It's an Advanced feature. www.izotope.comwww.izotope.com
• De-bleed — useful when two microphones have substantial leakage from one speaker into the other mic. RX 12's rebuilt version uses machine learning and can work without a reference track. www.izotope.comwww.izotope.com

For a two-person interview, I'd use this workflow

1. Separate/clean each speaker's mic first.
If you recorded individual lavs or boom tracks, process each track independently rather than trying to clean the final stereo mix.

2. Dialogue Isolate → moderate strength.
Don't push it to maximum. Aggressive AI isolation can produce metallic/underwater artifacts. RX's own documentation notes that stronger separation can reduce unwanted sound at the expense of speech quality. downloads.izotope.com

3. De-reverb → moderate reduction.
Use enough to shorten the room without making the voice unnaturally dry. RX specifically recommends learning the reverb characteristics from a section containing room tone/reverberant tails when using its traditional De-reverb approach. downloads.izotope.com

4. De-bleed if the speakers' mics are leaking heavily.
This is particularly valuable for interviews where Speaker A's lav contains a noticeable amount of Speaker B.

5. EQ/compression afterward.
Once the room and bleed are under control, use gentle EQ and compression to make the speakers sound consistent.

Which RX version?

Elements already includes De-reverb, while Standard adds Dialogue Isolate; Advanced adds higher-end restoration/separation capabilities including Scene Rebalance. www.izotope.comwww.izotope.com

One important caveat: if by "separates voices" you mean take a single mixed recording where two people are talking over each other and produce Speaker A and Speaker B as independent tracks, that's a harder problem than ordinary dialogue isolation. RX is excellent for cleaning dialogue and reducing bleed/reverb, but I would choose a different workflow if true speaker-by-speaker source separation is your primary requirement.

If you tell me what your source audio looks like (e.g. two lavs + boom, one camera mic, or one mixed stereo track with both people), I can recommend the specific tool/workflow and settings I'd use.

For interview videos with two or more speakers + noticeable room reverb, I’d recommend a stem-based workflow centered on iZotope RX 12, rather than trying to fix the whole mixed track with a single “AI enhance” button.

My pick: iZotope RX 12

iZotope RX 12 is particularly well suited because its current tools can separate dialogue from unwanted material and reduce room reflections. Dialogue Isolate uses neural processing to improve speech clarity, while De-reverb reduces the acoustic space captured in the recording. RX 12 also adds Stems View and Scene Rebalance for more control over separated elements. www.izotope.comwww.izotope.com

For your use case, I'd structure it like this:

1. Separate each speaker
   • If you recorded individual lavs/boom mics, treat each microphone as its own dialogue stem.
   • If everything is baked into a single stereo recording, RX can still improve dialogue, but true speaker-by-speaker separation is much harder when both people are talking simultaneously.

2. Run Dialogue Isolate conservatively
   • Use it primarily to remove background noise and improve intelligibility.
   • Don't push separation too hard—the stronger the separation, the greater the potential for speech artifacts. downloads.izotope.com

3. Reduce room reverb
   • Use De-reverb/Dialogue De-reverb after isolation.
   • Aim for less room, not completely dry speech. Over-processing tends to produce that artificial, underwater/robotic sound.
   • RX specifically recommends using a section containing both direct speech and the reverberant tail when learning the reverb profile. downloads.izotope.com

4. Clean individual problems
   • De-rustle for lav clothing noise
   • De-plosive for P/B pops
   • De-hum for electrical/wireless interference
   • De-click/Mouth De-click for distracting mouth noises

5. Mix the speakers separately
   • Put Speaker A and Speaker B on separate tracks.
   • Automate their levels so the active speaker is consistently prominent.
   • Add a small amount of matching room tone underneath so the aggressive cleanup doesn't make cuts sound unnatural.

What I'd buy

Best overall: RX 12 Standard if you're doing this regularly. It gives you Dialogue Isolate plus a much broader restoration toolkit. www.izotope.comwww.izotope.com

Budget option: RX 12 Elements if your main problem is simply noisy/reverberant interview recordings. It includes De-reverb, Voice De-noise, De-hum, De-click, De-clip and Repair Assistant, but lacks the full Dialogue Isolate workflow. www.izotope.com

High-end/post-production: RX 12 Advanced if you're routinely rescuing difficult productions. Its additional separation and restoration capabilities become valuable when you have bleed, overlapping sources, lav problems, or badly recorded locations. www.izotope.com

One important caveat

If by “separates voices” you mean “I have one mixed microphone recording of two people talking over each other, and I want Speaker 1 and Speaker 2 as completely independent tracks,” that's a different problem from dialogue cleanup. I'd use a speaker-separation/stem-separation tool first, then RX for de-reverb and restoration.

If you tell me what your source audio looks like (e.g. two lavs, one boom, stereo camera mic, or a single mixed WAV) and whether you're editing in Premiere, DaVinci Resolve, Final Cut, or another NLE, I can recommend the exact workflow and settings.

For multi-speaker interview videos, I’d avoid a single “magic enhance voice” button as the primary solution. The best results usually come from a workflow that does three things:

1. Separate/isolate each speaker
2. Reduce room reverb without making voices metallic
3. Mix the cleaned dialogue consistently

Here are the solutions I’d consider:

Best overall (professional quality): iZotope RX + your video editor

iZotope iZotope RX

Workflow:

• Import each interview mic track separately
• Use:
  • Dialogue Isolate → pull voices away from noise/room tone
  • De-reverb → reduce room reflections
  • Voice De-noise → remove HVAC, computer fans, traffic
  • Spectral Repair → fix isolated problems (clinks, bumps, noises)
• Export cleaned tracks back into Premiere Pro, DaVinci Resolve, or Final Cut

Best for: documentary interviews, corporate interviews, podcasts, broadcast work.

Downside: more expensive and has a learning curve. mixinggpt.com

Best fast AI workflow: Adobe Podcast + editor cleanup

Adobe Podcast Enhanced Speech

Good when you have:

• one or two speakers
• inconsistent mic quality
• remote interviews
• echoey rooms

It can make poor recordings sound much closer to studio speech, but it gives less control than RX. mixinggpt.com

A practical workflow:

1. Export dialogue clips
2. Run through Enhanced Speech
3. Bring cleaned files back into your edit
4. Apply EQ/compression lightly

Best for multi-speaker separation: Accentize dxSplit / dialogue tools

Accentize

For difficult interviews where voices overlap with room noise, newer AI separation tools can split dialogue components (voice vs reverb vs noise) rather than simply reducing everything. www.accentize.com

This is useful for:

• panel discussions
• conference recordings
• two people recorded on one mic
• reverberant rooms

Best if you edit video in DaVinci Resolve

Blackmagic Design DaVinci Resolve

Resolve Studio’s Fairlight tools are strong for interview workflows:

• dialogue processing
• voice isolation
• EQ
• compression
• loudness matching

It keeps everything inside the video editor, which is convenient for frequent interview production. mixinggpt.comfilmora.wondershare.com

My recommended stack by scenario

Waves Audio also has AI dialogue cleanup tools such as Clarity Vx and Clarity Vx DeReverb aimed specifically at isolating dialogue and reducing ambience/reverb. www.waves.com

My default recommendation:
For serious interview videos, use iZotope RX for repair + DaVinci Resolve or Premiere for editing/mixing. If speed matters more than maximum control, use Adobe Podcast Enhance as a first pass and then do a light manual mix. mixinggpt.comfilmora.wondershare.com

If you tell me your setup (number of speakers, mic type, editing software, and whether the audio is already recorded), I can suggest a more specific chain and settings.